perm filename LSPMAN.RPG[UP,DOC]1 blob sn#323802 filedate 1977-12-16 generic text, type C, neo UTF8
COMMENT ⊗   VALID 00235 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00036 00002	**DRAFT**                         The Language                         **DRAFT**
C00038 00003	**DRAFT**                     General Information                      **DRAFT**
C00043 00004	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00046 00005	**DRAFT**                     General Information                      **DRAFT**
C00050 00006	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00054 00007	**DRAFT**                     General Information                      **DRAFT**
C00057 00008	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00058 00009	**DRAFT**                         Data Objects                         **DRAFT**
C00063 00010	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00068 00011	**DRAFT**                         Data Objects                         **DRAFT**
C00073 00012	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00077 00013	**DRAFT**                         Data Objects                         **DRAFT**
C00079 00014	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00080 00015	**DRAFT**                  The Basic Actions of LISP                   **DRAFT**
C00085 00016	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00089 00017	**DRAFT**                  The Basic Actions of LISP                   **DRAFT**
C00094 00018	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00097 00019	**DRAFT**                  The Basic Actions of LISP                   **DRAFT**
C00101 00020	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00105 00021	**DRAFT**                  The Basic Actions of LISP                   **DRAFT**
C00109 00022	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00112 00023	**DRAFT**                  The Basic Actions of LISP                   **DRAFT**
C00116 00024	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00119 00025	**DRAFT**                  The Basic Actions of LISP                   **DRAFT**
C00121 00026	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00124 00027	**DRAFT**                    Function Descriptions                     **DRAFT**
C00129 00028	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00130 00029	**DRAFT**                          Predicates                          **DRAFT**
C00133 00030	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00136 00031	**DRAFT**                          Predicates                          **DRAFT**
C00140 00032	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00143 00033	**DRAFT**                          Predicates                          **DRAFT**
C00144 00034	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00145 00035	**DRAFT**                        The Evaluator                         **DRAFT**
C00148 00036	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00152 00037	**DRAFT**                        The Evaluator                         **DRAFT**
C00156 00038	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00159 00039	**DRAFT**                        The Evaluator                         **DRAFT**
C00163 00040	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00166 00041	**DRAFT**                        The Evaluator                         **DRAFT**
C00169 00042	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00171 00043	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00174 00044	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00177 00045	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00179 00046	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00182 00047	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00185 00048	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00188 00049	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00191 00050	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00192 00051	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00195 00052	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00199 00053	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00203 00054	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00206 00055	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00209 00056	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00212 00057	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00215 00058	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00218 00059	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00222 00060	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00224 00061	**DRAFT**                 Manipulating List Structure                  **DRAFT**
C00228 00062	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00231 00063	**DRAFT**                       Flow of Control                        **DRAFT**
C00235 00064	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00238 00065	**DRAFT**                       Flow of Control                        **DRAFT**
C00242 00066	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00246 00067	**DRAFT**                       Flow of Control                        **DRAFT**
C00250 00068	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00255 00069	**DRAFT**                       Flow of Control                        **DRAFT**
C00258 00070	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00261 00071	**DRAFT**                       Flow of Control                        **DRAFT**
C00264 00072	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00267 00073	**DRAFT**                       Flow of Control                        **DRAFT**
C00269 00074	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00273 00075	**DRAFT**                       Flow of Control                        **DRAFT**
C00276 00076	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00277 00077	**DRAFT**                        Atomic Symbols                        **DRAFT**
C00281 00078	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00284 00079	**DRAFT**                        Atomic Symbols                        **DRAFT**
C00286 00080	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00290 00081	**DRAFT**                        Atomic Symbols                        **DRAFT**
C00293 00082	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00296 00083	**DRAFT**                        Atomic Symbols                        **DRAFT**
C00299 00084	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00302 00085	**DRAFT**                        Atomic Symbols                        **DRAFT**
C00304 00086	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00308 00087	**DRAFT**                        Atomic Symbols                        **DRAFT**
C00312 00088	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00313 00089	**DRAFT**                        Atomic Symbols                        **DRAFT**
C00316 00090	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00320 00091	**DRAFT**                        Atomic Symbols                        **DRAFT**
C00323 00092	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00324 00093	**DRAFT**                           Numbers                            **DRAFT**
C00326 00094	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00328 00095	**DRAFT**                           Numbers                            **DRAFT**
C00331 00096	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00333 00097	**DRAFT**                           Numbers                            **DRAFT**
C00335 00098	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00337 00099	**DRAFT**                           Numbers                            **DRAFT**
C00340 00100	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00343 00101	**DRAFT**                           Numbers                            **DRAFT**
C00344 00102	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00347 00103	**DRAFT**                           Numbers                            **DRAFT**
C00350 00104	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00351 00105	**DRAFT**                           Numbers                            **DRAFT**
C00354 00106	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00356 00107	**DRAFT**                           Numbers                            **DRAFT**
C00358 00108	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00360 00109	**DRAFT**                           Numbers                            **DRAFT**
C00363 00110	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00366 00111	**DRAFT**                           Numbers                            **DRAFT**
C00369 00112	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00371 00113	**DRAFT**                    Character Manipulation                    **DRAFT**
C00374 00114	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00378 00115	**DRAFT**                    Character Manipulation                    **DRAFT**
C00381 00116	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00382 00117	**DRAFT**                    Character Manipulation                    **DRAFT**
C00385 00118	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00387 00119	**DRAFT**                            Arrays                            **DRAFT**
C00392 00120	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00397 00121	**DRAFT**                            Arrays                            **DRAFT**
C00402 00122	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00405 00123	**DRAFT**                            Arrays                            **DRAFT**
C00409 00124	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00413 00125	**DRAFT**                            Arrays                            **DRAFT**
C00417 00126	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00418 00127	**DRAFT**                      Mapping Functions                       **DRAFT**
C00422 00128	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00425 00129	**DRAFT**                      Mapping Functions                       **DRAFT**
C00429 00130	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00430 00131	**DRAFT**                          The System                          **DRAFT**
C00435 00132	**DRAFT**                          The System                          **DRAFT**
C00436 00133	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00440 00134	**DRAFT**                          The System                          **DRAFT**
C00444 00135	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00446 00136	**DRAFT**                          The System                          **DRAFT**
C00451 00137	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00453 00138	**DRAFT**                          The System                          **DRAFT**
C00456 00139	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00460 00140	**DRAFT**                          The System                          **DRAFT**
C00465 00141	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00469 00142	**DRAFT**                          The System                          **DRAFT**
C00474 00143	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00478 00144	**DRAFT**                          The System                          **DRAFT**
C00481 00145	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00483 00146	**DRAFT**                          The System                          **DRAFT**
C00488 00147	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00493 00148	**DRAFT**                          The System                          **DRAFT**
C00498 00149	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00502 00150	**DRAFT**                          The System                          **DRAFT**
C00506 00151	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00511 00152	**DRAFT**                          The System                          **DRAFT**
C00515 00153	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00519 00154	**DRAFT**                          The System                          **DRAFT**
C00523 00155	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00527 00156	**DRAFT**                          The System                          **DRAFT**
C00531 00157	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00536 00158	**DRAFT**                          The System                          **DRAFT**
C00538 00159	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00542 00160	**DRAFT**                          The System                          **DRAFT**
C00546 00161	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00550 00162	**DRAFT**                          The System                          **DRAFT**
C00554 00163	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00559 00164	**DRAFT**                          The System                          **DRAFT**
C00563 00165	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00565 00166	**DRAFT**                          The System                          **DRAFT**
C00569 00167	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00574 00168	**DRAFT**                          The System                          **DRAFT**
C00577 00169	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00581 00170	**DRAFT**                          The System                          **DRAFT**
C00583 00171	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00586 00172	**DRAFT**                          The System                          **DRAFT**
C00590 00173	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00594 00174	**DRAFT**                          The System                          **DRAFT**
C00598 00175	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00603 00176	**DRAFT**                          The System                          **DRAFT**
C00608 00177	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00612 00178	**DRAFT**                          The System                          **DRAFT**
C00616 00179	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00621 00180	**DRAFT**                          The System                          **DRAFT**
C00626 00181	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00631 00182	**DRAFT**                          The System                          **DRAFT**
C00635 00183	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00640 00184	**DRAFT**                          The System                          **DRAFT**
C00645 00185	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00649 00186	**DRAFT**                          The System                          **DRAFT**
C00654 00187	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00658 00188	**DRAFT**                          The System                          **DRAFT**
C00660 00189	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00664 00190	**DRAFT**                          The System                          **DRAFT**
C00666 00191	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00670 00192	**DRAFT**                          The System                          **DRAFT**
C00674 00193	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00677 00194	**DRAFT**                          The System                          **DRAFT**
C00682 00195	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00687 00196	**DRAFT**                          The System                          **DRAFT**
C00691 00197	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00696 00198	**DRAFT**                          The System                          **DRAFT**
C00701 00199	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00702 00200	**DRAFT**                          The System                          **DRAFT**
C00706 00201	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00711 00202	**DRAFT**                          The System                          **DRAFT**
C00715 00203	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00718 00204	**DRAFT**                          The System                          **DRAFT**
C00722 00205	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00726 00206	**DRAFT**                          The System                          **DRAFT**
C00730 00207	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00735 00208	**DRAFT**                          The System                          **DRAFT**
C00740 00209	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00744 00210	**DRAFT**                          The System                          **DRAFT**
C00748 00211	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00754 00212	**DRAFT**                          The System                          **DRAFT**
C00759 00213	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00763 00214	**DRAFT**                          The System                          **DRAFT**
C00768 00215	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00772 00216	**DRAFT**                          The System                          **DRAFT**
C00777 00217	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00781 00218	**DRAFT**                          The System                          **DRAFT**
C00785 00219	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00789 00220	**DRAFT**                          The System                          **DRAFT**
C00793 00221	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00797 00222	**DRAFT**                          The System                          **DRAFT**
C00802 00223	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00807 00224	**DRAFT**                          The System                          **DRAFT**
C00811 00225	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00814 00226	**DRAFT**                          The System                          **DRAFT**
C00818 00227	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00822 00228	**DRAFT**                          The System                          **DRAFT**
C00826 00229	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00830 00230	**DRAFT**                          The System                          **DRAFT**
C00833 00231	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00836 00232	**DRAFT**                          The System                          **DRAFT**
C00839 00233	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00842 00234	**DRAFT**                          The System                          **DRAFT**
C00846 00235	**DRAFT**                   Maclisp Reference Manual                   **DRAFT**
C00848 ENDMK
C⊗;
**DRAFT**                         The Language                         **DRAFT**


                             Part 1 - The Language

                               Table of Contents



1.      General Information  . . . . . . . . . . . . . . . . . . . . . . . . 1-1
1.1     The Maclisp Language . . . . . . . . . . . . . . . . . . . . . . . . 1-1
1.2     Structure of the Manual  . . . . . . . . . . . . . . . . . . . . . . 1-3
1.3     Notational Conventions . . . . . . . . . . . . . . . . . . . . . . . 1-4

2.      Data Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-7

3.      The Basic Actions of LISP  . . . . . . . . . . . . . . . . . . . . .1-13
3.1     Binding of Variables . . . . . . . . . . . . . . . . . . . . . . . .1-13
3.2     Evaluation of Forms  . . . . . . . . . . . . . . . . . . . . . . . .1-15
3.3     Application of Functions . . . . . . . . . . . . . . . . . . . . . .1-17
3.4     Special Forms  . . . . . . . . . . . . . . . . . . . . . . . . . . .1-21
3.5     Binding Context Pointers . . . . . . . . . . . . . . . . . . . . . .1-24

























December 5, 1977                                                        Page 1-1
**DRAFT**                     General Information                      **DRAFT**




1.  General Information


1.1  The Maclisp Language


   Maclisp is a dialect of  Lisp developed at M.I.T.'s Project MAC  and M.I.T.'s
Artificial Intelligence Laboratory  for use in artificial  intelligence research
and  related fields.   Maclisp  is descended  from the  commonly-known  Lisp 1.5
dialect; however, many features of the language have been changed or augmented.

   This document is intended both as a reference source for the language  and as
a user's guide to three implementations.  These are, in chronological order, the
M.I.T. Artificial Intelligence Lab's  implementation on the DEC  pdp-10 computer
under  their  operating   system  ITS,  hereafter   referred  to  as   "the  ITS
implementation,"  Project MAC's  implementation  on Honeywell's  version  of the
Multics system, hereafter referred  to as "the Multics implementation,"  and the
version  that runs  on  the DEC  pdp-10  under DEC's  TOPS-10  operating system,
hereafter  called "the  DEC-10 implementation."  The DEC-10  implementation also
runs under  TENEX by  means of  a TOPS-10  emulator.  Since  the ITS  and DEC-10
implementations are closely related, they are sometimes referred to collectively
as  the  pdp-10  implementation.    There  are  reputed  to  be   several  other
implementations.

   These implementations  are mostly  compatible; however,  some implementations
have extra features designed to exploit peculiar features of the system on which
they run, and some implementations are temporarily missing some  features.  Most
programs  will work  on any  implementation, although  it is  possible  to write
machine-dependent code if you try hard enough.

   The Maclisp system  is structured as an  environment, which is  essentially a
set  of  names and  bindings  of those  names  to data  structures  and function
definitions.   The  environment contains  a  large number  of  useful functions.
These functions can be used through an interpreter to define other functions, to
control the environment, to do useful work, etc.

   The interpreter is the basic user  interface to the system.  This is  how the
user enters "commands." When Maclisp is not doing anything else, such as running
a program, it waits for the user  to enter a Lisp form.  This form  is evaluated
and  the value  is  printed out.   The  form may  call  upon one  of  the system


December 5, 1977                     ∪1-1.                              Page 1-1
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


functions (or a user-defined function,  of course) to perform some  useful task.
The evaluation  of a  form may  initiate the  execution of  a large  and complex
program,  perhaps never  returning to  the "top  level" interpreter,  or  it may
perform some  simple action and  immediately wait for  the user to  type another
form.

   It is also possible to get  into the interpreter while a program  is running,
using  the break  facility.  This  is primarily  used in  debugging  and related
programming activities.

   The functions invoked by the top-level interpreter may be  executable machine
programs, or they may themselves  be interpreted.  This is entirely a  matter of
choice and convenience.  The system functions are mostly machine programs.  User
functions are usually first used interpretively.  After they work,  the compiler
may be applied  to them, turning  them into machine  programs which can  then be
loaded into the environment.

   All of this is done  within a single consistent language, Lisp,  whose virtue
is that the data structure is simple and general enough that programs may easily
operate on programs, and that the program structure is simple and general enough
that it can be used as a command language.























Page 1-2                             ∪1-1.1                     December 5, 1977
**DRAFT**                     General Information                      **DRAFT**


1.2  Structure of the Manual


   The manual is generally  structured into sections on particular  topics; each
section contains  explanatory text and  function definitions,  interspersed.  In
general,  each  section  contains both  elementary  and  complex  material, with
complexity increasing toward the end of the section.  An axiomatic, step-by-step
development is not used.  Frequently  the more complex information in  a section
will assume knowledge from other sections which appear later in the manual.  The
new user is advised to skip around, reading early chapters and early sections of
chapters first.

   Often descriptions of Lisp functions will be given not only in prose but also
in terms of other Lisp functions.  These are as accurate as possible, but should
not be  taken too  literally.  Their  main purpose is  to serve  as a  source of
examples.

   Accessing information in the manual is dependent on both the user's  level of
ability and the purpose for which  she or he is using the manual.   Though cover
to cover reading is not recommended (though not excluded), it is  suggested that
someone who has  never previously seen this  manual browse through  it, touching
the beginning of each  subdivision that is listed  in the Table of  Contents, in
order to familiarize himself or herself with the material that it  contains.  To
find an answer  to some particular  question, one must  use one of  the provided
access methods.  Since the manual is structured by topics one can use  the Table
of Contents that is found at the beginning of the manual, and the  more detailed
tables of contents  found at the  beginning of each of  the six major  parts, to
find where information of a general class will be found.  Entry into  the manual
is also facilitated by  the Glossary and the  Concept Index, which are  found at
the end.   Also at  the end of  the manual  are a Function  Index and  an Atomic
Symbol Index which are  probably most useful to  a regular and repeated  user of
the dialect, or to  an experienced user of  another dialect, who wishes  to find
out the answer to a question about a specific function.  When one section of the
manual assumes knowledge of another section a page number reference to the other
section will generally be given.









December 5, 1977                     ∪1-1.2                             Page 1-3
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


1.3  Notational Conventions


   There are some conventions of  notation that must be mentioned at  this time,
due to their being used in examples.

   Most numbers are in octal  radix (base eight).  Numbers with a  decimal point
and spelled-out numbers are in decimal radix.  It is important to  remember that
by default Maclisp inputs and outputs  all numbers in octal radix.  If  you want
to change this, see the variables base and ibase.

   A combination  of the characters  equal sign and  greater than  symbol, "=>",
will be used in examples of  Lisp code to mean evaluation.  For instance,  "F =>
V" means that evaluating the form F produces the value V.

   All uses of the phrase "Lisp reader," unless further qualified, refer to that
part of the Lisp  system which reads input, and  not to the person  reading this
document.

   The terms  "S-expression" and "Lisp  object" are synonyms  for "any  piece of
Lisp data."

   The character "$" always stands for dollar-sign, never for "alt mode," unless
that is specifically stated.

   The two characters  accent acute, "'", and  semi-colon, ";", are  examples of
what are called macro characters.  Though the macro character facility, which is
explained in Part 5, is not of immediate interest to a new user of  the dialect,
these two macro characters come preset by the Lisp system and are  useful.  When
the Lisp reader encounters an accent acute, or quote mark, it reads in  the next
S-expression and encloses it in  a quote-form, which prevents evaluation  of the
S-expression.  That is:

                                   'some-atom

                                  turns into:

                               (quote some-atom)

                                      and

                                 '(cons 'a 'b)


Page 1-4                             ∪1-1.3                     December 5, 1977
**DRAFT**                     General Information                      **DRAFT**


                                   turns into

                       (quote (cons (quote a) (quote b)))

   The semi-colon (;) is used  as a commenting character.  When the  Lisp reader
encounters it, the remainder of the line is discarded.

   The  term  "newline"  is used  to  refer  to that  character  or  sequence of
characters which indicates the end of a line.  This is implementation dependent.
In Multics Maclisp, newline is the Multics newline character, octal 012.  In ITS
Maclisp, newline  is carriage  return (octal 015),  optionally followed  by line
feed (octal  012.)  In dec-10  Maclisp, newline is  carriage return  followed by
line feed.

   All  Lisp  examples  in  this  manual  are  written  according  to  the  case
conventions of the Multics implementation, which uses both upper and  lower case
letters  and spells  the names  of most  system functions  in lower  case.  Some
implementations of  Maclisp use only  upper case letters  because they  exist on
systems which are not, or have not always been, equipped with  terminals capable
of  generating and  displaying  the full  ascii character  set.   However, these
implementations will accept input in lower case and translate it to  upper case,
unless the user has explicitly said not to.






















December 5, 1977                     ∪1-1.3                             Page 1-5
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**














































Page 1-6                             ∪1-1.3                     December 5, 1977
**DRAFT**                         Data Objects                         **DRAFT**




2.  Data Objects


   Lisp works with pieces of data called "objects" or "S-expressions." These can
be simple "atomic" objects or  complex objects compounded out of  other objects.
Functions,  the basic  units of  a Lisp  program, are  also objects  and  may be
manipulated as data.

   Objects  come in  several types.   All  types are  manifest; that  is,  it is
possible for a  program to tell what  type an object is  just by looking  at the
object itself, so it  is not necessary to declare  the types of variables  as in
some other languages.  One can  make declarations, however, in order to  aid the
compiler in producing optimal code.  (See part 4.2.)

   It is important to know that  Lisp represents objects as pointers, so  that a
storage cell (a  "variable") will hold  any object, and  the same object  may be
held by several different storage cells.  For example, the same identical object
may be a component of two different compound objects.

   The data-types are divided into  three broad classes:  the atomic  types, the
non-atomic types, and  the composite types.  Objects  are divided into  the same
three classes  according to their  type.  Atomic objects  are basic  units which
cannot be broken down by ordinary chemical means (car and cdr), while non-atomic
objects are structures constructed out of other objects.  Composite  objects are
indivisible, atomic,  entities which  have other  objects associated  with them.
These other objects may be examined and replaced.

   The atomic data types are numbers, atomic symbols, strings, and subr-objects.
Atomic symbols can also be regarded as composite.  See below.

   In Lisp numbers can be represented by three types of atomic objects: fixnums,
flonums, and bignums.  A fixnum  is a fixed-point binary integer whose  range of
values  is  machine-dependent.   A  flonum  is  a  floating-point  number  whose
precision and range of values  are machine-dependent.  A bignum is  an infinite-
precision integer.  It is impossible to get "overflow" in bignum  arithmetic, as
any  integer  can  be  represented by  a  bignum.   However,  fixnum  and flonum
arithmetic is faster than bignum arithmetic and requires less memory.  Sometimes
the  word  "fixnum"  is used  to  include  both fixnums  and  bignums  (i.e. all
integers); in  this manual,  however, the word  "fixnum" will  never be  used to
include bignums unless that is explicitly stated.


December 5, 1977                     ∪1-2.                              Page 1-7
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   The  printed  representations  for  numbers  are  as  follows:  a  fixnum  is
represented  as a  sequence of  digits in  a specified  base, usually  octal.  A
trailing decimal point indicates a  decimal base.  A flonum is represented  as a
set of digits containing an embedded or leading decimal point and/or  a trailing
exponent.  The exponent is introduced by  an upper or lower case "e".   A bignum
looks like a fixnum except that it has enough digits that it will not fit within
the range available to  fixnums.  Any number may be  preceded by a + or  - sign.
Some examples of fixnums are 4,  -1232, -191., +46.   An example of a  bignum is
1565656565656565656565656565656565.  Some  examples of  flonums are:   4.0, .01,
-6e5, 4.2e-1.

   One of the most important Lisp data types is the atomic symbol.  In fact, the
word "atom" is often used to mean just atomic symbols, and not the  other atomic
types.  An atomic symbol has associated with it a name, a value, and  possibly a
list  of "properties".   The name  is  a sequence  of characters,  which  is the
printed representation  of the  atomic symbol.   This name  is often  called the
"pname," or  "print-name." A pname  may contain any  ascii character  except the
null character, which  causes trouble in  some implementations.  For  example, a
certain atomic symbol  would be represented externally  as foo; internally  as a
structure containing the value, the pname "foo", and the properties.

   There are  two special atomic  symbols, t and  nil.  These always  have their
respective selves as values and their values may not be changed.  nil is used as
a  "marker"  in many  contexts;  it is  essential  to the  construction  of data
structures  such as  lists.  t  is usually  used when  an antithesis  to  nil is
required for some purpose, e.g.  to represent the logical conditions  "true" and
"false." Another property of the special  atomic symbol nil is that its  car and
its cdr are always nil.

   The value  of an  atomic symbol  can be any  object of  any type.   There are
functions to set  and get the  value of a  symbol.  Because atomic  symbols have
values associated with them,  they can be used  as variables in programs  and as
"dummy arguments"  in functions.  It  is also possible  for an atomic  symbol to
have no value, in which case it is said to be "undefined" or "unbound."

   The property list of an atomic symbol is explained on page 2-52.  It  is used
for such things  as recording the fact  that an atomic symbol  is the name  of a
function.

   An atomic symbol  with one or  no characters in its  pname is often  called a
"character object" and used to represent an ascii character.  The  atomic symbol
with a zero-length  pname represents the ascii  null character, and  the symbols


Page 1-8                             ∪1-2.                      December 5, 1977
**DRAFT**                         Data Objects                         **DRAFT**


with  one-character  pnames  represent  the  character  which  is  their  pname.
Functions which take character objects as input usually also accept a string one
character long  or a  fixnum equal to  the ascii-code  value for  the character.
Character objects are always interned on the obarray (see page 2-58).

   Another  Lisp data  type is  the string.   This is  a sequence  of characters
(possibly zero-length).  Strings are used  to hold messages to be typed  out and
to manipulate text when the structure of the text is not appropriate for the use
of "list processing."  The printed representation of  a string is a  sequence of
characters enclosed in double-quotes, e.g. "foo".   If a " is to be  included in
the string, it is written twice, e.g. "foo""bar" is foo"bar.  In implementations
without strings,  atomic symbols are  used instead.  The  pdp-10 implementations
currently lack strings.

   A "subr-object" is a special atomic data-type whose use is normally hidden in
the  implementation.  A  subr-object  represents executable  machine  code.  The
functions built  into the Lisp  system are subr-objects,  as are  user functions
that have been compiled.  A  subr-object has no printed representation,  so each
system function has an atomic symbol  which serves as its name.  The  symbol has
the subr-object as a property.

   One composite  data type  is the  array.  An  array consists  of a  number of
cells, each of  which may contain  any Lisp object.  The  cells of an  array are
accessed by subscripting; each cell is  named by a tuple of integers.   An array
may have one or more dimensions; the upper limit on the number of  dimensions is
implementation-defined.  An array is not always associated with an atomic symbol
which is its name.  Rather,  an array is always designated by  an array-pointer,
which is  a special kind  of atomic Lisp  object.  Frequently,  an array-pointer
will be placed on  the property list of a  symbol under the indicator  array and
then that symbol will be used as  the name of the array, since symbols  can have
mnemonic names and  a reasonable printed representation.   See page 2-89  for an
explanation of how to create, use, and delete arrays.

   Another composite data  type is the file-object,  which is described  on part
5.3.

   The  sole  non-atomic  data  type  is the  "cons."   A  cons  is  a structure
containing  two  components,  called  the "car"  and  the  "cdr"  for historical
reasons.  (These are  names of fields  in an IBM  7094 machine word.)  These two
components may be any Lisp object,  even another cons (in fact, they  could even
be the same cons).  In this way complex structures can be built up out of simple
conses.  Internally a cons is represented in a form similar to:


December 5, 1977                     ∪1-2.                              Page 1-9
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



               ←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←
               |                 |                 |
               |     car         |      cdr        |
               ←λ|←←←←←←←←←←←←←←←←←←λ|←←←←←←←←←←←←←←←←←←λ|

where the boxes represent  cells of memory large  enough to hold a  pointer, and
"car" and "cdr"  are two pointers to  objects.  The printed representation  of a
cons is the "dotted-pair" notation (A . B) where A is the car and B is the cdr.

   Another way  to write the  internal representation of  a cons, which  is more
convenient for large structures, is:

                ---λ> o -----> cdr
                    |
                    |
                    V
                   car

   There are three  Lisp functions associated with  conses: cons, car,  and cdr.
The  function cons  combines its  two  arguments into  a cons;  (1 .  2)  can be
generated by evaluating (cons 1 2).  The function car returns the  car component
of its argument, and the function cdr returns the cdr component of its argument.

   One type of structure, built out of conses, that is used quite often,  is the
"list."  A  list is  a row  of objects, of  arbitrary length.   A list  of three
things 1, 2, and 3  is constructed by (cons 1 (cons  2 (cons 3 nil))); nil  is a
special atom that is used  to mark the end of  a list.  The structure of  a list
can be diagrammed as:

                ---λ> o ----> o ----> o ----> nil
                    |       |       |
                    |       |       |
                    V       V       V
                    1       2       3

   From this it can be  seen that the car of  a list is its first  element, that
the cdr of a list is a list  of the elements after the first, and that  the list
of no elements is the same as nil.

   This list of 1,  2, and 3 could be  represented in the dot-notation  used for
conses as  (1 . (2 . (3  . nil))).  However, a more convenient notation  for the


Page 1-10                            ∪1-2.                      December 5, 1977
**DRAFT**                         Data Objects                         **DRAFT**


printed representation of lists has been defined:  the "list-notation" (1  2 3).
It is  also possible to  have a hybrid  of the two  notations which is  used for
structures which are almost  a list except that they  end in an atom  other than
nil.  For example, (A . (B . (C . D))) can be represented as (A B C . D).

   A list not  containing any elements is  perfectly legal and  frequently used.
This zero-length list is  identified with the atom nil.   It may be typed  in as
either nil or ().




































December 5, 1977                     ∪1-2.                             Page 1-11
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**














































Page 1-12                            ∪1-2.                      December 5, 1977
**DRAFT**                  The Basic Actions of LISP                   **DRAFT**




3.  The Basic Actions of LISP


3.1  Binding of Variables


   The  basic  primitives  of  programming in  Lisp  are  variables,  forms, and
functions.  A variable is an atomic symbol which has a value associated with it;
the symbol is said to  be bound to that value.   The value may of course  be any
Lisp object whatsoever.  The  atomic symbol acts simply  as a name by  which the
program may refer to the value while it is processing it.

   This is similar to the  concept of variables in other  programming languages.
However, Lisp's concept of the scope  of names is subtly different from  that of
most "block-structured" languages.  At  a given moment, a variable  may actually
have several bindings in existence.   Only the most recent, or  current binding,
can be used.  When a new binding  is created, the previous one is pushed  onto a
stack.  It will become accessible again when the binding which superseded  it is
removed.   Creation  and removal  of  bindings is  synchronized  with subroutine
calling  (and with  certain  special forms  described below)  so  this mechanism
corresponds  closely  to  the "local  variables"  concept  of  other programming
languages.   However,  Lisp considers  that  there is  only  one  variable whose
binding changes, rather than several separate variables which happen to have the
same  name.  Any  reference  to a  variable,  even from  outside  the particular
program which gave it its current binding, gets the current binding and  not one
determined by "scope  rules."  It is possible  to simulate the other  concept of
scope of names by using binding context pointers, which are described later (see
page 1-22).

   Unlike many other languages, Lisp  does not combine the concepts of  name and
storage.  Many languages associate with  a variable (a name) a piece  of storage
which can hold one object of a particular type, such as a floating point number.
The variable's  value resides in  this storage.  It  is then impossible  for two
variables to really have "the same" value; one could have a copy of the value of
another but not the same identical object.

   The situation in Lisp is quite  different.  Binding a variable to a  value is
not copying the value into storage associated with that variable.   Values exist
as separate objects  in their own  right and in  their own storage.   Binding is
simply an association between a  variable and a value; consequently there  is no


December 5, 1977                     ∪1-3.                             Page 1-13
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


reason why two variables cannot have truly identical values.  Similarly, erasing
the binding between a variable and its value does not destroy or throw  away the
value; it simply breaks  the association.  Of course,  if there is no  other use
for the value the storage it occupies will eventually be reclaimed by the system
and put to more productive use.

   Often these processes of creating a new binding of a variable to a  value and
reverting to  a previous binding  are referred to  as binding and  unbinding the
variable, respectively.

   A slightly different way of creating a binding between a variable and a value
is assignment.  When  a variable is  bound to a  value, the previous  binding is
saved and can be restored, but when  a variable has a value assigned to  it, the
previous binding  is not  saved, but is  simply replaced.   Thus binding  may be
regarded  as creating  a new  level  of usage  of a  variable,  while assignment
switches a variable to a different value within the same level.  For instance, a
subroutine  or function  may bind  a variable  to an  initial value  when  it is
entered, and  then proceed to  make use of  that variable, possibly  assigning a
different value to it  from time to time.   The initial binding of  the variable
establishes the (temporary) ownership of that variable by the subroutine.

   Due to the subtlety of  the distinction between binding and  assignment, some
people  have  proposed that  assignment  be eliminated  wherever  possible.  The
Maclisp do function can often be useful in this regard.

   There are several program constructs by which a variable can be bound.  These
will be explained after forms and functions have been introduced.

















Page 1-14                            ∪1-3.1                     December 5, 1977
**DRAFT**                  The Basic Actions of LISP                   **DRAFT**


3.2  Evaluation of Forms


   The  process of  "executing" a  Lisp program  consists of  the  evaluation of
forms.  Evaluation takes a form and produces from it a value (any  Lisp object),
according to  a strict  set of  rules which  might be  regarded as  the complete
semantics of Lisp.

   If the form  is atomic, it is  evaluated in a way  which depends on  its data
type.  An atomic symbol is a variable; it evaluates to the value to which  it is
currently bound.  If it is not bound, an error occurs.  (See part 3.4.) A number
or a string is a literal  constant; it evaluates to itself.  The  special atomic
symbols t and nil are also treated as constants.  A constant can also be created
by use of the quote special form; the value of (quote x) is x.

   If  the form  is a  list, its  first element  specifies the  operation  to be
performed, and its remaining elements specify arguments to that operation.  Non-
atomic forms  come in  two types:   special forms,  which include  the necessary
programming  operations  such  as  assignment  and  conditionals,  and  function
references,  in which  the "operation"  is a  function which  is applied  to the
specified  arguments.   Thus  functional  composition  is  the  method  by which
programs are built up out of  parts - as distinguished from composition  of data
structures, for  example.  Lisp functions  correspond closely to  subroutines in
other programming languages.

   A function  may be  either a primitive  which is  directly executable  by the
machine,  called a  subr  (short for  "subroutine"),  or a  function  defined by
composition  of  functions  and  special  forms,  called  an  expr   (short  for
"expression.") Most subrs are built in to the language, but it is possible for a
user to convert his exprs into  subrs by using the compiler (see part  4.)  This
gains speed and compactness at some cost in debugging features.

   There is additional complexity because special forms are actually implemented
as if they were function references.  There is a special type of subr  called an
fsubr  which is  used  for this  purpose.  An  fsubr  is permitted  to  make any
arbitrary  interpretation  of  its  argument  specification  list,   instead  of
following the standard procedure which is described below.  It is  also possible
to define a special form by an expr, which is then called a fexpr.  Most  of the
built-in special forms are handled specially by the compiler.  They are compiled
as the appropriate code rather than as a call to the fsubr.

   Other types  of functions are  lsubr, which  is just a  subr with  a variable


December 5, 1977                     ∪1-3.2                            Page 1-15
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


number  of  arguments,  lexpr,  which  is an  expr  with  a  variable  number of
arguments, and  macro, which is  a type of  special form whose  result is  not a
value, but another form; this allows a "transformational" type of semantics.

   Consider the form

                                (F A1 A2 ... An)


   The evaluator first  examines F to  see if it is  a function which  defines a
special form, i.e. an fsubr, a fexpr, or a macro.  If so, F is consulted  and it
decides how to produce  a value.  If not, F  must be an ordinary  function.  The
sub-forms  A1 through  An are  evaluated, producing  n arguments,  and  then the
definition of F is applied  to the arguments.  (Application is described  in the
following section.) This yields a result (some Lisp object), which is then taken
as the value of the form.

   An atomic  form of some  random type, such  as a subr-object,  a file,  or an
array-pointer, evaluates to  something random, often  itself; or else  causes an
error depending on the convenience  of the implementation.  Note that  an array-
pointer is different from  an atomic symbol which happens  to be the name  of an
array; such an atomic symbol is evaluated the same as any other atomic symbol.






















Page 1-16                            ∪1-3.2                     December 5, 1977
**DRAFT**                  The Basic Actions of LISP                   **DRAFT**


3.3  Application of Functions


   When a non-atomic form is  evaluated, the function specified by that  form is
combined with  the arguments specified  by that form  to produce a  value.  This
process  is  called application;  the  function is  said  to be  applied  to the
arguments.

   The first  step in application  is to convert  the function-specifier  into a
functional  expression  (sometimes  confusingly  called  a  functional  form.) A
functional expression is a Lisp object  which is stylized so that Lisp  can know
how to apply it to arguments.   The rules for this conversion will  be described
after the types of functional expressions have been explained.

   There are basically two types of functional expression.   A lambda-expression
is a functional expression which specifies some variables which are to  be bound
to the arguments, and  some forms which are  to be evaluated.  One  would expect
the forms to depend on the variables.  The value of the last form is used as the
value  of the  application of  the lambda-expression.   Any preceding  forms are
present purely for their side-effects.  A lambda-expression looks like:

                     (lambda (a b c d)
                         form1
                         form2
                         form3)

   Here a,  b, c,  and d  are the variables  to be  bound to  the values  of the
arguments,  called the  lambda-variables.  If  at a  certain moment  the current
binding of a was the one  created by this lambda-expression, a would be  said to
be lambda-bound.   Clearly this  lambda-expression is  a function  which accepts
four arguments.  The application of the functional expression to  four arguments
produces a value by evaluating form1, then form2, and then form3.  The  value of
form3 is the value of the whole form.  For example, the value of the form

                    ((lambda (a b) b) 3 4)

is 4.  The  functional expression used  is a very  simple one which  accepts two
arguments and returns the second one.

   If we grant the existence of a primitive addition operation, whose functional
expression may be designated by +, then the value of the form



December 5, 1977                     ∪1-3.3                            Page 1-17
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



                    ((lambda (a b) (+ a b)) 3 4)

is 7.  Actually,

                    (+ 3 4)

evaluates to the same thing.

   The  second basic  type of  functional  expression is  the subr,  which  is a
program  directly executable  by the  machine.  The  arguments of  the  form are
conveyed  to  this  program  in a  machine-dependent  manner,  it  performs some
arbitrary computation, and it returns a result.  The built in primitives  of the
language are subrs, and the user may write lambda-expressions which make  use of
these subrs to define  his own functions.  The  compiler may be used  to convert
user functions into subrs if extra efficiency is required.

   It  is  extremely  convenient  to  be  able  to  assign  names  to functional
expressions.  Otherwise the  definition of a function  would have to  be written
out in full each time it was used, which would be impossibly cumbersome.

   Lisp uses atomic symbols to name functions.  The "property list" mechanism is
used to associate an atomic symbol with a functional expression.  (See page 2-52
for an  explanation of  property lists.)  Because the  binding mechanism  is not
used, it is  possible for the  same name to  be used for  both a variable  and a
function with no conflict.  Usually the defun special form is used  to establish
the association between a function name and a functional expression.

   Thus, the car of a form  may be either a functional expression itself,  or an
atomic symbol which names a functional expression.  In the latter case, the name
of  the  "property"  which  associates  the  symbol  with  the  expression gives
additional information:

   A lambda-expression is normally placed under the expr property.  This defines
an ordinary expr.

   If  a lambda-expression  is placed  under the  fexpr property,  it  defines a
special form.  In that  case, the first lambda-variable  is bound to the  cdr of
the form being evaluated.  For example, if foo is a fexpr, and (foo (a b) (c d))
is evaluated, then  foo's lambda-variable would  be bound to  ((a b) (c  d)).  A
second lambda-variable may optionally be included in a fexpr.  It will  be bound
to a "binding  context pointer" to  the context of  the evaluation of  the form.
(See page 1-22 for the details of binding context pointers.)

Page 1-18                            ∪1-3.3                     December 5, 1977
**DRAFT**                  The Basic Actions of LISP                   **DRAFT**


   If a  lambda-expression with  one lambda-variable is  placed under  the macro
property,  it defines  the "macro"  special form  mentioned above.   The lambda-
expression is applied to the entire form, as a single argument, and the value is
a new form that is evaluated in place of the original form.

   If a subr-object is placed under the subr property, it defines a subr.   If a
subr-object is placed  under the fsubr property,  it defines a special  form.  A
subr-object  under  the lsubr  property  defines a  subr  which  accepts varying
numbers of arguments.

   There  are some  additional refinements.   A lambda-expression  which accepts
varying numbers of arguments, called a lexpr, looks as follows:

                    (lambda n
                        form1
                        form2)

The  single,  unparenthesized,  lambda-variable  n is  bound  to  the  number of
arguments.  The function arg, described on page 2-12, may be used to  obtain the
arguments.

   Another  property  which  resembles a  functional  property  is  the autoload
property.  If Lisp encounters an autoload property while searching  the property
list of a  symbol for functional  properties, it loads  in the file  of compiled
functions  specified by  the property,  then searches  the property  list again.
Presumably the file would contain  a definition for the function  being applied,
and  that  definition would  be  found the  second  time through.   In  this way
packages  of  functions  which  are  not  always  used  can  be  present  in the
environment only when needed.

   An array may also  be used as a  function.  The arguments are  the subscripts
and the  value is the  contents of the  selected cell of  the array.   An atomic
symbol  with an  array property  appearing in  the function  position in  a form
causes that array to be used.

   If the function-specifier of a form doesn't meet any of the above tests, Lisp
evaluates it and tries again.  In this way, "functional variables" and "computed
functions" can  be used.   However, it is  better to  use the  funcall function.
(See page 2-13.)

   There are some other cases of lesser importance:



December 5, 1977                     ∪1-3.3                            Page 1-19
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   There is an obscure type of functional expression called  a label-expression.
It looks like
                        (label name (lambda (...) ...))

The  atomic symbol  name  is bound  to  the enclosed  lambda-expression  for the
duration of the application of the label-expression.  Thus if name is used  as a
functional variable this temporary definition  will be used.  This is  mostly of
historical interest and is rarely used in actual programming.

   Another type  of functional  expression is the  funarg.  A  funarg is  a list
beginning with the atomic symbol  funarg, as you might expect, and  containing a
function and a binding context pointer.  Applying a funarg causes  the contained
function to  be applied in  the contained binding  context instead of  the usual
context.  funargs are created by the *function special form.

   An expr property may be an atomic symbol rather than a lambda-expression.  In
this case, the atomic  symbol is used as  the function.  The original  symbol is
simply a synonym for it.

   In  addition to  the variety  of application  just described,  which  is used
internally by  the evaluation procedure,  there is a  similar but  not identical
application procedure available through the function apply.  The main difference
is that the function and the arguments are passed to apply separately.  They are
not encoded into a form, consequently macros are not accepted by this version of
application.  Note that what  is passed to apply is  a list of arguments,  not a
list of expressions which, evaluated, would yield arguments.


















Page 1-20                            ∪1-3.3                     December 5, 1977
**DRAFT**                  The Basic Actions of LISP                   **DRAFT**


3.4  Special Forms


   This section  briefly describes some  of the special  forms in  Maclisp.  For
full details on a specific special form, consult the Function Index in the back.

Constants

     (quote x) evaluates to the S-expression x.

     (function x)  evaluates to  the functional expression  x.  There  is little
     real  difference  between  quote  and function.   The  latter  is  simply a
     mnemonic reminder to anyone who reads the program - including  the compiler
     - that the specified expression is supposed to be some kind of function.

Conditionals

     Conditionals control whether or not certain forms are  evaluated, depending
     on the results of evaluating other forms.  Thus both the value and the side
     effects of the conditional form can be controlled.

        (cond (predicate form1 form2...) (predicate form1 form2...)...)

     is a general conditional form.  The lists of a predicate and some forms are
     called clauses.  The  cond is evaluated by  considering the clauses  one by
     one in the order they are written.  The predicate of a clause is evaluated,
     and if the result is true, that is, anything other than nil, then the forms
     in that clause are evaluated and the cond is finished without examining the
     remaining clauses.  If the result is not true, i.e. if it is nil,  then the
     next clause is examined in the same way.  If all the clauses are exhausted,
     that is not an error.  The value of a cond is the value of the last form it
     evaluates, which could be  nil if no predicate is  true, or the value  of a
     predicate if that predicate is true but has no forms in its clause.

     (and form1 form2 form3...) evaluates  the forms in succession until  one is
     nil or the  forms are exhausted,  and the result is  the value of  the last
     form evaluated.

     (or form1 form2 form3...) evaluates  the forms until one is non-nil  or the
     forms  are  exhausted,  and  the  result is  the  value  of  the  last form
     evaluated.



December 5, 1977                     ∪1-3.4                            Page 1-21
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


Non-Local Exits

     (catch form tag) evaluates the  form, but if the special form  (throw value
     tag)  is encountered,  and the  tags are  the same,  the  catch immediately
     returns the value without further ado.  See page 2-44 for the full details.

Iteration

     (prog  (variable...)  form-or-tag ...)  allows  Fortranoid  "programs" with
     goto's, local variables, and return's to be written.

     (do ...) is the special form for iteration.  See page 2-38 for  the details
     of prog and do.

Defining Functions

     (defun  name  (arg1  arg2...)  form1  form2...)  defines  an  (interpreted)
     function.  See page 2-60 for full details.

Error Control

     (break name t) causes ";bkpt name"  to be typed out and gives control  to a
     read-eval-print loop so that the  user can examine and change the  state of
     the world.  When he is satisfied, the user can cause the break to  return a
     value.  See page 3-5 for the details of break.

     (errset form) evaluates the form, but if an error occurs the  errset simply
     returns nil.  If no error occurs, the value is a list whose  single element
     is what the value of the form would have been without errset.

Assignment

     (setq var1 value1 var2 value2...) assigns the values to the variables.  The
     values are forms which are evaluated.

     (store (array  subscript1 subscript2...)  value) assigns  the value  to the
     array cell selected by subscripting.  See part 2.8 for  further information
     on arrays.






Page 1-22                            ∪1-3.4                     December 5, 1977
**DRAFT**                  The Basic Actions of LISP                   **DRAFT**


Miscellaneous Parameters

     (status  name -optional  args-) returns  miscellaneous parameters  of LISP.
     name is a mnemonic name for what is to be done.

     (sstatus name -optional args-) sets miscellaneous parameters.

     See part 3.7 for the details of status and sstatus.

Pretty-Printing

     (grindef x) prettily prints the  value and function definition (if  any) of
     the atomic symbol  x.  Indentation is used  to reveal structure,  the quote
     special form is represented by ', etc. See part 6.3 for the details.

Tracing

     (trace name)  causes the function  name to print  a message whenever  it is
     called and whenever  it returns.  See page  3-34 for the many  features and
     options of trace.
























December 5, 1977                     ∪1-3.4                            Page 1-23
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


3.5  Binding Context Pointers


   There  is a  special type  of  object called  a binding  context  pointer, or
sometimes an "a-list pointer", which can  be used to refer to a  binding context
(a set  of bindings of  variables and  values which was  extant at  a particular
instant.) Due to the stack implementation of Maclisp, a binding  context pointer
is only valid while control is  nested within the binding context it  names.  It
is not  possible to exit  from within a  binding context but  keep it  around by
retaining a pointer to it.

   A binding context pointer is either a negative fixnum or nil.  nil  means the
"global" or "top level" binding context.  The negative fixnum is a special value
of implementation dependent  meaning which should be  obtained only from  one of
the four following sources:  the function evalframe, the function  errframe, the
special form *function, or the second lambda-variable of a fexpr.

   The only use for  binding context pointers is  to pass them to  the functions
eval and  apply to  specify the  binding context  in which  variables are  to be
evaluated  and  assignments  are  to  be  performed  during  that  evaluation or
application.  Binding  context pointers are  also used internally  by *function.
When it generates a funarg, it  puts in the funarg the functional  expression it
was  given and  a binding  context pointer  designating the  binding environment
current at the time *function was called.




















Page 1-24                            ∪1-3.5                     December 5, 1977
**DRAFT**                    Function Descriptions                     **DRAFT**


                         Part 2 - Function Descriptions

                               Table of Contents



1.      Predicates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-1

2.      The Evaluator  . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-7

3.      Manipulating List Structure  . . . . . . . . . . . . . . . . . . . .2-15
3.1     Conses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-15
3.2     Lists  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-18
3.3     Alteration of List Structure . . . . . . . . . . . . . . . . . . . .2-23
3.4     Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-25
3.5     Sorting  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-31
3.6     Hunks  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-33

4.      Flow of Control  . . . . . . . . . . . . . . . . . . . . . . . . . .2-35
4.1     Conditionals . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-36
4.2     Iteration  . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-38
4.3     Non-local Exits  . . . . . . . . . . . . . . . . . . . . . . . . . .2-44
4.4     Causing and Controlling Errors . . . . . . . . . . . . . . . . . . .2-46

5.      Atomic Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . .2-49
5.1     The Value Cell . . . . . . . . . . . . . . . . . . . . . . . . . . .2-49
5.2     The Property List  . . . . . . . . . . . . . . . . . . . . . . . . .2-52
5.3     The Print-Name . . . . . . . . . . . . . . . . . . . . . . . . . . .2-56
5.4     Interning of Symbols . . . . . . . . . . . . . . . . . . . . . . . .2-58
5.5     Defining Atomic Symbols as Functions . . . . . . . . . . . . . . . .2-61

6.      Numbers  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-65
6.1     Number Predicates  . . . . . . . . . . . . . . . . . . . . . . . . .2-65
6.2     Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-67
6.3     Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-69
6.4     Arithmetic . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-71
6.5     Exponentiation and Logarithm Functions . . . . . . . . . . . . . . .2-79
6.6     Trigonometric Functions  . . . . . . . . . . . . . . . . . . . . . .2-80
6.7     Random Functions . . . . . . . . . . . . . . . . . . . . . . . . . .2-81
6.8     Logical Operations on Numbers  . . . . . . . . . . . . . . . . . . .2-82

7.      Character Manipulation . . . . . . . . . . . . . . . . . . . . . . .2-85
7.1     Character Objects  . . . . . . . . . . . . . . . . . . . . . . . . .2-85

December 5, 1977                                                        Page 2-1
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


7.2     Character Strings  . . . . . . . . . . . . . . . . . . . . . . . . .2-89

8.      Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-91

9.      Mapping Functions  . . . . . . . . . . . . . . . . . . . . . . . . .2-99







































Page 2-2                                                        December 5, 1977
**DRAFT**                          Predicates                          **DRAFT**




1.  Predicates


   A  predicate is  a  function which  tests  for some  condition  involving its
argument and returns t if that condition is true, or nil if it is not true.

   The  following  predicates are  for  checking data  types.   These predicates
return t if their argument is of the type indicated by the name of the function,
nil if it is of some other type.  Note that the name of most predicates  ends in
the letter p, by convention.


atom                SUBR 1 arg

     The atom predicate returns nil if its argument is a dotted-pair or  a list,
     or t  if it  is any kind  of atomic  object such as  a number,  a character
     string, or an atomic symbol.


symbolp             SUBR 1 arg

     The symbolp predicate returns t if its argument is an atomic symbol, or nil
     if it is anything else.


fixp                SUBR 1 arg

     The fixp  predicate returns  t if  its argument  is a  fixnum or  a bignum,
     otherwise nil.


floatp              SUBR 1 arg

     The floatp predicate returns  t if its argument is  a flonum, nil if  it is
     not.


bigp                SUBR 1 arg

     The  predicate  bigp  returns  t  if its  argument  is  a  bignum,  and nil
     otherwise.

December 5, 1977                     ∪2-1.                              Page 2-1
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


numberp             SUBR 1 arg

     The numberp predicate returns t if its argument is any kind of  number, nil
     if it is not.


hunkp               SUBR 1 arg

     The hunkp predicate returns t if its argument is a hunk (see page  2-33 for
     a discussion of  hunks).  hunkp does not  consider list cells to  be hunks.
     This predicate does not exist in the Multics implementation.


typep               SUBR 1 arg

     typep is a general  function for constructing type-predicates.   It returns
     an atomic symbol describing the type of its argument, chosen from the list

             (fixnum flonum bignum list symbol string array random)

     symbol means  atomic symbol.   list means a  list or  a cons.   array means
     array-pointer.   random  is for  all  types  that don't  fit  in  any other
     category.  Thus numberp could have been defined by:

        (defun numberp (x)
           (and (memq (typep x) '(fixnum flonum bignum))
                t))


The following two functions only exist in the Multics implementation.


stringp             SUBR 1 arg

     The stringp predicate returns t if its argument is a string, otherwise nil.


subrp               SUBR 1 arg

     The subrp predicate returns  t if its argument  is a "subr" object,  i.e. a
     pointer to the machine code for a compiled or system function.  Example:

            (subrp (get 'car 'subr)) => t

Page 2-2                             ∪2-1.                      December 5, 1977
**DRAFT**                          Predicates                          **DRAFT**


The following are a more miscellaneous set of predicates.


eq                  SUBR 2 args

     (eq x y) => t  if x and y are  exactly the same object, nil  otherwise (cf.
     equal).   It  should be  noted  that things  that  print the  same  are not
     necessarily eq to each other.   In particular, numbers with the  same value
     need not be eq, and two similar lists are usually not eq.  In  general, two
     atomic symbols  with the same  print-name are eq,  but it is  possible with
     maknam or multiple obarrays to generate symbols which have the  same print-
     name but are not eq.  Examples:

                (eq 'a 'b) => nil
                (eq 'a 'a) => t
                (eq '(a . b) '(a . b)) => nil (usually)
                (eq (cons 'a 'b) (cons 'a 'b)) => nil (always)
                (setq x '(a . b)) (eq x x) => t since it is
                             the same copy of (a . b) in both arguments.
                (setq x (setq y 17)) (eq x y) => t or nil
                              depending on the implementation.  You can
                              never rely on numbers being eq.


equal               SUBR 2 args

     The equal  predicate returns  t if its  arguments are  similar (isomorphic)
     objects. (cf.  eq) Two numbers  are equal  if they have  the same  value (a
     flonum is never equal to a  fixnum though).  Two strings are equal  if they
     have the same length, and the characters composing them are the  same.  All
     other atomic  objects are equal  if and  only if they  are eq.   For dotted
     pairs and lists, equal is defined recursively as the two car's  being equal
     and the two cdr's being equal.  Thus equal could have been defined by:











December 5, 1977                     ∪2-1.                              Page 2-3
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



            (defun equal (x y)
                (or (eq x y)
                    (and (numberp x) (numberp y) (numequal x y))
                    (and (not (atom x))
                         (not (atom y))
                         (equal (car x) (car y))
                         (equal (cdr x) (cdr y)))))

            if there was an auxiliary function for numeric equality:

           (defun numequal (x y)
                 (and (eq (typep x) (typep y))
                      (zerop (difference x y))))

     This  numequal function  is not  the same  as the  Maclisp numeric-equality
     function, =, because the latter only compares non-big numbers.

     As a consequence of  the above definition, it  can be seen that  equal need
     not  terminate when  applied  to looped  list structure.   In  addition, eq
     always implies equal.  An intuitive definition of equal (which is not quite
     correct) is that two objects are  equal if they look the same  when printed
     out.


not                 SUBR 1 arg

     not returns t if its argument is nil, otherwise nil.


null                SUBR 1 arg

     This is  the same  as not.   Both functions  are provided  for the  sake of
     clarity.  null should  be used to  check if something  is nil and  return a
     logical value.  not should be used to invert the sense of a  logical value.
     Even though Lisp uses nil to represent logical "false," you  shouldn't make
     understanding your program depend on this.  For example, one often writes

        (cond ((not (null x)) ... )
              ( ... ))




Page 2-4                             ∪2-1.                      December 5, 1977
**DRAFT**                          Predicates                          **DRAFT**



        rather than

        (cond (x ... )
              ( ... ))

     There is no  loss of efficiency since  these will compile into  exactly the
     same instructions.


   See also the number predicates (page 2-63).

































December 5, 1977                     ∪2-1.                              Page 2-5
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**














































Page 2-6                             ∪2-1.                      December 5, 1977
**DRAFT**                        The Evaluator                         **DRAFT**




2.  The Evaluator


eval                LSUBR 1 or 2 args

     (eval  x) evaluates  x, as  a form,  atomic or  otherwise, and  returns the
     result.

     (eval x  p) evaluates  x in the  context specified  by the  binding context
     pointer p.  Example:

          (setq x 43 foo 'bar)
          (eval (list 'cons x 'foo))
               => (43 . bar)


apply               LSUBR 2 or 3 args

     (apply f y) applies the function f to the list of arguments y.  Unless f is
     an fsubr or fexpr, such as cond or and, which evaluates its arguments  in a
     funny way, the  arguments in the list  y are used without  being evaluated.
     Examples:

            (setq f '+) (apply f '(1 2 3)) => 6
            (setq f '-) (apply f '(1 2 3)) => -4
            (apply 'cons '((+ 2 3) 4)) =>
                     ((+ 2 3) . 4) not (5 . 4)

     (apply  f  y  p)  works  like apply  with  two  arguments  except  that the
     application is  done with  the variable bindings  specified by  the binding
     context pointer p.


quote               FSUBR

     The special form (quote x) returns x without trying to evaluate  it.  quote
     is used to include constants in a form.  For convenience, the read function
     normally  converts any  S-expression preceded  by the  apostrophe  or acute
     accent character (') into the quote special form.  For example,



December 5, 1977                     ∪2-2.                              Page 2-7
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



                    (setq x '(some list))

     is converted by the reader to:

                    (setq x (quote (some list)))

     which when evaluated causes the variable  x to be set to the  constant list
     value  shown.   For more  information  on input  syntax,  see  the detailed
     discussion in part 5.1.

     quote could have been defined by:

          (defun quote fexpr (x) (car x))


function            FSUBR

     function is like quote except that its argument is a functional expression.
     To  the interpreter,  quote and  function are  identical, but  the compiler
     needs  to be  able to  distinguish between  a random  piece of  data, which
     should be left alone, and a function, which should be compiled into machine
     code.  Example:

        (mapcar (function (lambda (p q)
                             (cond ((eq p '*) q)
                                   (t (list p '= q)) )))
                first-list-of-things
                (compute-another-list) )

     calls  mapcar with  three arguments,  the first  of which  is  the function
     defined  by  the  lambda-expression.  The  actual  value  passed  to mapcar
     depends on whether the form  has been compiled.  If it is  interpreted, the
     lambda-expression  written above  will be  passed.  If  it is  compiled, an
     automatically-generated  atomic  symbol  with  the  compiled  code  for the
     lambda-expression as its subr property will be passed.  The usual  thing to
     do with functional arguments is to invoke them via apply or  funcall, which
     accept both the compiled and the interpreted functional forms.

     function makes no attempt to solve the "funarg problem."   *function should
     be used for this purpose.



Page 2-8                             ∪2-2.                      December 5, 1977
**DRAFT**                        The Evaluator                         **DRAFT**


*function           FSUBR

     The value of (*function f) is  a "funarg" of the function f.  A  funarg can
     be used like a function.  It has the additional property that it contains a
     binding context pointer so that the values of variables are bound  the same
     during  the  application of  the  funarg as  at  the time  it  was created,
     provided that the binding environment in which the funarg was created still
     exists on the stack. Hence if  foo is a function that accepts  a functional
     argument, such as

          (defun foo (f)
              (append one-value (f the-other-value) ))

          or, better

          (defun foo (f)
              (append one-value (funcall f the-other-value) ))

          then

          (foo (*function bar))

          works, but

          (foo (prog (x y z)
                  (do something)
                  (return (*function bar)) ))

     does  not if  bar intends  to reference  the prog  variables x,  y,  and z.
     *function is intended to help  solve the "funarg problem," however  it only
     works in some easy cases.  Funargs generated by *function are  intended for
     use as functional arguments and cannot be returned as values  of functional
     applications.  Thus, the user should be careful in his use of  *function to
     make  sure that  his use  does not  exceed the  limitations of  the Maclisp
     funarg mechanism.

     It is possible to assign a  value to a variable when a previous  binding of
     that variable has  been made current by  a funarg.  The assignment  will be
     executed in  the proper  context.  (This has  not always  been the  case in
     Maclisp; it is a fairly new feature.)

     A funarg has the form
                        (funarg function . context-ptr)

December 5, 1977                     ∪2-2.                              Page 2-9
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


comment             FSUBR

     comment  ignores  its  arguments and  returns  the  atomic  symbol comment.
     Example:

        (defun foo (x)
            (cond ((null x) 0)
                  (t (comment x has something in it)
                     (1+ (foo (cdr x))))))

     Usually it is preferable to comment code using the  semicolon-macro feature
     of the standard input syntax.  This allows the user to add comments  to his
     code which are ignored by the lisp reader.

     Example:

        (defun foo (x)
            (cond ((null x) 0)
                  (t (1+ (foo (cdr x))))     ;x has something in it
              ))

     A  problem  with such  comments  is that  they  are discarded  when  the S-
     expression is read into lisp.  If it is edited within lisp and printed back
     into  a file,  the comments  will be  lost.  However,  most users  edit the
     original file and read the changes into lisp, since this allows them to use
     the editor of their choice.  Thus this is not a real problem.


prog2               LSUBR 2 or more args

     The expressions in a prog2 form are evaluated from left to right, as in any
     lsubr-form.  The  result is  the second argument.   prog2 is  most commonly
     used to evaluate an expression with side effects, then return a value which
     needs to be computed before the side effects happen.
     Examples:

            (prog2 (do-this) (do-that)) ;just get 2 things evaluated

            (setq x (prog2 nil y        ;parallel assignment
                           (setq y x))) ;which exchanges x and y

            (defun prog2 nargs (arg 2)) ;a lexpr definition for prog2


Page 2-10                            ∪2-2.                      December 5, 1977
**DRAFT**                        The Evaluator                         **DRAFT**


progn               LSUBR 1 or more args

     The expressions in a progn form are evaluated from left to right, as usual,
     and the result is the value of  the last one.  In other words, progn  is an
     lsubr which does  nothing but return  its last argument.   Although lambda-
     expressions, prog-forms, do-forms, cond-forms, and iog-forms all  use progn
     implicitly, that is, they allow  multiple forms in their bodies,  there are
     occasions when one needs to evaluate a number of forms for side-effects and
     make them appear to be a single form.  progn serves this purpose. Example:

          (progn (setq a (cdr frob)) (eq (car a) (cadr a)))

          might be used as the antecedent of a cond clause.

        progn could have been defined by:

        (defun progn nargs
            (and (> nargs 0)
                 (arg nargs)))


progv               FSUBR

     progv is a special form to provide the user with extra control over lambda-
     binding.   It binds  a list  of variables  to a  list of  values,  and then
     evaluates  some forms.   The  lists of  variables and  values  are computed
     quantities; this is what makes progv different from lambda, prog, and do.

                  (progv var-list value-list form1 form2 ... )

     first evaluates var-list and  value-list.  Then the variables are  bound to
     the values.   In compiled  code the  variables must  be special,  since the
     compiler has no way of  knowing what symbols might appear in  the var-list.
     If too few values are  supplied, the remaining variables are bound  to nil.
     If too many values are supplied, the excess values are ignored.

     After the variables have been bound to the values, the forms are evaluated,
     and finally the variable bindings  are undone.  The result returned  is the
     value of the last form.  Note that the "body" of a progv is similar to that
     of progn, not that of prog.
     Example:



December 5, 1977                     ∪2-2.                             Page 2-11
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



        (setq a 'foo b 'bar)

        (progv (list a b 'b) (list b) (list a b foo bar))
            => (foo nil bar nil)

     During the evaluation of the body  of this progv, foo is bound to  bar, bar
     is bound to nil, b is bound to nil, and a remains bound to foo.


arg                 SUBR 1 arg

     (arg nil),  when evaluated  during the  application of  a lexpr,  gives the
     number of arguments supplied to that lexpr.  This is primarily  a debugging
     aid, since lexprs  also receive their number  of arguments as the  value of
     their lambda-variable.

     (arg i), when evaluated during the application of a lexpr, gives  the value
     of the i'th argument to the lexpr.  i must be a fixnum in this case.  It is
     an  error if  i is  less than  1 or  greater than  the number  of arguments
     supplied to the lexpr.

     Example:

        (defun foo nargs            ;define a lexpr foo.
            (print (arg 2))         ;print the second argument.
            (+ (arg 1)              ;return the sum of the first
               (arg (- nargs 1))))  ;and next to last arguments.


setarg              SUBR 2 args

     setarg is used only during the  application of a lexpr.  (setarg i  x) sets
     the  lexpr's i'th  argument to  x.  i  must be  greater than  zero  and not
     greater than the number of arguments passed to the lexpr.  After  (setarg i
     x) has been done, (arg i) will return x.








Page 2-12                            ∪2-2.                      December 5, 1977
**DRAFT**                        The Evaluator                         **DRAFT**


listify             SUBR 1 arg

     (listify n)  efficiently manufactures  a list of  n of  the arguments  of a
     lexpr.   With a  positive argument  n, it  returns a  list of  the  first n
     arguments of the lexpr.  With a  negative argument n, it returns a  list of
     the last (abs n) arguments of the lexpr.  Basically, it works as if defined
     as follows:

            (defun listify (n)
                 (cond ((minusp n)
                        (listify1 (arg nil) (+ (arg nil) n 1)))
                       (t
                        (listify1 n 1)) ))

            (defun listify1 (n m)      ; auxiliary function.
                 (do ((i n (1- i))
                      (result nil (cons (arg i) result)))
                     ((< i m) result) ))


funcall             LSUBR 1 or more args

     (funcall f a1 a2  ... an) calls the function  f with the arguments  a1, a2,
     ..., an.   It is similar  to apply except  that the separate  arguments are
     given to funcall, rather than a list  of arguments.  If f is a fexpr  or an
     fsubr there must be exactly one argument.  f may not be a macro.  Example:

                (setq cons 'plus)
                (cons 1 2) => (1 . 2)
                (funcall cons 1 2) => 3


subrcall            FSUBR

     subrcall  is  used  to  invoke  a  subr-pointer  directly,  rather  than by
     referring  to  an atomic  symbol  of  which the  subr-pointer  is  the subr
     property.  The form is:

                         (subrcall type p a1 a2 ... an)

     All arguments except the first  are evaluated.  type is the type  of result
     expected: fixnum, flonum, or nil (any  type).  p is the subr pointer  to be


December 5, 1977                     ∪2-2.                             Page 2-13
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     called.   a1  through  an are  the  arguments  to be  passed  to  the subr.
     subrcall compiles into efficient machine code.


lsubrcall           FSUBR

     lsubrcall is identical to subrcall except that the subr-pointer  called has
     to be an lsubr instead of a subr.  This is because many Lisps use different
     internal calling sequences for lsubrs than for subrs.


arraycall           FSUBR

     arraycall is similar to subrcall and lsubrcall except that an array-pointer
     is used instead  of a subr-pointer.  The  first argument of  arraycall must
     correspond to the type  that the array was  given when it was  created.  An
     arraycall expression may be used as the first argument to store.



























Page 2-14                            ∪2-2.                      December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**




3.  Manipulating List Structure


3.1  Conses


car                 SUBR 1 arg

     Returns the first component of a cons.

     Example:  (car '(a b)) => a


cdr                 SUBR 1 arg

     Returns the second component of a cons.

     Example:  (cdr '(a b c)) => (b c)


car                 SWITCH

cdr                 SWITCH

     Officially car and cdr are only applicable to lists.  However, as  a matter
     of convenience the car and cdr of nil are nil.  This allows programs to car
     and cdr off the ends of  lists without having to check, which  is sometimes
     helpful.  Furthermore, some old programs apply car and cdr to objects other
     than lists in order to  hack with the internal representation.   To provide
     control over this, the value of car can be set to control which  data types
     are subject to the car operation.  Similarly, the value of cdr controls the
     cdr  operation.   Illegal operations  will  cause errors.   For  reasons of
     efficiency, this error checking is only enabled in (*rset t) mode (see page
     3-28) and is mostly turned  off in compiled programs.  The values  to which
     the switches may be set are:







December 5, 1977                     ∪2-3.                             Page 2-15
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



          Value      Operation applicable to

          list       lists.
          nil        lists and nil.
          symbol     lists, nil, and symbols.
          t          anything.

     The default value of the switches is nil.


c...r               SUBR 1 arg

     All the compositions of up to four car's and cdr's are defined as functions
     in their own right.  The names begin with c and end with r, and  in between
     is a sequence of a's and d's corresponding to the composition  performed by
     the function.

     For example,
                     (cddadr x) = (cdr (cdr (car (cdr x))))

     Some  of the  most commonly  used  ones are:  cadr, which  gets  the second
     element of a list; caddr, which  gets the third element of a  list; cadddr,
     which gets the fourth element of a list; caar, to car twice.

     The car'ing and cdr'ing  operations of these functions have  error checking
     under the control of the car and cdr switches explained above, just  as the
     car and cdr functions themselves do.


cons                SUBR 2 args

     This is a primitive  function to construct a  new dotted pair whose  car is
     the first argument to cons, and  whose cdr is the second argument  to cons.
     Thus the following  identities hold (except  when numbers are  involved; as
     always numbers are not well-behaved with respect to eq):

           (eq (car (cons x y)) x) => t
           (eq (cdr (cons x y)) y) => t





Page 2-16                            ∪2-3.1                     December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**



        Examples:
            (cons 'a 'b) => (a . b)
            (cons 'a (cons 'b (cons 'c nil))) => (a b c)
            (cons 'a '(b c d e f)) => (a b c d e f)


ncons               SUBR 1 arg

     (ncons x) = (cons x nil) = (list x)


xcons               SUBR 2 args

     xcons ("exchange cons") is like cons except that the order of  arguments is
     reversed.

     Example:
                            (xcons 'a 'b) => (b . a)

























December 5, 1977                     ∪2-3.1                            Page 2-17
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


3.2  Lists


last                SUBR 1 arg

     last returns the last cons of the list which is its argument.

        Example:
                (setq x '(a b c d))
                (last x) => (d)
                (rplacd (last x) '(e f))
                x => (a b c d e f)

     last could have been defined by:

        (defun last (x)
            (cond ((null x) x)
                  ((null (cdr x)) x)
                  ((last (cdr x))) ))

     In some implementations,  the null check above  may be replaced by  an atom
     check, which  will catch  dotted lists.   Code which  depends on  this fact
     should not be  written though, because  all implementations are  subject to
     change on this point.


length              SUBR 1 arg

     length  returns the  length of  its argument,  which must  be a  list.  The
     length of a list is the number of top-level conses in it.

        Examples:
                (length nil) => 0
                (length '(a b c d)) => 4
                (length '(a (b c) d)) => 3

     length could have been defined by:







Page 2-18                            ∪2-3.2                     December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**



        (defun length (x)
            (cond ((null x) 0)
                  ((1+ (length (cdr x)))) ))
        or by:

        (defun length (x)
            (do ((n 0 (1+ n))
                 (y x (cdr y)))
                ((null y) n) ))

     The warning about dotted lists given under last applies also to length.


list                LSUBR 0 or more args

     list constructs and returns a list of its arguments.

     Example:
              (list 3 4 'a (car '(b . c)) (+ 6 -2)) => (3 4 a b 4)

     list could have been defined by:

        (defun list nargs
            (do ((n nargs (1- n))
                 (s nil (cons (arg n) s)))
                ((zerop n) s) ))
     (This depends on parallel assignment to the control variables of do.)


append              LSUBR 0 or more args

     The arguments  to append  are lists.   The result  is a  list which  is the
     concatenation of the arguments.  The arguments are not changed (cf. nconc).
     For example,

             (append '(a b c) '(d e f) nil '(g)) => (a b c d e f g)

     To make a copy of  the top level of a list,  that is, to copy the  list but
     not its elements, use (append x nil).




December 5, 1977                     ∪2-3.2                            Page 2-19
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     A  version of  append  which only  accepts  two arguments  could  have been
     defined by:

        (defun append2 (x y)
            (cond ((null x) y)
                  ((cons (car x) (append2 (cdr x) y)) )))

     The generalization to  any number of arguments  could then be made  using a
     lexpr:

        (defun append argcount
            (do ((i (1- argcount) (1- i))
                 (val (arg argcount) (append2 (arg i) val)))
                ((zerop i) val) ))


reverse             SUBR 1 arg

     Given a list as argument, reverse creates a new list whose elements are the
     elements of its argument taken  in reverse order.  reverse does  not modify
     its argument, unlike nreverse which is faster but does modify its argument.
     Example:

                   (reverse '(a b (c d) e)) => (e (c d) b a)

        reverse could have been defined by:

        (defun reverse (x)
            (do ((l x (cdr l))         ; scan down argument,
                 (r nil                ; putting each element
                    (cons (car l) r))) ; into list, until
                ((null l) r)))         ; no more elements.


nconc               LSUBR 0 or more args

     nconc takes lists as arguments.   It returns a list which is  the arguments
     concatenated  together.  The  arguments  are changed,  rather  than copied.
     (cf. append)

     Example:
                   (nconc '(a b c) '(d e f)) => (a b c d e f)


Page 2-20                            ∪2-3.2                     December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**


     Note that the constant (a b c) has  now been changed to (a b c d e  f).  If
     this form is evaluated again, it will yield (a b c d e f d e f).  This is a
     danger you always have to watch out for when using nconc.

     nconc could have been defined by:

        (defun nconc (x y)      ;for simplicity, this definition
            (cond ((null x) y)  ;only works for 2 arguments.
                  (t (rplacd (last x) y) ;hook y onto x
                      x)))      ;and return the modified x.


nreverse            SUBR 1 arg

     nreverse reverses its  argument, which should be  a list.  The  argument is
     destroyed by rplacd's all through the list (cf. reverse).

     Example:
                         (nreverse '(a b c)) => (c b a)

        nreverse could have been defined by:

        (defun nreverse  (x)
            (cond ((null x) nil)
                  ((nreverse1 x nil))))

        (defun nreverse1 (x y)          ;auxiliary function
            (cond ((null (cdr x)) (rplacd x y))
                  ((nreverse1 (cdr x) (rplacd x y)))))
                  ;; this last call depends on order of argument evaluation.


nreconc             SUBR 2 args

     (nreconc x y) is exactly the same as (nconc (nreverse x) y) except  that it
     is more efficient.

     nreconc could have been defined by:

        (defun nreconc (x y)
            (cond ((null x) y)
                  ((nreverse1 x y)) ))


December 5, 1977                     ∪2-3.2                            Page 2-21
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     using the same nreverse1 as above.











































Page 2-22                            ∪2-3.2                     December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**


3.3  Alteration of List Structure


   The functions  rplaca and  rplacd are  used to  make alterations  in already-
existing list structure.   The structure is  not copied but  physically altered;
hence caution should  be exercised when using  these functions as  strange side-
effects can occur if portions of list structure become shared unbeknownst to the
programmer.  The nconc, nreverse,  and nreconc functions already  described have
the same property.   However, they are normally  not used for  this side-effect;
rather, the list-structure modification is purely for efficiency  and compatible
non-modifying functions are provided.

rplaca              SUBR 2 args

     (rplaca  x y)  changes the  car of  x to  y and  returns (the  modified) x.
     Example:

            (setq g '(a b c))

            (rplaca (cdr g) 'd) => (d c)

     Now g => (a d c)


rplacd              SUBR 2 args

     (rplacd  x y)  changes the  cdr of  x to  y and  returns (the  modified) x.
     Example:

            (setq x '(a b c))

            (rplacd x 'd) => (a . d)

     Now x => (a . d)

     See also setplist (page 2-55).








December 5, 1977                     ∪2-3.3                            Page 2-23
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


subst               SUBR 3 args

     (subst x y z) substitutes x for all occurrences of y in z, and  returns the
     modified copy  of z.   The original  z is  unchanged, as  subst recursively
     copies all of z replacing elements eq to y as it goes.  If x and y are nil,
     z  is  just  copied, which  is  a  convenient way  to  copy  arbitrary list
     structure.

     Example:

        (subst 'Tempest 'Hurricane
               '(Shakespeare wrote (The Hurricane)))
            => (Shakespeare wrote (The Tempest))

     subst could have been defined by:

        (defun subst (x y z)
            (cond ((eq z y) x)    ;if item eq to y, replace.
                  ((atom z) z)    ;if no substructure, return arg.
                  ((cons (subst x y (car z))  ;otherwise recurse.
                         (subst x y (cdr z))))))


sublis              SUBR 2 args

     sublis  makes substitutions  for atomic  symbols in  an  S-expression.  The
     first argument  to sublis is  an association list  (see the  next section).
     The second argument  is the S-expression in  which substitutions are  to be
     made.  sublis looks at all atomic symbols in the S-expression; if an atomic
     symbol appears in  the association list occurrences  of it are  replaced by
     the object it is associated with.  The argument is not modified; new conses
     are created where necessary and only where necessary, so the  newly created
     structure shares as much of its substructure as possible with the old.  For
     example, if  no substitutions  are made,  the result  is eq  to the  old S-
     expression.
     Example:

            (sublis '((x . 100) (z . zprime))
                    '(plus x (minus g z x p) 4))
               => (plus 100 (minus g zprime 100 p) 4)

     In some implementations sublis works by putting temporary sublis properties
     on the atomic symbols in the dotted pairs, so beware.

Page 2-24                            ∪2-3.3                     December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**


3.4  Tables


   Maclisp includes several functions which simplify the maintenance  of tabular
data structures of  several varieties.  The simplest  is a plain list  of items,
which models (approximately) the concept  of a set.  There are functions  to add
(cons), remove (delete, delq), and search for (member, memq) items in a list.

   Association lists are very commonly  used.  An association list is a  list of
dotted pairs.   The car  of each pair  is a  "key" and the  cdr is  "data".  The
functions assoc and assq may be used to retrieve the data, given the key.

   Structured records can  be stored as association  lists or as  stereotyped S-
expressions  where each  element of  the structure  has a  certain  car-cdr path
associated with it.  There  are no built-in functions  for these but it  easy to
define macros to implement them (see part 6.2).

   Simple list-structure is very convenient, but may not be efficient enough for
large data bases because  it takes a long time  to search a long  list.  Maclisp
includes some hashing functions  (sxhash, maknum) which aid in  the construction
of more efficient, hairier structures.


member              SUBR 2 args

     (member x y) returns nil if x is not a member of the list y.  Otherwise, it
     returns the  portion of y  beginning with the  first occurrence of  x.  The
     comparison is made by equal.  y is searched on the top level only.

        Example:
                (member 'x '(1 2 3 4)) => nil
                (member 'x '(a (x y) c x d e x f)) => (x d e x f)


     Note that the  value returned by  member is eq to  the portion of  the list
     beginning with x.  Thus rplaca on the result of member may be used,  if you
     first check to make sure member did not return nil.
     Example:

        (catch (rplaca (or (member x z)
                           (throw nil lose))
                       y)
               lose)

December 5, 1977                     ∪2-3.4                            Page 2-25
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     member could have been defined by:

        (defun member (x y)
            (cond ((null y) nil)
                  ((equal x (car y)) y)
                  ((member x (cdr y))) ))


memq                SUBR 2 args

     memq  is like  member, except  eq is  used for  the comparison,  instead of
     equal.  memq could have been defined by:

        (defun memq (x y)
            (cond ((null y) nil)
                  ((eq x (car y)) y)
                  ((memq x (cdr y))) ))


delete              LSUBR 2 or 3 args

     (delete  x  y) returns  the  list y  with  all top-level  occurrences  of x
     removed.  equal  is used for  the comparison.  The  argument y  is actually
     modified (rplacd'ed) when instances of x are spliced out.  delete should be
     used for value, not for effect.  That is, use

        (setq a (delete 'b a))

        rather than

        (delete 'b a))


        The latter is not equivalent when the first element of the value of a is
        b.

     (delete x y n) is like (delete x y) except only the first n instances  of x
     are deleted.  n is allowed to be zero.  If n is greater than the  number of
     occurrences of  x in the  list, all occurrences  of x in  the list  will be
     deleted.

     Example:


Page 2-26                            ∪2-3.4                     December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**



        (delete 'a '(b a c (a b) d a e)) => (b c (a b) d e)

        delete could have been defined by:

        (defun delete nargs        ; lexpr for 2 or 3 args
            (delete1 (arg 1)       ; pass along arguments...
                     (arg 2)
                     (cond ((= nargs 3) (arg 3))
                           (123456789.)))) ; infinity

        (defun delete1 (x y n)       ;auxiliary function
            (cond ((or (null y) (zerop n)) y)
                  ((equal x (car y)) (delete1 x
                                              (cdr y)
                                              (1- n)))
                  ((rplacd y (delete1 x (cdr y) n)))))


delq                LSUBR 2 or 3 args

     delq  is the  same as  delete except  that eq  is used  for  the comparison
     instead of equal.


sxhash              SUBR 1 arg

     sxhash computes a hash code of an S-expression, and returns it as a fixnum,
     which may be positive or negative.   A property of sxhash is that  (equal x
     y) implies  (= (sxhash x)  (sxhash y)).  The  number returned by  sxhash is
     some  possibly  large  number  in the  range  allowed  by  fixnums.   It is
     guaranteed that:

     1) sxhash for an atomic symbol will always be positive.

     2) sxhash  of any particular  expression will be  constant in  a particular
     implementation for all time, probably.

     3)  Two  different  implementations  may  hash  the  same  expression  into
     different values.

     4) sxhash of any object of type random will be zero.


December 5, 1977                     ∪2-3.4                            Page 2-27
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     5) sxhash of a fixnum will = that fixnum.

        Here is an example of how to use sxhash in maintaining
        hash tables of S-expressions:

     (defun knownp (x)    ;look up x in the table
       (prog (i bkt)
         (setq i (plus 76 (remainder (sxhash x) 77)))
           ;The remainder should be reasonably randomized between
           ;-76 and 76, thus table size must be > 175 octal.
         (setq bkt (table i))
           ;bkt is thus a list of all those expressions that hash
           ;into the same number as does x.
         (return (member x bkt))))

     To write an "intern" for S-expressions, one could

     (defun sintern (x)
       (prog (bkt i tem)
         (setq bkt (table (setq i (+ 2n-2 (\ (sxhash x) 2n-1)))))
             ;2n-1 and 2n-1 stand for a power of 2 minus one and
             ;minus two respectively.  This is a good choice to
             ;randomize the result of the remainder operation.
         (return (cond ((setq tem (member x bkt))
                        (car tem))
                       (t (store (table i) (cons x bkt))
                          x)))))


assoc               SUBR 2 args

     (assoc x y) looks  up x in the association  list (list of dotted  pairs) y.
     The value is the first dotted pair whose car is equal to x, or nil if there
     is none such.

        Examples:
             (assoc 'r '((a . b) (c . d) (r . x) (s . y) (r . z)))
                =>  (r . x)

             (assoc 'fooo '((foo . bar) (zoo . goo))) => nil

     It is okay to rplacd the result of assoc as long as it is not nil,  if your
     intention is to "update" the "table" that was assoc's second argument.

Page 2-28                            ∪2-3.4                     December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**



        Example:
            (setq values '((x .  100) (y  . 200) (z .  50)))
            (assoc 'y values) => (y . 200)
            (rplacd (assoc 'y values)  201)
            (assoc 'y values)  => (y . 201) now
               (One should always be careful about using rplacd however)

     A typical  trick is  to say (cdr  (assoc x  y)).  Since the  cdr of  nil is
     guaranteed to be nil, this yields nil if no pair is found (or if a  pair is
     found whose cdr is nil.)

        assoc could have been defined by:

        (defun assoc (x y)
            (cond ((null y) nil)
                  ((equal x (caar y)) (car y))
                  ((assoc x (cdr y))) ))


assq                SUBR 2 args

     assq is  like assoc except  that the comparison  uses eq instead  of equal.
     assq could have been defined by:

        (defun assq (x y)
            (cond ((null y) nil)
                  ((eq x (caar y)) (car y))
                  ((assq x (cdr y))) ))


sassoc              SUBR 3 args

     (sassoc x y  z) is like  (assoc x y) except  that if x  is not found  in y,
     instead of  returning nil sassoc  calls the function  z with  no arguments.
     sassoc could have been defined by:

        (defun sassoc (x y z)
            (or (assoc x y)
                (apply z nil)))

     sassoc  and sassq  (see below)  are of  limited use.   These  are primarily
     leftovers from Lisp 1.5.

December 5, 1977                     ∪2-3.4                            Page 2-29
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


sassq               SUBR 3 args

     (sassq x  y z)  is like (assq  x y)  except that if  x is  not found  in y,
     instead of  returning nil  sassq calls  the function  z with  no arguments.
     sassq could have been defined by:

        (defun sassq (x y z)
            (or (assq x y)
                (apply z nil)))


maknum              SUBR 1 arg

     (maknum x) returns a positive fixnum which is unique to the object  x; that
     is, (maknum x) and  (maknum y) are numerically equal  if and only if  (eq x
     y).  This can be used in hashing.

     In the  pdp-10 implementations,  maknum returns the  memory address  of its
     argument.   In  the  Multics  implementation,  an  internal  hash  table is
     employed.

     Note  that unlike  sxhash, maknum  will  not return  the same  value  on an
     expression which has been printed out and read back in again.


munkam              SUBR 1 arg

     munkam is the  opposite of maknum.  Given  a number, it returns  the object
     which was given to maknum to  get that number.  It is inadvisable  to apply
     munkam to a number which did not come from maknum.














Page 2-30                            ∪2-3.4                     December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**


3.5  Sorting


   Several functions are provided for sorting arrays and lists.  These functions
use algorithms which always terminate no matter what sorting predicate  is used,
provided  only that  the predicate  always terminates.   The array  sort  is not
necessarily stable, that  is equal items may  not stay in their  original order.
However the list sort is stable.

   After sorting, the argument (be it list or array) is rearranged internally so
as  to  be completely  ordered.   In the  case  of an  array  argument,  this is
accomplished by permuting the elements of the array, while in the list case, the
list is  reordered by  rplacd's in  the same  manner as  nreverse.  Thus  if the
argument should not  be clobbered, the  user must sort  a copy of  the argument,
obtainable by fillarray or append, as appropriate.

   Should the comparison predicate cause an error, such as a wrong type argument
error, the state of  the list or array  being sorted is undefined.   However, if
the error is corrected the sort will, of course, proceed correctly.

   Both sort and sortcar handle the  case in which their second argument  is the
function alphalessp in a more  efficient manner than usual.  This  efficiency is
primarily due to elimination of argument checks at comparison time.


sort                SUBR 2 args

     The first argument to sort is an array (or list), the second a predicate of
     two arguments.  Note that a "number array" cannot be sorted.  The predicate
     must be applicable to all the objects in the array or list.   The predicate
     should take  two arguments,  and return non-nil  if and  only if  the first
     argument is strictly less than the second (in some appropriate sense).

     The sort function proceeds to sort the contents of the array or  list under
     the  ordering imposed  by  the predicate,  and  returns the  array  or list
     modified into sorted  order, i.e. its  modified first argument.   Note that
     since  sorting  requires  many  comparisons, and  thus  many  calls  to the
     predicate,  sorting will  be much  faster if  the predicate  is  a compiled
     function rather than interpreted.





December 5, 1977                     ∪2-3.5                            Page 2-31
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



        Example:

        (defun mostcar (x)
            (cond ((atom x) x)
                  ((mostcar (car x)))))

        (sort 'fooarray
              (function (lambda (x y)
                (alphalessp (mostcar x) (mostcar y)))))

     If fooarray contained these items before the sort:

        (Tokens (The lion sleeps tonight))
        (Carpenters (Close to you))
        ((Rolling Stones) (Brown sugar))
        ((Beach Boys) (I get around))
        (Beatles (I want to hold your hand))

     then after the sort fooarray would contain:

        ((Beach Boys) (I get around))
        (Beatles (I want to hold your hand))
        (Carpenters (Close to you))
        ((Rolling Stones) (Brown sugar))
        (Tokens (The lion sleeps tonight))


sortcar             SUBR 2 args

     sortcar is  exactly like sort,  but the  items in the  array or  list being
     sorted should all be non-atomic.  sortcar takes the car of each item before
     handing two items to the predicate.   Thus sortcar is to sort as  mapcar is
     to maplist.










Page 2-32                            ∪2-3.5                     December 5, 1977
**DRAFT**                 Manipulating List Structure                  **DRAFT**


3.6  Hunks


   This section applies only to the pdp10 implementation.

   Hunks are a generalization  of conses, useful in constructing  more efficient
data structures.  A hunk is like  a cons but it has more components;  hunks come
in several  convenient sizes.   The advantage of  an n-element  hunk over  an n-
element list is that the hunk occupies less space (half as much if n is  a power
of 2).   The elements  of a  hunk can  be referenced  more efficiently  than the
elements  of a  list, since  the compiler  knows the  relative locations  of the
components and addresses them directly.

   The advantage of  lists over hunks is  flexibility; lists can be  any length,
can vary  in length, can  be altered by  rplacd, and can  be manipulated  with a
library  of  useful  searching, sorting,  and  combining  operations, previously
described in this chapter.

   Another feature of  hunks is that at  times one may treat  a hunk as  a cons,
ignore the  extra components.   This allows the  construction of  list structure
which has extra "frobs" stuck on at certain points.  The atom function  does not
consider hunks to be atomic; it returns nil if given a hunk.

   print represents hunks using  an extended form of dot-notation;  read however
does not yet understand this notation.  See the writeup on print.


hunk                LSUBR 0 or more args

     hunk takes any number of arguments, and returns a hunk whose components are
     the arguments.  The  first argument is  the car, and  the last is  the cdr;
     that is, the arguments are in the order 1, 2, 3, ..., N-1, 0.  This  is the
     same order as print and makhunk use.

     The  maximum  size  of  a  hunk is  128  components.   This  may  vary from
     implementation to implementation.

     With  no arguments,  hunk returns  nil.  With  one or  two  arguments, hunk
     returns a cons.





December 5, 1977                     ∪2-3.6                            Page 2-33
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


cxr                 SUBR 2 args

     (cxr n h) returns the n'th component of the hunk h.  car of a  hunk returns
     the 1st component, and cdr of a hunk returns the 0th component.


rplacx              SUBR 3 args

     (rplacx n h z) replaces the n'th component of the hunk h with z.  The value
     of rplacx is its (modified) second argument.  rplaca of a hunk replaces its
     1st component, and rplacd of a hunk replaces its 0th component.


makhunk             SUBR 1 arg

     (makhunk n), where  n is a fixnum,  creates and returns an  n-element hunk,
     filled with nils.  (makhunk  l), where l is  a list, creates and  returns a
     hunk of the  appropriate length, initialized from  l.  This is  like (apply
     'hunk l).

     Like hunk, makhunk will return nil or a cons if you ask for a hunk of 0, 1,
     or 2 elements.


hunksize            SUBR 1 arg

     hunksize returns the number of components in its argument.  hunksize of nil
     is 0 and hunksize of a cons is 2.


hunkp               VARIABLE

     If the value of hunkp is nil, the functions print, equal, and purcopy treat
     hunks as conses, as most other system functions do.  The extra elements are
     simply ignored.  If  the value of hunkp  is non-nil, which is  the default,
     these three functions deal with all the elements.  sxhash always deals with
     all the elements.







Page 2-34                            ∪2-3.6                     December 5, 1977
**DRAFT**                       Flow of Control                        **DRAFT**




4.  Flow of Control


   Maclisp provides a variety of structures for flow of control.

   Functional application is the basic method for construction of programs.  All
operations are written as the  application of a function to  arguments.  Maclisp
programs  are often  written  as a  large  collection of  small  functions which
implement simple operations.   Some of the functions  work by calling  others of
the functions, thus defining some operations in terms of others.

   Recursion  exists  when  a  function  calls  itself.   This  is  analogous to
mathematical induction.

   Iteration is a control structure present in most languages.  It is similar to
recursion but sometimes less useful and sometimes more useful.  Maclisp contains
a generalized iteration facility.  The iteration facility also permits those who
like "gotos" to use them.

   Conditionals allow control to branch  depending on the value of  a predicate.
and  and or  are basically  one-arm conditionals,  while cond  is  a generalized
multi-armed conditional.

   Nonlocal  exits are  similar to  a  return, except  that the  return  is from
several levels of  function calling rather than  just one, and is  determined at
run time.   These are  mostly used for  applications such  as escaping  from the
middle of a function when it is discovered that the algorithm is not applicable.

   Errors are  a type of  non-local exit  used by the  Lisp interpreter  when it
discovers a condition that it does not like.  Errors have the additional feature
of correctability, which  allows a user-specified  function (most often  a break
loop), to get a chance to come in and correct the error or at least inspect what
was happening  and determine what  caused it, before  the nonlocal  exit occurs.
This is explained in detail on part 3.4.

   Maclisp does not directly provide "hairy control structure" such  as multiple
processes, backtracking, or continuations.





December 5, 1977                     ∪2-4.                             Page 2-35
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


4.1  Conditionals


and                 FSUBR

     (and form1 form2...) evaluates the forms one at a time, from left to right.
     If  any  form  evaluates  to  nil,  and  immediately  returns  nil  without
     evaluating the  remaining forms.   If all the  forms evaluate  non-nil, and
     returns  the value  of the  last one.   and can  be used  both  for logical
     operations, where  nil stands for  False and  t stands for  True, and  as a
     conditional expression.

        Examples:
             (and x y)

             (and (setq temp (assq x y))
                  (rplacd temp z))

             (and (null (errset (something)))
                  (princ "There was an error."))

     Note: (and) => t, which is the identity for this operation.


or                  FSUBR

     (or form1 form2...) evaluates the forms one by one from left to  right.  If
     a form evaluates to nil, or  proceeds to evaluate the next form.   If there
     are no more  forms, or returns  nil.  But if  a form evaluates  non-nil, or
     immediately returns that value without evaluating any remaining  forms.  or
     can be used both for logical  operations, where nil stands for False  and t
     for True, and as a conditional expression.

     Note:  (or) => nil, the identity for this operation.


cond                FSUBR

     The  cond  special form  consists  of  the word  cond  followed  by several
     clauses.  Each  clause consists  of a  predicate followed  by zero  or more
     forms.  Sometimes the predicate is called the antecedent and the  forms are
     called the consequents.


Page 2-36                            ∪2-4.1                     December 5, 1977
**DRAFT**                       Flow of Control                        **DRAFT**



        (cond (antecedent consequent consequent...)
              (antecedent ...)
              ... )

     The idea is  that each clause  represents a case  which is selected  if its
     predicate is satisfied and the predicates of all preceding clauses  are not
     satisfied.  When a case is selected, its consequent forms are evaluated.

     cond  processes  its  clauses  in order  from  left  to  right.   First the
     predicate of the current clause  is evaluated.  If the result is  nil, cond
     advances to the next clause.   Otherwise, the cdr of the clause  is treated
     as a list of forms, or consequents, which are evaluated in order  from left
     to  right.   After   evaluating  the  consequents,  cond   returns  without
     inspecting any remaining  clauses.  The value of  the cond special  form is
     the value of the last consequent evaluated, or the value of  the antecedent
     if there were no consequents in  the clause.  If cond runs out  of clauses,
     that is, if every antecedent is  nil, that is, if no case is  selected, the
     value of the cond is nil.

        Example:
            (cond ((zerop x)    ;First clause:
                   (+ y 3))     ; (zerop x) is antecedent.
                                ; (+ y 3) is consequent.
                  ((null y)     ;A clause with 2 consequents:
                   (setq x 4)   ; this
                   (cons x z))  ; and this.
                  (z)           ;A clause with no consequents:
                                ; the antecedent is just z.
               )                ;This is the end of the cond.

     This is like the traditional Lisp 1.5 cond except that it is  not necessary
     to have exactly one consequent in each clause, and it is permissible to run
     out of clauses.










December 5, 1977                     ∪2-4.1                            Page 2-37
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


4.2  Iteration


prog                FSUBR

     prog  is  the "program"  special  form.  It  provides  temporary variables,
     sequential evaluation of statements, and the ability to do "gotos."  A prog
     looks something like:

        (prog (var1 var2...)
         tag1
            statement1
            statement2
         tag2
            statement3
            . . .
          )

     var1, var2,  ... are  temporary variables.   When the  prog is  entered the
     values of  these variables  are saved.  When  the prog  is exited  they are
     restored.  The variables are initialized  to nil when the prog  is entered,
     thus they are said  to be "bound to  nil" by the prog.   However, variables
     which have been declared fixnum or  flonum will be initialized to 0  or 0.0
     instead,  but  only in  compiled  programs.  You  should  be  careful about
     relying on the initial value of prog-variables.

     The part of a prog after the temporary variable list is the body.   An item
     in the body may be an atomic symbol or a number, which is a tag, or  a non-
     atomic form, which is a statement.

     prog,   after  binding   the  temporary   variables,  processes   its  body
     sequentially.  tags  are skipped  over;  statements  are evaluated  but the
     values are ignored.  If the end  of the body is reached, prog  returns nil.
     If (return x) is evaluated, prog stops processing its body and  returns the
     value x.   If (go tag)  is evaluated, prog  jumps to the  part of  the body
     labelled with the tag.   The argument to go  is not evaluated unless  it is
     non-atomic.

     It should be noted  that the Maclisp prog is  an extension of the  Lisp 1.5
     prog, in  that go's and  return's may  occur in more  places than  Lisp 1.5
     allowed.  However, the Lisp compilers implemented on ITS, Multics,  and the
     DECsystem 10 for Maclisp require that go's and return's be lexically within


Page 2-38                            ∪2-4.2                     December 5, 1977
**DRAFT**                       Flow of Control                        **DRAFT**


     the scope of the prog. This makes a function which does not contain a prog,
     but which does contain a go or return uncompilable.

     See also the do special form,  which uses a body similar to prog.   The do,
     catch, and  throw special forms  are included in  Maclisp as an  attempt to
     encourage goto-less programming style,  which leads to more  readable, more
     easily  maintained  code.   The  programmer  is  recommended  to  use these
     functions instead of prog wherever reasonable.

     Example:

        (prog (x y z)  ;x, y, z are prog variables  - temporaries.
           (setq y (car w) z (cdr w))     ;w is a free variable.
        loop
           (cond ((null y) (return x))
                 ((null z) (go err)))
        rejoin
           (setq x (cons (cons (car y) (car z))
                         x))
           (setq y (cdr y)
                 z (cdr z))
           (go loop)
        err
           (break are-you-sure? t)
           (setq z y)
           (go rejoin))


do                  FSUBR

     The do  special form  provides a  generalized "do  loop" facility,  with an
     arbitrary number of "index variables" whose values are saved when the do is
     entered and restored when it is  left, i.e. they are bound by the  do.  The
     index  variables  are  used  in the  iteration  performed  by  do.   At the
     beginning they are initialized to specified values, and then at the  end of
     each trip  around the loop  the values of  the index variables  are changed
     according  to  specified rules.   do  allows the  programmer  to  specify a
     predicate which determines when the iteration will terminate.  The value to
     be returned as the result of the form may optionally be specified.

     do comes in two varieties.



December 5, 1977                     ∪2-4.2                            Page 2-39
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     The newer variety of do looks like:

        (do ((var init repeat)...)
            (end-test exit-form...)
            body...)

     The  first item  in the  form  is a  list of  zero or  more  index variable
     specifiers.   Each index  variable specifier  is a  list of  the name  of a
     variable var,  an initial value  init, which defaults  to nil  (or possibly
     zero, as mentioned under prog) if it is omitted, and a repeat value repeat.
     If repeat is omitted, the var is not changed between loops.

     All  assignment  to  the  index variables  is  done  in  parallel.   At the
     beginning of  the first iteration,  all the inits  are evaluated,  then the
     vars are saved, then the vars  are setq'ed to the values of the  inits.  To
     put it another way, the vars  are lambda-bound to the values of  the inits.
     Note  that the  inits are  evaluated  before the  vars are  bound.   At the
     beginning of  each succeeding  iteration those vars  that have  repeats get
     setq'ed  to the  values of  their respective  repeats.  Note  that  all the
     repeats are evaluated before any of the vars is changed.

     The second element  of the do-form  is a list  of an end  testing predicate
     end-test and zero or more forms, the exit-forms.  At the beginning  of each
     iteration, after processing of the repeats, the end-test is  evaluated.  If
     the result  is nil, execution  proceeds with  the body of  the do.   If the
     result is not nil, the exit-forms are evaluated from left to right and then
     do returns.  The value of the do is the value of the last exit-form, or nil
     if there were no exit-forms.   Note that the second element of  the do-form
     resembles a cond clause.

     If the second element  of the form is nil,  there is no end-test  nor exit-
     forms, and the body of the do is executed only once.  In this type of do it
     is an  error to  have repeats.   This type of  do is  a "prog  with initial
     values."

     If the second element  of the form is  the S-expression (nil), there  is no
     end-test or exit-forms, and the body  of the do is executed over  and over.
     This is  a "do  forever." The  infinite loop  can be  terminated by  use of
     return or throw.

     The remainder of the do-form constitutes a prog-body.  When the end  of the
     body is reached, the next iteration  of the do begins.  If return  is used,
     do returns the indicated value and no more iterations occur.

Page 2-40                            ∪2-4.2                     December 5, 1977
**DRAFT**                       Flow of Control                        **DRAFT**


     The older variety of do is:

     (do var init repeat end-test body...)

     The first time through the loop  var gets the value of init;  the remaining
     times through the loop it  gets the value of repeat, which  is re-evaluated
     each time.  Note that init is  evaluated before the value of var  is saved.
     After var is set, end-test is evaluated.  If it is non-nil, the do finishes
     and returns nil.  If the end-test is nil, the body of the loop is executed.
     The body  is like a  prog body.  go  may be used.   If return is  used, its
     argument is the value of the do.   If the end of the prog body  is reached,
     another loop begins.

     Examples of the older variety of do:

        (setq n (cadr (arraydims x)))
        (do i 0 (1+ i) (= i n)
                (store (x i) 0))      ;zeroes out the array x

        (do zz x (cdr zz) (or (null zz) (zerop (f (car zz)))))
                           ; this applies f to each element of x
                           ; continuously until f returns zero.

     Examples of the new form of do:

             (do ((n (cadr (arraydims x)))
                  (i 0 (1+ i)))
                 ((= i n)
               (store (x i) 0))
                     ;this is like the example above,
                     ;except n is local to the do

             (do ((x) (y) (z)) (nil) body)
        is like
             (prog (x y z) body)

     except  that when  it runs  off the  end  of  the body,  do loops  but prog
     returns nil.  On the other hand,

             (do ((x) (y) (z)) nil body)

        is identical to the prog above (it does not loop.)


December 5, 1977                     ∪2-4.2                            Page 2-41
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



        (do ((x y (f x))) ((p x)) body)

        is like

        (do x y (f x) (p x) body)

     The construction

        (do ((x e (cdr x)) (oldx x x)) ((null x)) body)

     exploits parallel assignment to  index variables.  On the  first iteration,
     the value of oldx  is whatever value x had  before the do was  entered.  On
     succeeding iterations, oldx contains the  value that x had on  the previous
     iteration.

     In either form of do, the body may contain no forms at all.  Very  often an
     iterative algorithm can be  most clearly expressed entirely in  the repeats
     and exit-forms of a new-style do, and the body is empty.

        (do ((x x (cdr x))
             (y y (cdr y))
             (z nil (cons (f x y) z))) ;exploits parallel
            ((or (null x) (null y))    ; assignment.
             (nreverse z))             ;typical use of nreverse.
            )                          ;no do-body required.

        is like (maplist 'f x y).


go                  FSUBR

     The (go tag) special form is used  to do a "go-to" within the body of  a do
     or a prog.  If  the tag is an atom,  it is not evaluated.  Otherwise  it is
     evaluated and should yield an atom.  Then go transfers control to the point
     in the  body labelled by  a tag  eq or =  to the one  given.  (Tags  may be
     either atomic symbols or numbers).  If there is no such tag in the body, it
     is an unseen-go-tag error.

     "Computed" go's should be avoided in compiled code, or altogether.




Page 2-42                            ∪2-4.2                     December 5, 1977
**DRAFT**                       Flow of Control                        **DRAFT**



        Example:

        (prog (x y z)
          (setq x some frob)
        loop
          do something
          (and some predicate (go loop))  ;regular go
          do something more
          (go (cond ((minusp x) 'loop)    ;"computed go"
                    (t 'endtag)))
        endtag
          (return z))


return              SUBR 1 arg

     return  is used  to return  from a  prog or  a do.   The value  of return's
     argument  is returned  by prog  or  do as  its value.   In  addition, break
     recognizes  the typed-in  S-expression (return  value) specially.   If this
     form is typed at a break, value will be evaluated and returned as the value
     of break.  If  not at the  top level of  a form typed  at a break,  and not
     inside a prog or do, return will cause a fail-act error.
     Example:

        (do ((x x (cdr x))
             (n 0 (* n 2)))
            ((null x) n)
         (cond ((atom (car x))
                (setq n (1+ n)))
               ((memq (caar x) '(sys boom bleah))
                (return n))))












December 5, 1977                     ∪2-4.2                            Page 2-43
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


4.3  Non-local Exits


catch               FSUBR

     catch is the Maclisp function for doing structured non-local exits.  (catch
     x) evaluates x and returns its value, except that if during  the evaluation
     of x  (throw y) should  be evaluated, catch  immediately returns  y without
     further evaluating x.

     catch may also be used with a second argument, not evaluated, which is used
     as a tag to distinguish between  nested catches.  (catch x b) will  catch a
     (throw y b)  but not a  (throw y z).  throw  with only one  argument always
     throws to the  innermost catch.  catch with  only one argument  catches any
     throw.  It is an error if throw is done when there is no suitable catch.
     Example:

        (catch (mapcar (function (lambda (x)
                                   (cond ((minusp x)
                                          (throw x negative))
                                         (t (f x)) )))
                       y)
               negative)

     which returns  a list  of f  of each  element of  y if  y is  all positive,
     otherwise the first negative member of y.

     The user of catch and throw is advised to stick to the 2 argument versions,
     which are  no less efficient,  and tend to  reduce the likelihood  of bugs.
     The one argument versions  exist primarily as an  easy way to fix  old Lisp
     programs  which  use  errset  and err  for  non-local  exits.   This latter
     practice is  rather confusing, because  err and errset  are supposed  to be
     used for error handling, not general program control.

     The catch-tag break is used by the break function.









Page 2-44                            ∪2-4.3                     December 5, 1977
**DRAFT**                       Flow of Control                        **DRAFT**


throw               FSUBR

     throw is used with catch as a structured non-local exit mechanism.

     (throw x) evaluates x and throws the value back to the most recent catch.

     (throw x tag) throws the value of x back to the most recent  catch labelled
     with tag or unlabelled.  catch'es with tags not eq to tag are skipped over.
     x is evaluated but tag is not.

     See the description of catch for further details.

































December 5, 1977                     ∪2-4.3                            Page 2-45
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


4.4  Causing and Controlling Errors


   See the complete description of the Maclisp error system (part 3.4)  for more
information about how these functions work.

error               LSUBR 0 to 3 args

     This is a function which  allows user functions to signal their  own errors
     using the Maclisp error system.

     (error) is the same as (err).

     (error message)  signals a simple  error; no datum  is printed and  no user
     interrupt is signalled.  The error message typed out is message.

     (error message datum)  signals an error with  message as the message  to be
     typed out and datum as the Lisp object to be printed in the  error message.
     No user interrupt is signalled.

     (error message datum  uint-chn) signals an error  but first signals  a user
     interrupt on channel uint-chn, provided  that there is such a  channel, and
     it has a  non-nil service function,  and the special  conditions concerning
     errset (see page  3-16) are satisfied.  uint-chn  is the name of  the user-
     interrupt channel to  be used (an atomic  symbol); see part 3.4.2.   If the
     service function returns  an atom, error goes  ahead and signals  a regular
     error.  If the service function returns a list, error returns as  its value
     the car of that list.  In this case it was a "correctable" error.   This is
     the only case  in which error  will return; in  all other cases  control is
     thrown back to top level, or to the nearest enclosing errset.


errset              FSUBR

     The special  form (errset  form flag) is  used to  trap an  expected error.
     errset evaluates the form.  If an error occurs during the evaluation of the
     form,  the error  is prevented  from escaping  from inside  the  errset and
     errset returns nil.  If no errors occur, a list of one element,  the result
     of the evaluation, is returned.  The result is listified so that there will
     no ambiguity if it is nil.  errset may also be made to return any arbitrary
     value by use of the err function.



Page 2-46                            ∪2-4.4                     December 5, 1977
**DRAFT**                       Flow of Control                        **DRAFT**


     The flag is optional.  If present, it is evaluated before the form.   If it
     is nil,  no error message  will be  printed if an  error occurs  during the
     evaluation of the form.  If it is  not nil, or if it is omitted,  any error
     messages generated will be printed.

     Examples:

     If you are not sure x is a number:

                           (errset (setq x (add1 x)))

     This example may not work in compiled code if the compiler chooses to open-
     code the  add1 rather than  calling the add1  subroutine.  In  general, one
     must be extremely foolhardy to depend on error checking in compiled code.

     To suppress the error message if the value of a is not an atomic symbol:

                             (errset (set a b) nil)

     To do the same but generate one's own message:

              (or (errset (set a b) nil)
                  (error '(not a variable) a))


err                 FSUBR

     (err) causes an error which is handled the same as a Lisp error except that
     there is no preliminary user interrupt, and no message is typed out.

     (err x)  is like (err)  except that  if control returns  to an  errset, the
     value of the errset will be the result of evaluating x, instead of nil.

     (err x nil) is the same as (err x).

     (err x t) is like (err x) except that x is not evaluated until  just before
     the errset returns it.  That is, x is evaluated after unwinding the pdl and
     restoring the bindings.

     Note: some people use err  and errset where catch and throw  are indicated.
     This is a very poor programming practice.  See writeups of catch  and throw
     for details.


December 5, 1977                     ∪2-4.4                            Page 2-47
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**














































Page 2-48                            ∪2-4.4                     December 5, 1977
**DRAFT**                        Atomic Symbols                        **DRAFT**




5.  Atomic Symbols


5.1  The Value Cell


   Each atomic symbol has associated with  it a value cell, which is a  piece of
storage that can refer to one  Lisp object.  This object is called  the symbol's
value, since it is what is returned if the symbol is evaluated.  The  binding of
atomic  symbols  to  values  allows  them to  be  used  in  programming  the way
"variables" are used in other languages.

   The value cell can also be empty,  in which case the symbol has no  value and
is said  to be  unbound or  undefined.  This is  the initial  state of  a newly-
created atomic symbol.  Attempting to evaluate an unbound symbol causes an error
to be signalled.

   An object can be  placed into a symbol's  value cell by lambda-binding  or by
assignment.   (See page  1-11.)  The  difference is  in how  closely  the value-
changing is associated with control structure and in whether it is  considered a
side-effect.


setq                FSUBR

     The  setq  special form  is  used  to assign  values  to  variables (atomic
     symbols.) setq processes  the elements of  its form in  pairs, sequentially
     from left  to right.   The first  member of  each pair  is a  variable, the
     second is a form  which evaluates to a  value.  The form is  evaluated, but
     the variable is not.  The value-binding  of the variable is made to  be the
     value specified.  You must  not setq the special atomic-symbol  constants t
     and nil.  The value returned by  setq is the last value assigned,  i.e. the
     result of the evaluation of the last element of the setq-form.

     Example:  (setq x (+ 1 2 3) y (cons x nil))

     This returns (6) and gives x a value of 6 and y a value of (6).

     Note that the first assignment is completed before the second assignment is
     started, resulting in the second use of x getting the value assigned in the
     first pair of the setq.

December 5, 1977                     ∪2-5.                             Page 2-49
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


set                 SUBR 2 args

     set is like setq except that the first argument is evaluated; also set only
     takes one pair of arguments.  The first argument must evaluate to an atomic
     symbol, whose value  is changed to the  value of the second  argument.  set
     returns the value of its second argument.  Example:

            (set (cond ((predicate) 'atom1) (t 'atom2))
                        'stba)

     evaluates to stba and gives either atom1 or atom2 a value of stba.

        set could have been defined by:

        (defun set (x y)
            (eval (list 'setq x (list 'quote y))))

        Alternatively, setq could have been defined by:

        (defun setq fexpr (x)
          ((lambda (var val rest)
             (set var val)
             (cond ((null rest) val)
                   ((apply (function setq) rest)) ))   ;if more, recurse
           (car x)
           (eval (cadr x))
           (cddr x)))


symeval             SUBR 1 arg

     (symeval a) returns the  value of a, which  must be an atomic  symbol.  The
     compiler produces highly  optimal code for  symeval, making it  much better
     than eval when the value of  a symbol needs to be taken and  the particular
     symbol to be used varies.









Page 2-50                            ∪2-5.1                     December 5, 1977
**DRAFT**                        Atomic Symbols                        **DRAFT**


boundp              SUBR 1 arg

     The argument to boundp must be an  atomic symbol.  If it has a value,  t is
     returned.  Otherwise nil is returned.


makunbound          SUBR 1 arg

     The argument to makunbound must be an atomic symbol.  Its value is removed,
     i.e. it becomes unbound.

        Example:
                (setq a 1)
                a => 1
                (makunbound 'a)
                a => unbnd-vrbl error.

     makunbound returns its argument.


























December 5, 1977                     ∪2-5.1                            Page 2-51
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


5.2  The Property List


   A property-list  is a list  with an  even number of  elements.  Each  pair of
elements constitutes a property; the first element is called the "indicator" and
the second is called the  "value" or, loosely, the "property." The  indicator is
generally an atomic symbol which serves as the name of the property.   The value
is any Lisp object.

   For  example, one  type of  functional  property uses  the atom  expr  as its
indicator.  In the case of an expr-property, the value is a list  beginning with
lambda.

   An example of a property list with two properties on it is:

                    (expr (lambda (x) (plus 14 x)) foobar t)

   The first property has indicator expr and value (lambda (x) (plus 14 x)), the
second property has indicator foobar and value t.

   Each  atomic symbol  has associated  with it  a property-list,  which  can be
retrieved with the  plist function.  It is  also possible to  have "disembodied"
property  lists  which are  not  associated  with any  symbol.   These  keep the
property  list on  their cdr,  so the  form of  a disembodied  property  list is
(<anything> . plist).  The way  to create a disembodied property list  is (ncons
nil).  Atomic symbols also (usually) keep their property list on their  cdr, but
you aren't allowed  to know that.   Use the plist  function to get  the property
list of a symbol.

   Property lists are useful for associating "attributes" with symbols.  Maclisp
uses properties to remember function definitions.  The compiler  uses properties
internally  to keep  track of  some of  what it  knows about  the program  it is
compiling.

   The user  familiar with Lisp  1.5 will  want to note  that the  property list
"flags" which are allowed  on Lisp 1.5 property  lists do not exist  in Maclisp.
However, the same effect can be  achieved by using properties with a value  of t
or nil.

   Some property names are used  internally by Maclisp, and should  therefore be
avoided in user code.  These include args, array, autoload, expr,  fexpr, fsubr,
lsubr, macro, pname, sublis, subr, value, used by the Lisp system proper; arith,


Page 2-52                            ∪2-5.2                     December 5, 1977
**DRAFT**                        Atomic Symbols                        **DRAFT**


*array,  atomindex,  *expr,  *fexpr,  *lexpr,  numfun,  number,  numvar,  ohome,
special, sym, used by the compiler; grindfn, grindmacro, used by the grinder.


get                 SUBR 2 args

     (get x y) gets x's y-property.  x can be an atomic symbol or  a disembodied
     property list.  The value of x's y-property is returned, unless x has no y-
     property in which case nil is returned.   It is not an error for x to  be a
     number, but nil will always be returned since numbers do not  have property
     lists.
     Example:

        (get 'foo 'bar)
          => nil                 ;initially foo has no bar property
        (putprop 'foo 'zoo 'bar) ;give foo a bar property
          => zoo
        (get 'foo 'bar)          ;retrieve that property
          => zoo
        (plist 'foo)             ;look at foo's property list
        => (bar zoo ...other properties...)

        get could have been defined by:

        (defun get (x y)
          (do ((z (cond ((numberp x) nil)
                        ((atom x) (plist x))
                        (t (cdr x)))
                  (cddr z)))
              ((or (null z) (eq y (car z)))
               (cadr z))))

     This relies  on the  fact that  the car  and the  cdr of  nil are  nil, and
     therefore (cadr z) is nil if z is nil.


getl                SUBR 2 args

     (getl x y) is  like get except that y  is a list of indicators  rather than
     just a single indicator.  getl searches x's property list until  a property
     whose indicator appears in the list y is found.



December 5, 1977                     ∪2-5.2                            Page 2-53
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     The portion of x's property list beginning with the first such  property is
     returned.  The car of this is the indicator (property name) and the cadr is
     the property value.  getl returns nil if none of the indicators in y appear
     on the property list of x.  getl could have been defined by:

        (defun getl (x pl)
           (do ((q (plist x) (cddr q))) ; scan down P-list of x
               ((or (null q) (memq (car q) pl))
                q)))

     This  definition is  simplified and  doesn't take  numbers  and disembodied
     property lists into account.


putprop             SUBR 3 args

     (putprop x y  z) gives x  a z-property of  y and returns  y.   x may  be an
     atomic symbol or a disembodied property list.  After somebody does (putprop
     x y z), (get x z) will return y.

     Example:
                          (putprop 'Nixon 'not 'crook)

     If the symbol already  has a property with  the same name that  property is
     removed first.  This ensures that getl will always find the  property which
     was put on most recently.  For instance, if you were to redefine an expr as
     a subr,  and then  redefine it  as an  expr again,  this effect  of putprop
     causes the evaluator to find the latest definition always.

     A lisp definition of the basic putprop without the complications of numbers
     and disembodied property lists might be:

        (defun putprop (x y z)
          (remprop x z)
          (setplist x (cons z (cons y (plist x))))
          y)








Page 2-54                            ∪2-5.2                     December 5, 1977
**DRAFT**                        Atomic Symbols                        **DRAFT**


defprop             FSUBR

     defprop  is a  version  of putprop  with no  argument-evaluation,  which is
     sometimes more convenient for typing.  For instance,

                          (defprop foo bar oftenwith)

     is equivalent to

                         (putprop 'foo 'bar 'oftenwith)


remprop             SUBR 2 args

     (remprop x y)  removes x's y-property, by  splicing it out of  x's property
     list.   The value  is nil  if x  had no  y-property.  If  x did  have  a y-
     property, the value is a list  whose car is the property, and whose  cdr is
     part of x's property list, similar to (cdr (getl x '(y))).

     x may be an atomic symbol or a disembodied property list.  Example:

                              (remprop 'foo 'expr)

     undefines the function foo, assuming it was defined by

                              (defun foo (x) ... )


plist               SUBR 1 arg

     (plist x) returns the property list of the atomic symbol x.


setplist            SUBR 2 args

     (setplist x l) sets the property list of the atomic symbol x to l.  This is
     to  be used  with  caution, since  in some  implementations  property lists
     contain internal system properties  which are essential to the  workings of
     the Lisp system.





December 5, 1977                     ∪2-5.2                            Page 2-55
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


5.3  The Print-Name


   Each  atomic symbol  has an  associated character  string called  its "print-
name," or  "pname" for  short.  This character  string is  used as  the external
representation  of the  symbol.  If  the string  is typed  in, it  is read  as a
reference to the symbol.  If the  symbol is to be print'ed, the string  is typed
out.

   See also page 2-83 for some other functions which have to do with pnames.


samepnamep          SUBR 2 args

     The arguments to samepnamep must evaluate to atomic symbols or to character
     strings.  The result is t if they have the same pname, nil  otherwise.  The
     pname  of  a  character  string is  considered  to  be  the  string itself.
     Examples:

               (samepnamep 'xyz (maknam '(x y z))) => t

               (samepnamep 'xyz (maknam '(w x y))) => nil

               (samepnamep 'x "x") => t


alphalessp          SUBR 2 args

     (alphalessp x  y), where x  and y evaluate  to atomic symbols  or character
     strings, returns t if the  pname of x occurs earlier in  alphabetical order
     than the pname of y.  The  pname of a character string is considered  to be
     the string itself.  Examples:

               (alphalessp 'x 'x1) => t

               (alphalessp 'z 'q) => nil

               (alphalessp "x" 'y) => t

     Note that the "alphabetical order" used by alphalessp is actually the ASCII
     collating sequence.   Consequently all upper  case letters come  before all
     lower case letters.


Page 2-56                            ∪2-5.3                     December 5, 1977
**DRAFT**                        Atomic Symbols                        **DRAFT**


pnget               SUBR 2 args

     (pnget symbol  n) returns  the pname  of the  symbol as  a list  of fixnums
     containing  packed  n-bit bytes.   The  legal  values of  n  depend  on the
     implementation; in the pdp-10 implementation, 6 (SIXBIT) and 7  (ASCII) are
     allowed.  If this seems obscure, that's because it is.  Example:

        (pnget 'MUMBLERATOR 7) =>
                    (-311246236550 -351327625542 -270←33)


pnput               SUBR 2 args

     This is a sort of inverse  of pnget.  (pnput (pnget foo 7) flag)  returns a
     symbol with the same pname as foo.  The symbol is interned if flag  is non-
     nil.




























December 5, 1977                     ∪2-5.3                            Page 2-57
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


5.4  Interning of Symbols


   One normally wants  to refer to  the same (eq)  atomic symbol every  time the
same  pname  is typed.   Maclisp  implements  this through  what  is  called the
obarray.  The obarray is a hash-table of atomic symbols.  These symbols are said
to be interned, or registered in the obarray.  Whenever a pname is read  in Lisp
input, the  obarray is searched  for a symbol  with the same  pname.  If  one is
found, the pname is considered to refer to that symbol.  If not, a new symbol is
created and added to the obarray.

   The  representation  of an  obarray  is a  Lisp  array.  The  first  510. (or
thereabouts) elements  of the array  contain lists which  are buckets of  a hash
table.  The  last 128. elements  of the array  contain the  "character objects,"
symbols  with   1-character  pnames.    (These  entries   contain  nil   if  the
corresponding  symbol has  not  yet been  interned.) The  character  objects are
treated specially  for efficiency.  There  are usually one  or two  unused array
elements between these two areas.

   In order to allow for multiple name spaces, Maclisp allows multiple obarrays.
An  obarray  can  be  made  "current"  by  binding  the  symbol  obarray  to the
appropriate  array-pointer.  See  page  2-89 for  details on  how  to manipulate
obarrays and arrays in general.

   It is  possible to have  a symbol  interned on several  obarrays at  the same
time.  It is also possible to have two different (non-eq) symbols with  the same
pname interned  on different  obarrays.  Furthermore  it is  possible to  have a
symbol  which is  not interned  on any  obarray, which  is called  an uninterned
symbol.   These  are  useful  for  purely-internal  functions,  but   can  cause
difficulty in debugging  since they can't be  accessed directly.  Such  a symbol
can be accessed via some data structure that contains it, set up by  the program
that created it.

   Normally symbols  are never removed  from obarrays.  It  is possible  for the
user to explicitly remove  a symbol from the  current obarray.  There is  also a
feature by  which "truly  worthless" symbols may  be removed  automatically (see
page 3-58).







Page 2-58                            ∪2-5.4                     December 5, 1977
**DRAFT**                        Atomic Symbols                        **DRAFT**


intern              SUBR 1 arg

     (intern x), where x is  an atomic symbol, returns the unique  atomic symbol
     which is  "interned on the  obarray" and has  the same pname  as x.   If no
     symbol on the  current obarray has  the same pname  as x, then  intern will
     place x itself on the obarray, and return it as the value.


remob               SUBR 1 arg

     The argument to  remob must be  an atomic symbol.   It is removed  from the
     current obarray if it is  interned on that obarray.  This makes  the atomic
     symbol inaccessible to any S-expressions  that may be read in or  loaded in
     the future.  remob returns nil.


copysymbol          SUBR 2 args

     A subr  of two arguments.   The first  argument must be  a symbol,  and the
     second should be t  or nil.  The result  is a new, uninterned  symbol, with
     the same pname as the argument.  "Uninterned" means that the symbol has not
     been placed on any  obarray.  If the second  argument is t, the  new symbol
     will be given the  same value as the original  and will have a copy  of its
     property  list.   Thus the  new  will start  out  with the  same  value and
     properties as the  old, but if  it is setq'ed  or putprop'ed, the  value or
     property of the old  will not be changed.   If the second argument  is nil,
     the new  symbol has no  value and no  properties (except  possibly internal
     system properties.)


gensym              LSUBR 0 or 1 args

     gensym creates and returns a new atomic symbol, which is not interned on an
     obarray  (and therefore  is not  recognized by  read.) The  atomic symbol's
     pname is of the form prefix number, e.g. g0001.  The number  is incremented
     each time.

     If gensym  is given  an argument,  a numeric  argument is  used to  set the
     number.  The pname of an atomic-symbol argument is used to set  the prefix.
     For example:




December 5, 1977                     ∪2-5.4                            Page 2-59
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



          if   (gensym) => g0007
          then (gensym 'foo) => f0008
               (gensym 40) => f0032
          and  (gensym) => f0033

     Note that the number is in  decimal and always four digits, and  the prefix
     is always one character.




































Page 2-60                            ∪2-5.4                     December 5, 1977
**DRAFT**                        Atomic Symbols                        **DRAFT**


5.5  Defining Atomic Symbols as Functions


   Atomic symbols may be used as  names for functions.  This is done  by putting
the actual function (a subr-object or a lambda-expression) on the  property list
of  the  atomic  symbol  as  a "functional  property,"  i.e.  under  one  of the
indicators expr, fexpr, macro, subr, lsubr, or fsubr.

   Array  properties  (see  page  2-89) are  also  considered  to  be functional
properties, so an atomic symbol which is  the name of an array is also  the name
of a function, the accessing function of that array.

   When  an atomic  symbol  which is  the name  of  a function  is  applied, the
function which it names is substituted.


defun               FSUBR

     defun is used for defining functions.  The general form is:

                (defun name  type (lambda-variable...)
                       body...)

     However, name and type may  be interchanged.  type, which is  optional, may
     be expr, fexpr, or macro.  If it is omitted, expr is assumed.  Examples:

          (defun addone (x) (1+ x))             ;defines an expr

          (defun quot fexpr (x) (car x))        ;defines a fexpr

          (defun fexpr quot (x) (car x))        ;is the same

          (defun zzz expr x                     ;this is how you
                     (foo (arg 1)(arg 2)))      ; define a lexpr.

     The first example above is really just defining a synonym.  Another  way to
     do this is:
                            (defprop addone 1+ expr)

     That is, an atomic  functional property indicates synonyming.  This  can be
     particularly useful to  define a macro  by an expr or  fexpr, or even  by a
     subr.


December 5, 1977                     ∪2-5.5                            Page 2-61
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     The functions defprop and putprop may also be used for defining functions.

     There is a feature by which, when a file of functions has been compiled and
     loaded into  the lisp  environment, the file  may be  edited and  then only
     those  functions  which  were  changed  may  be  loaded   for  interpretive
     execution.  This is done by compiling with the "E" switch, and then reading
     in the source  file with the variable  defun bound non-nil.   Each function
     will have an  expr-hash property maintained,  which contains the  sxhash of
     the interpreted definition of  the function.  defun will only  redefine the
     function if this hash-code  has changed.  This feature is  rather dangerous
     since reasonable alterations to the function definition may not  change the
     sxhash  and  consequently may  not  take effect.   Because  of  its general
     losingness, this feature is only available in the pdp-10 implementation and
     sometimes not even there.

        defun could have been defined by:

        (defun defun fexpr (x)  ;circular, but you get the idea
          (prog (name type body)

            ; first, analyze the form, get arguments.
            (cond ((memq (car x) '(expr fexpr macro))
                   (setq type (car x)
                         name (cadr x)
                         body (cddr x)))
                  ((memq (cadr x) '(expr fexpr macro))
                   (setq name (car x)
                         type (cadr x)
                         body (cddr x)))
                  ((setq name (car x)
                         type 'expr
                         body (cdr x))))

            (setq body (cons 'lambda body))










Page 2-62                            ∪2-5.5                     December 5, 1977
**DRAFT**                        Atomic Symbols                        **DRAFT**



            ; now, check for expr-hash hair.
            (cond ((and defun
                        (get name 'expr-hash)
                        (= (get name 'expr-hash)
                           (sxhash body)))
                     )
                  ; actually make the definition.
                  ((putprop name body type)))
            (return name)))


args                LSUBR 1 or 2 args

     (args f) tells you the number of arguments expected by the function  f.  If
     f wants n  arguments, args returns (nil  . n).  If f  can take from m  to n
     arguments, args returns  (m . n).  If  f is an fsubr  or a lexpr,  expr, or
     fexpr, the results are meaningless.

     (args f x), where x is (nil .  n) or (m . n), sets the number  of arguments
     desired  by  the function  f.   This only  works  for  compiled, non-system
     functions.


sysp                SUBR 1 arg

     The sysp predicate  takes an atomic symbol  as an argument.  If  the atomic
     symbol is the name of a system function (and has not been  redefined), sysp
     returns  the type  of  function (subr,  lsubr, or  fsubr).   Otherwise sysp
     returns nil.  Examples:

            (sysp 'foo) => nil  (presumably)

            (sysp 'car) => subr

            (sysp 'cond) => fsubr








December 5, 1977                     ∪2-5.5                            Page 2-63
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**














































Page 2-64                            ∪2-5.5                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**




6.  Numbers


For a description of the various types of numbers used in Maclisp, see part 1.2.

6.1  Number Predicates


zerop               SUBR 1 arg

     The zerop  predicate returns  t if its  argument is  fixnum zero  or flonum
     zero.  (There  is no  bignum zero.)  Otherwise  it returns  nil.  It  is an
     error if the argument is not a number.  If that is possible signp should be
     used.


plusp               SUBR 1 arg

     The plusp  predicate returns  t if  its argument  is strictly  greater than
     zero, nil if it is zero or negative.  It is an error if the argument is not
     a number.


minusp              SUBR 1 arg

     The minusp predicate returns t if its argument is a negative number, nil if
     it is  a non-negative  number.  It  is an error  if the  argument is  not a
     number.


oddp                SUBR 1 arg

     The oddp predicate  returns t if its  argument is an odd  number, otherwise
     nil.  The argument must be a fixnum or a bignum.








December 5, 1977                     ∪2-6.                             Page 2-65
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


signp               FSUBR

     The signp  predicate is used  to test the  sign of a  number.  (signp  c x)
     returns t  if x's  sign satisfies the  test c,  nil if it  does not.   x is
     evaluated but c is not.  The result is always nil if x is not a  number.  c
     can be one of the following:

                  l       means    x<0
                  le        "      x<λ←0
                  e         "      x=0
                  n         "      x=λ/0
                  ge        "      x>λ←0
                  g         "      x>0

     Examples:
         (signp le -1) => t
         (signp n 0) => nil
         (signp g '(foo . bar)) => nil


haulong             SUBR 1 arg

     (haulong x) returns the number of significant bits in x.  x can be a fixnum
     or a  bignum.  The result  is the  least integer not  less than  the base-2
     logarithm of |x|+1.

        Examples:
            (haulong 0) => 0
            (haulong 3) => 2
            (haulong -7) => 3
            (haulong 12345671234567) => 40.













Page 2-66                            ∪2-6.1                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**


6.2  Comparison


=                   SUBR 2 args

     (= x  y) is  t if x  and y  are numerically equal.   x and  y must  be both
     fixnums or both flonums.  Use equal to compare bignums.


greaterp            LSUBR 2 or more args

     greaterp compares its arguments, which must be numbers, from left to right.
     If any argument is not greater than the next, greaterp returns nil.  But if
     the  arguments  to  greaterp  are strictly  decreasing,  the  result  is t.
     Examples:

                (dπeaterp 4 3) => t
                (greaterp 1 1) => nil
                (greaterp 4.0 3.6 -2) => t
                (greaterp 4 3 1 2 0) => nil


>                   SUBR 2 args

     (> x y) is t if x is  strictly greater than y, and nil otherwise.  x  and y
     must be both fixnums or both flonums.


lessp               LSUBR 2 or more args

     lessp compares its  arguments, which must be  numbers, from left  to right.
     If any argument is not less  than the next, lessp returns nil.  But  if the
     arguments to lessp are strictly increasing, the result is t.  Examples:

                (lessp 3 4) => t
                (lessp 1 1) => nil
                (lessp -2 3.6 4) => t
                (lessp 0 2 1 3 4) => nil






December 5, 1977                     ∪2-6.2                            Page 2-67
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


<                   SUBR 2 args

     (< x y) is t if x is strictly less than y, and nil otherwise.  x and y must
     be both fixnums or both flonums.


max                 LSUBR 1 or more args

     max returns the  largest of its arguments,  which must be numbers.   If any
     argument is a flonum, the result will be a flonum.  Otherwise, it will be a
     fixnum or a bignum depending on its magnitude.


min                 LSUBR 1 or more args

     min returns the smallest of  its arguments, which must be numbers.   If any
     argument is a flonum, the result will be a flonum.  Otherwise, it will be a
     fixnum or a bignum depending on its magnitude.


























Page 2-68                            ∪2-6.2                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**


6.3  Conversion


fix                 SUBR 1 arg

     (fix x)  converts x to  a fixnum  or a bignum  depending on  its magnitude.
     Examples:

          (fix 7.3) => 7
          (fix -1.2) => -2
          (fix 104) => 104


ifix                SUBR 1 arg

     (ifix x) converts x  from a flonum to a  fixnum.  ifix will never  return a
     bignum, unlike fix.  This allows it to be efficiently open-coded.   This is
     not the  same function  as IFIX in  Fortran; rounding  is down  rather than
     towards zero.  It is like ENTIER in Algol.

     ifix does not exist in the Multics implementation.


float               SUBR 1 arg

     (float x) converts x to a flonum.  Example:

          (float 4) => 4.0
          (float 3.27) => 3.27


abs                 SUBR 1 arg

     (abs x) => |x|,  the absolute value of the  number x.  abs could  have been
     defined by:

        (defun abs (x) (cond ((minusp x) (minus x))
                             (t x) ))






December 5, 1977                     ∪2-6.3                            Page 2-69
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


minus               SUBR 1 arg

     minus  returns the  negative of  its  argument, which  can be  any  kind of
     number.  Examples:

                (minus 1) => -1
                (minus -3.6) => 3.6


haipart             SUBR 2 args

     (haipart  x  n) extracts  n  leading  or trailing  bits  from  the internal
     representation of x.  x may be a  fixnum or a bignum.  n must be  a fixnum.
     The value  is returned  as a  fixnum or a  bignum.  If  n is  positive, the
     result  contains  the  n  high-order significant  bits  of  |x|.   If  n is
     negative, the  result contains the  |n| low-order bits  of |x|.  If  |n| is
     bigger than the number of significant bits in x, |x| is returned.

        Examples:
            (haipart 34567 7) => 162

            (haipart 34567 -5) => 27

            (haipart -34567 -5) => 27




















Page 2-70                            ∪2-6.3                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**


6.4  Arithmetic


                               General Arithmetic

   These functions will  perform arithmetic on any  kind of numbers,  and always
yield an  exact result, except  when used with  flonums.  (Flonums  have limited
precision and range.) Conversions to flonum or bignum representation are done as
needed.  Flonum representation will be used if any of the arguments are flonums;
otherwise fixnum  representation will be  used if the  result can fit  in fixnum
form, or bignum representation if it cannot.

   The two sections after this one describe other arithmetic functions which are
more efficient but less powerful.

plus                LSUBR 0 or more args

     plus returns the sum of its arguments, which may be any kind of numbers.


difference          LSUBR 1 or more args

     difference returns its first argument minus the rest of its  arguments.  It
     works for any kind of numbers.


times               LSUBR 0 or more args

     times  returns the  product of  its arguments.   It works  for any  kind of
     numbers.


quotient            LSUBR 1 or more args

     quotient returns its first argument  divided by the rest of  its arguments.
     The arguments may any kind of number.

        Examples:
          (quotient 3 2)  => 1      ;fixnum division truncates.

          (quotient 3 2.0) =>  1.5  ;but flonum division does not.λ.

          (quotient 6.0 1.5  2.0) => 2.0

December 5, 1977                     ∪2-6.4                            Page 2-71
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


add1                SUBR 1 arg

     (add1 x) => x+1.  x may be any kind of number.


sub1                SUBR 1 arg

     (sub1 x) => x-1.  x may be any kind of number.


remainder           SUBR 2 args

     (remainder x y) =>  the remainder of the division  of x by y.  The  sign of
     the remainder is the same as the sign of the dividend.  The  arguments must
     be fixnums or bignums.


gcd                 SUBR 2 args

     (gcd x y) => the greatest common divisor of x and y.  The arguments must be
     fixnums or bignums.


expt                SUBR 2 args

                   z
     (expt x z) = x

     The exponent z may be  a bignum if the base x  is 0, 1, or -1;  otherwise z
     should be a fixnum.  x may be any kind of number.

     As a special feature,  expt allows its second  argument to be a  flonum, in
     which  case  the  first  argument   is  converted  to  a  flonum   and  the
     exponentiation  is  performed  in floating  point,  using  logarithms.  The
     result is a flonum in this case.


*dif                SUBR 2 args

     *dif is  a subr  form of difference.   It is  documented here  because some
     people  use  it.   There  is  no  reason  to  use  it,  since  the compiler
     automatically converts difference into *dif as required.


Page 2-72                            ∪2-6.4                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**


*quo                SUBR 2 args

     *quo is a subr form of quotient.  It is documented here because some people
     use it.   There is no  reason to use  it, since the  compiler automatically
     converts quotient into *quo as required.







































December 5, 1977                     ∪2-6.4                            Page 2-73
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


                               Fixnum Arithmetic


   These functions require their arguments to be fixnums and produce only fixnum
results.   If the  true result,  which  would be  returned by  the  more general
functions described previously, is too large to be represented as a  fixnum, the
result actually returned will be truncated to an implementation-dependent number
of  bits,  which  is  36.  (including  the  sign)  in  the  Multics  and  pdp-10
implementations.   The  compiler   produces  highly-optimized  code   for  these
operations.

+                   LSUBR 0 or more args

     + returns the sum of its arguments.  The arguments must be fixnums, and the
     result is always a fixnum.  Examples:

                (+ 2 6 -1) => 7
                (+ 3) => 3      ;trivial case
                (+) => 0        ;identity element


-                   LSUBR 0 or more args

     This  is  the  fixnum-only subtraction  function.   With  one  argument, it
     returns the number's negation.  With more than one argument, it returns the
     first argument minus the rest of the arguments.

                (-) => 0, the identity element
                (- 3) => -3
                (- 5 3) => 2
                (- 2 6 -1) => -3
        etc.


*                   LSUBR 0 or more args

     * returns the product of its arguments.  Examples:

                (* 4 5 -6) => -120.
                (* 3) => 3           ;trivial case
                (*) => 1             ;identity element



Page 2-74                            ∪2-6.4                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**


/                   LSUBR 0 or more args

     This is the fixnum-only  division function.  The arguments must  be fixnums
     and the result of the division is truncated to an integer and returned as a
     fixnum.  Note that the name of this function must be typed in as  //, since
     Lisp uses / as an escape character.

     If used with more than one  argument, it divides the first argument  by the
     rest of  the arguments.   If used with  only one  argument, it  returns the
     fixnum reciprocal of that number, which is -1, 0, 1, or undefined depending
     on whether the number is -1, large, 1, or 0.

               (//) => 1, the identity element.
               (// 20. 5) => 4
               (// 100. 3 2) => 16.
               etc.


1+                  SUBR 1 arg

     (1+ x) => x+1.  x must be a fixnum.  The result is always a fixnum.


1-                  SUBR 1 arg

     (1- x) => x-1.  x must be a fixnum.  The result is always a fixnum.


\                   SUBR 2 args

     (\ x y) returns the remainder of x divided by y, with the sign of x.  x and
     y must be fixnums.  Examples:

               (\ 5 2) => 1
               (\ 65. -9.) => 2
               (\ -65. 9.) => -2


\\                  SUBR 2 args

     This subr  of two arguments  is like gcd,  but only accepts  fixnums.  This
     makes it faster than gcd.


December 5, 1977                     ∪2-6.4                            Page 2-75
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


↑                   SUBR 2 args

     ↑ is the fixnum only  exponentiation function.  It is somewhat  faster than
     expt, but requires its arguments to be fixnums, uses fixnum arithmetic, and
     always returns a fixnum result, which will be incorrect if the  true result
     is too large to be represented as a fixnum.






































Page 2-76                            ∪2-6.4                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**


                               Flonum Arithmetic


   These functions  require their  arguments to be  flonums, and  always produce
flonum results.  If the true result is too large or too small to  be represented
as a flonum, an arithmetic underflow or overflow error will occur.  (In the pdp-
10  implementation these  errors  are not  detected in  compiled  programs.) The
compiler produces highly-optimized code for these operations.

+$                  LSUBR 0 or more args

     +$ returns the sum of its arguments.

        Examples:
                (+$ 4.1 3.14) => 7.24
                (+$  2.0 1.5  -3.6)  => -0.1
                (+$ 2.6) => 2.6         ;trivial case
                (+$)  => 0.0            ;identity element


-$                  LSUBR 0 or more args

     This  is the  flonum-only subtraction  function.  When  used with  only one
     argument,  it returns  the number's  negation.  Otherwise,  it  returns the
     first argument minus the rest of the arguments.

               (-$) => 0.0, the identity element
               (-$ x)  => the negation of x.
               (-$ 6.0 2.5) => 4.5
               (-$ 2.0 1.5 -3.6) => 3.1
               etc.


*$                  LSUBR 0 or more args

     *$ returns the product of its arguments.  Examples:

             (*$ 3.0 2.0 4.0) => 24.0
             (*$ 6.1) => 6.1            ;trivial case
             (*$) => 1.0                ;identity element




December 5, 1977                     ∪2-6.4                            Page 2-77
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


/$                  LSUBR 0 or more args

     This is  the flonum-only division  function.   Note that  the name  of this
     function must be typed in as //$, since Lisp uses / as an escape character.
     This function computes the reciprocal if given only one argument.  If given
     more than one argument, it divides the first by the rest.

               (//$) => 1.0, the identity element
               (//$ 5.0) => 0.2
               (//$ 6.28 3.14) => 2.0
               (//$ 100.0 3.0 2.0) => 16.5
               etc.


1+$                 SUBR 1 arg

     (1+$ x) => x+1.0.  x must be a flonum.  The result is always a flonum.


1-$                 SUBR 1 arg

     (1-$ x) => x-1.0.  x must be a flonum.  The result is always a flonum.


↑$                  SUBR 2 args

     ↑$ is the flonum-only exponentiation function.  The first argument  must be
     a flonum, the second must be a fixnum (repeat, a fixnum), and the result is
     a flonum.  To raise a flonum to a flonum power, use (expt x y) or  (exp (*$
     y (log x))).














Page 2-78                            ∪2-6.4                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**


6.5  Exponentiation and Logarithm Functions


sqrt                SUBR 1 arg

     (sqrt x) =>  a flonum which is  the square root of  the number x.   This is
     more  accurate than  (expt x  0.5).  The  following code,  which is  due to
     Gosper, should be written if the square root of a bignum is desired.  It is
     essentially a  Newton iteration, with  appropriate precautions  for integer
     truncation.

          (defun bsqrt (n)
             (bsqrt1 (abs n)
                     (expt 2 (// (1+ (haulong n)) 2))))

          (defun bsqrt1 (n guess)
                 ((lambda (next)
                    (cond ((lessp next guess)
                           (bsqrt1 n next))
                          (t guess)))
                  (quotient (plus guess (quotient n guess))
                            2)))


exp                 SUBR 1 arg

                 x
     (exp x) => e


log                 SUBR 1 arg

     (log x) => the natural logarithm of x.











December 5, 1977                     ∪2-6.5                            Page 2-79
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


6.6  Trigonometric Functions


sin                 SUBR 1 arg

     (sin x) gives the  trigonometric sine of x.  x  is in radians.  x may  be a
     fixnum or a flonum.


cos                 SUBR 1 arg

     (cos x) returns the cosine of x.  x is in radians.  x may be a fixnum  or a
     flonum.


atan                SUBR 2 args

     (atan x  y) returns  the arctangent  of x/y, in  radians.  x  and y  may be
     fixnums or flonums.  y may be 0 as long as x is not also 0.

























Page 2-80                            ∪2-6.6                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**


6.7  Random Functions


random              LSUBR 0 to 2 args

     (random) returns a random fixnum.

     (random nil) restarts the random sequence at its beginning.

     (random x), where x is a fixnum, returns a random fixnum between 0  and x-1
     inclusive.  A useful function is:

        (defun frandom ()
            (//$ (float (random 10000.)) 10000.0)))

     which returns a random flonum between 0.0 and 1.0.

     (random n1 n2) sets  the random number seed  from the pair of  integers n1,
     n2.


zunderflow          SWITCH

     If an intermediate  or final flonum  result in the  interpretive arithmetic
     functions  (times,  *$,  expt,  etc.)  is  too  small  in  magnitude  to be
     represented by the  machine, corrective action  will be taken  according to
     the zunderflow switch.

     If the value of zunderflow is non-nil, the offending result will be  set to
     0.0 and computation  will proceed.  If the  value of zunderflow is  nil, an
     error will be signalled.  nil is the initial value.

     In the pdp-10 implementation compiled code is not affected by zunderflow if
     the  arithmetic  in  question was  open-coded  by  the  compiler.  Instead,
     computation proceeds using a result with a binary exponent 256  higher than
     the correct exponent.  In  the Multics implementation zunderflow  works the
     same for compiled code as for interpreted code.

     See (sstatus divov), which controls division by zero (part 3.7).





December 5, 1977                     ∪2-6.7                            Page 2-81
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


6.8  Logical Operations on Numbers


   These  functions  may  be  used freely  for  bit  manipulation;  the compiler
recognizes them and produces efficient code.

boole               LSUBR 3 or more args

     (boole k x y) computes a bit by bit Boolean function of the fixnums x and y
     under the control of k.  k must  be a fixnum between 0 and 17  (octal).  If
     the  binary representation  of k  is  abcd, then  the truth  table  for the
     Boolean operation is:

              y
        ←←←←λ|←0λ←←←1λ←
          0| a  c
        x  |
          1| b  d

     If boole has more than three arguments, it goes from left to right; thus

                  (boole k x y z) = (boole k (boole k x y) z)

     The most common values for k are 1 (and), 7 (or), 6 (xor).  You can get the
     complement, or logical negation, of x by (boole 6 x -1).

           The following macros are often convenient:

        (defun logand macro (x)
          (subst (cdr x) 'f '(boole 1 . f)))

        (defun logor macro (x)
          (subst (cdr x) 'f '(boole 7 . f)))

        (defun logxor macro (x)
          (subst (cdr x) 'f '(boole 6 . f)))








Page 2-82                            ∪2-6.8                     December 5, 1977
**DRAFT**                           Numbers                            **DRAFT**


     Alternatively, these could be defined with macrodef (see part 6.2):

        (macrodef logand x (boole 1 . x))

        (macrodef logor x (boole 7 . x))

        (macrodef logxor x (boole 6 . x))


lsh                 SUBR 2 args

     (lsh x y), where x and y are fixnums, returns x shifted left y bits if y is
     positive, or  x shifted right  |y| bits if  y αis negative.   Zero-bits are
     shifted in to fill unused positions.  The result is undefined if |y|  > 36.
     The number 36 is implementation  dependent, but this is the number  used in
     both the Multics and pdp-10 implementations.  Examples:

                (lsh 4 1) => 10 (octal)
                (lsh 14 -2) => 3
                (lsh -1 1) => -2


rot                 SUBR 2 args

     (rot x y) returns as a fixnum the 36-bit representation of x,  rotated left
     y bits if y is positive, or rotated right |y| bits if y is negative.  x and
     y must be fixnums.   The results are undefined if  |y| > 36.  As  with lsh,
     the number 36 depends on the implementation.  Examples:

            (rot 1 2) => 4
            (rot -1 7) => -1
            (rot 601234 36.) => 601234
            (rot 1 -2) => 200000000000
            (rot -6 6) => -501


The following feature only exists in the pdp-10 implementation.

     The internal representation of flonums may be hacked using these functions.
     lsh or rot applied to  a flonum operates on the internal  representation of
     the  flonum and  returns a  fixnum  result.  For  example, (lsh  0.5  0) =>
     200400000000 (octal).  The following function also exists:


December 5, 1977                     ∪2-6.8                            Page 2-83
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


fsc                 SUBR 2 args

     (fsc  x y)  performs a  FSC instruction  on the  two numbers  x and  y, and
     returns the result as a flonum.  Consult the pdp-10 processor manual if you
     want to use this.

     x  and  y   may  be  fixnums  or   flonums;  fsc  just  uses   the  machine
     representations of the numbers.  If y is greater than 777777 octal, the FSC
     instruction is omitted and  the possibly-unnormalized flonum with  the same
     machine representation as x is returned.


































Page 2-84                            ∪2-6.8                     December 5, 1977
**DRAFT**                    Character Manipulation                    **DRAFT**




7.  Character Manipulation


7.1  Character Objects


   An  atomic symbol  with a  one-character pname  is often  called  a character
object  and  used to  represent  the ascii  character  which is  its  pname.  In
addition the atomic  symbol with a zero-length  pname represents the  ascii null
character.  Functions which take a character object as an argument  usually also
accept a string one character long or a fixnum equal to the ascii-code value for
the character.  Character objects are  always interned on the obarray  (see page
2-58), so they may be compared with the function eq.


ascii               SUBR 1 arg

     (ascii x), where x is a number, returns the character object for  the ascii
     code x.

     Examples:

     (ascii 101) => A

     (ascii 56) => /.


getchar             SUBR 2 args

     (getchar x n), where x is an  atomic symbol and n is a fixnum,  returns the
     n'th character  of x's pname;  n = 1  selects the leftmost  character.  The
     character is returned as a character object.  nil is returned if n  αis out
     of bounds.









December 5, 1977                     ∪2-7.                             Page 2-85
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


getcharn            SUBR 2 args

     getcharn is the same as getchar except that the character is returned  as a
     fixnum instead of a character object.


maknam              SUBR 1 arg

     maknam takes as its argument a list of characters and returns an uninterned
     atomic symbol whose pname is constructed from the list of  characters.  The
     characters  may  be  represented  either as  fixnums  (ascii  codes)  or as
     character objects.  Example:

                          (maknam '(a b 60 d)) => ab0d


implode             SUBR 1 arg

     implode is the  same as maknam except  that the resulting atomic  symbol is
     interned.  It is more efficient than doing (intern (maknam x)), although it
     is less efficient than plain maknam which should be used when  interning is
     not required.


readlist            SUBR 1 arg

     The argument to  readlist is a list  of characters.  The characters  may be
     represented either as fixnums  (ascii codes) or as character  objects.  The
     characters in the  list are assembled into  an S-expression as if  they had
     been typed  into read  (see part 5.1.)  If macro  characters are  used, any
     usage in the macro character function of read, readch, tyi, or  tyipeek not
     explicitly specifying an input  file takes input from  readlists's argument
     rather than from an I/O device or a file.  This causes macro  characters to
     work as you would expect.

        Examples:
            (readlist '(a b c)) => abc
            (readlist '( /( p r 151 n t /  /' f o o /) ))
                => (print (quote foo))  ;ascii 151 = "i"

     Note the use of the slashified special characters left  parenthesis, space,
     quote, right parenthesis in the argument to readlist.


Page 2-86                            ∪2-7.1                     December 5, 1977
**DRAFT**                    Character Manipulation                    **DRAFT**


explode             SUBR 1 arg

     (explode x)  returns a list  of characters, which  are the  characters that
     would be typed  out if (prin1 x)  were done, including slashes  for special
     characters but not including extra newlines inserted to  prevent characters
     from running  off the  right margin.   Each character  is represented  by a
     character object.  Example:

        (explode '(+ /12 3)) => ( /( + /  // /1  /2 /  /3 /) )
                ;Note the presence of slashified spaces in this list.


explodec            SUBR 1 arg

     (explodec x)  returns a list  of characters which  are the  characters that
     would be  typed out if  (princ x) were  done, not including  extra newlines
     inserted to prevent characters from running off the right  margin.  Special
     characters  are  not  slashified.   Each  character  is  represented  by  a
     character object.  Example:

        (explodec '(+ /12 3)) => ( /( + /  /1 /2 /  /3 /) )


exploden            SUBR 1 arg

     (exploden x)  returns a list  of characters which  are the  characters that
     would be  typed out if  (princ x) were  done, not including  extra newlines
     inserted to prevent characters from running off the right  margin.  Special
     characters are not slashified.   Each character is represented by  a number
     which is the ascii code for that character.  cf. explodec.  Example:

        (exploden '(+ /12 3)) => (50 53 40 61 62 40 63 51)


flatsize            SUBR 1 arg

     (flatsize x) returns  the number of characters  prin1 would use to  print x
     out.






December 5, 1977                     ∪2-7.1                            Page 2-87
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


flatc               SUBR 1 arg

     (flatc x) returns the number of characters princ would use to print  x out,
     without slashifying special characters.








































Page 2-88                            ∪2-7.1                     December 5, 1977
**DRAFT**                    Character Manipulation                    **DRAFT**


7.2  Character Strings


   These  character  string  functions  only exist  at  present  in  the Multics
implementation of Maclisp.  A predicate to test if your implementation has these
functions is
                            (status feature strings)

   These functions all accept atomic  symbols in place of strings  as arguments;
in this case  the pname of the  atomic symbol is used  as the string.   When the
value of one of these functions is described as a string, it is always  a string
and never an atomic symbol.  Also see the functions on page 2-56.


catenate            LSUBR 0 or more args

     The arguments are character strings.   The result is a string which  is all
     the arguments concatenated together.  Example:

          (catenate "foo" "-" "bar") => "foo-bar"


index               SUBR 2 args

     index is like the PL/I builtin function index.  The arguments are character
     strings.  The position  of the first occurrence  of the second  argument in
     the first is returned, or 0 if there is none.  Examples:

          (index "foobar" "ba") => 4
          (index "foobar" "baz") => 0
          (index "goobababa" "bab") => 4


stringlength        SUBR 1 arg

     The argument  to stringlength must  be a character  string.  The  number of
     characters in it is returned.  Examples:

          (stringlength "foo") => 3
          (stringlength "") => 0




December 5, 1977                     ∪2-7.2                            Page 2-89
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


substr              LSUBR 2 or 3 args

     This is like the  PL/I substr builtin.  (substr x  m n) returns a  string n
     characters long, which is a portion of the string x beginning with its m'th
     character and proceeding for n characters.  m and n must be fixnums, x must
     be a string.

     (substr x m) returns  the portion of the  string x beginning with  its m'th
     character and continuing until the end of the string.  Examples:

          (substr "foobar" 3 2) => "ob"
          (substr "resultmunger" 6) => "tmunger"


get←pname           SUBR 1 arg

     (get←pname x) returns the pname of  x as a character string.  x must  be an
     atomic symbol.


make←atom           SUBR 1 arg

     make←atom returns an atomic symbol,  uninterned, whose pname is given  as a
     character string argument.  Example:

          (make←atom "foo") => foo      ;which is not eq to a
                                        ;foo that is read in.

















Page 2-90                            ∪2-7.2                     December 5, 1977
**DRAFT**                            Arrays                            **DRAFT**




8.  Arrays


   As explained in part 1.2, an array is a group of cells which may contain Lisp
objects.  The individual cells are selected by numerical subscripts.

   An array is  designated by a special  atomic object called  an array-pointer.
Array-pointers can be returned by the array-creation functions array and *array.
An array-pointer  may either be  used directly  to refer to  the array,  or, for
convenience in  referring to  the array  through input/output  media, it  may be
placed on the property list of  an atomic symbol under the indicator  array, and
then that symbol can be used as the name of the array.

   There are several types of arrays.  The main types are ordinary arrays, whose
cells can hold any type of object, and number arrays, whose cells can  only hold
numbers.  Number arrays permit more efficient code to be compiled  for numerical
applications, and take less space than an ordinary array which contains the same
number of  numbers.  See  the array*  declaration (part  4.2) and  the arraycall
function (page 2-14).

   When an array is created its  type must be declared by giving a  "type code."
The type code  for ordinary arrays  is t.  For number  arrays, the type  code is
either fixnum or flonum.   A particular number array  can only hold one  type of
numbers because its cells contain the machine representation of the  number, not
the Lisp-object representation.

   Some other types  of arrays are:   un-garbage-collected arrays, with  a type-
code of  nil, which are  the same as  ordinary arrays except  that they  are not
protected  by  the garbage  collector  and  therefore can  be  used  for certain
esoteric  hacks;  obarrays, with  a  type-code  of obarray,  which  are  used to
maintain tables of known atomic symbols  so that the same atomic symbol  will be
referenced when the same pname is typed in; and readtables, with a  type-code of
readtable, which  are used to  remember the syntax  specifications for  the Lisp
input reader.  Normally, there is  only one readtable and one  obarray, supplied
by the  system, but the  user may create  additional readtables and  obarrays in
order to  provide special  non-Lisp environments or  to gain  additional control
over the Lisp environment.   Lisp functions such as read  can be made to  use an
additional readtable or obarray by re-binding the variable readtable or obarray,
respectively.



December 5, 1977                     ∪2-8.                             Page 2-91
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   An array-pointer may  also be dead,  in which case it  does not point  to any
array.  One of the functions array, *array, or *rearray may be used  to revivify
a dead array-pointer.

   The functions array and *array are used to create arrays.  The first argument
may be an atomic  symbol, which makes that atomic  symbol the name of  an array,
putting an array-pointer  on its property  list, or redefining  an array-pointer
that was already on the property list to point to the new  array.  Alternatively
the first argument may be an  array pointer, which causes that array  pointer to
be redefined to point to a new array, or it may be nil, which causes a new array
pointer to be  created and returned.  Except  in the latter case,  array returns
its first argument.  *array always  returns the array pointer, never  the atomic
symbol.

   A readtable or an obarray may not be created with  user-specified dimensions.
The dimensions are always determined  by Lisp.  Other types of arrays  allow any
reasonable number (at least 3,  anyway) of dimensions to be specified  when they
are  created.  The  subscripts range  from 0  up to  1 less  than  the dimension
specified.

   Ordinary  and un-garbage-collected  arrays  are initialized  to  nil.  Fixnum
arrays are initialized to 0.  Flonum arrays are initialized to 0.0.

   Obarrays are initialized  according to the third  argument given to  array or
*array.  nil causes a completely empty obarray to be created.  Not even nil will
be interned on this obarray.   t causes the current obarray (value of the symbol
obarray)  to be  copied.  An  array-pointer which  is an  obarray, or  an atomic
symbol which names an  obarray, causes that obarray  to be copied.  If  no third
argument is given, the current obarray is copied.

   Readtables are initialized  in a similar fashion.   If the third  argument of
array or *array is nil, then the current readtable is copied.  If it is  t, then
the  readtable  being  created  is  initialized  to  the  initial  standard Lisp
readtable,  including the  macro characters  ' and  ;.  (Note  that this  is the
opposite of the t-nil convention  for obarrays.  This is for  compatibility with
the makreadtable function, which  no longer exists.) An array-pointer  or symbol
of a readtable to be copied may  also be given.  If no third argument  is given,
the current readtable is copied.

   An array-pointer may be redefined  to an entirely different type and  size of
array, using  the *array  function.  It  remains the  same array-pointer,  eq to
itself.  If a variable was setq'ed to the array-pointer, that variable  will now


Page 2-92                            ∪2-8.                      December 5, 1977
**DRAFT**                            Arrays                            **DRAFT**


indicate the  new array.   If a symbol  has that  array-pointer on  its property
list, it will now be the name of the new array.

   The *rearray  function can  be used to  redefine the  size or  arrangement of
dimensions of an array without losing its contents, or to make  an array-pointer
not point to any array (become dead).  If there is only one argument, the array-
pointer is  killed, the  array's contents are  discarded, and  the array-pointer
becomes a "dead array" as described  above.  *array may now be used  to redefine
it as a new array.

   If more than one argument is  given to *rearray, they are the  same arguments
as to *array.  *rearray with more than one argument cannot be used to change the
type of an array, and cannot operate on a readtable or an obarray, but it can be
used  to  change  the  dimensions  of an  array.   The  modified  array  will be
initialized from its old contents rather than  nil, 0, or 0.0.  The elements are
taken  in row-major  order for  initialization purposes,  and if  there  are not
enough,  nil, 0,  or 0.0  will be  used to  fill the  remaining elements  of the
modified array, according to the type.

   The Multics implementation also has a type of arrays called  external arrays.
External  arrays  reside  in  a Multics  segment  rather  than  within  the Lisp
environment.  They  behave much like  fixnum arrays, and  should be  declared as
such to the compiler.  To create an external array, use a form such as

                      (array foo external pointer length)

The pointer is  a packed pointer  to the beginning of  the array, i.e.  a fixnum
whose first six octal digits are  the segment number and whose second  six octal
digits are the word  address.  The length is the  number of words in  the array.
External arrays can only have  one dimension, can only contain fixnums,  and are
not initialized  when they are  created.  They cannot  usefully be saved  by the
save function.  This  type of array can  be used for communication  between Lisp
programs and  Multics programs  or subsystems written  in other  languages, when
large amounts of numerical data or machine words must be passed back  and forth.
See also defpl1 (part 4.6).

   If you want the range of subscripts on arrays to be checked, it  is necessary
to set the *rset flag non-nil (i.e. run in (*rset t) mode) and to avoid  the use
of in-line array accessing  (i.e. the array* declaration) in  compiled programs.
The amount of checking performed when *rset is nil and/or compiled code  is used
depends on the implementation.



December 5, 1977                     ∪2-8.                             Page 2-93
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



      Here is an example of a use of arrays:

      (defun matrix-multiply (arr1 arr2 result)
         (and (eq (typep arr1) 'symbol)       ;convert arguments
              (setq arr1 (get arr1 'array)))  ;to array-pointers
         (and (eq (typep arr2) 'symbol)
              (setq arr2 (get arr2 'array)))
         (and (eq (typep result) 'symbol)
              (setq result (get result 'array)))
         (do ((ii (cadr (arraydims result)))  ;get relevant
              (jj (caddr (arraydims result))) ;dimensions
              (kk (cadr (arraydims arr2))))
             ()
           (do i 0 (1+ i) (= i ii)        ;result := arr1 x arr2
             (do j 0 (1+ j) (= j jj)
               (do ((k 0 (1+ k))
                    (r 0.0))
                   ((= k kk)
                    (store (arraycall flonum result i j) r))
                 (setq r (+$ r (*$ (arraycall flonum arr1 i k)
                                   (arraycall flonum arr2 k j)
                    )))))))
         result)


*array              LSUBR 3 or more args

     (*array x  y b1 b2  ... bn) defines  x to be  an n-dimensional  array.  The
     first subscript may  range from 0 to  b1 minus 1, the  second from 0  to b2
     minus 1,  etc.  y  is the  type of array,  as explained  above.  It  may be
     chosen from among:  t, nil, fixnum, flonum, readtable, obarray.


array               FSUBR

     (array x y b1 b2 ... bn) has the same effect as (*array (quote x) (quote y)
     b1 b2 ... bn).  This special form is provided for your typing convenience.






Page 2-94                            ∪2-8.                      December 5, 1977
**DRAFT**                            Arrays                            **DRAFT**


*rearray            LSUBR 1 or more args

     *rearray is used to redefine the dimensions of an array.

     (*rearray x) kills the array-pointer  x, or the array-pointer which  is the
     array property of the atomic symbol x.  The storage used by  the associated
     array is reclaimed.  The value returned is  t if x was an array, nil  if it
     was not.

     (*rearray x type dim1 dim2 ...  dimn) is like (*array x type dim1  dim2 ...
     dimn) except that the contents of the previously existing array named x are
     copied into  the new array  named x.  If  it is a  multi-dimensional array,
     row-major order  is used.  This  means the last  subscript varies  the most
     rapidly as the array is traversed.


store               FSUBR

     The special form (store array-ref value) is used to store an object  into a
     particular cell  of an array.   The first element  of the  form, array-ref,
     must be a subscripted reference to an array, or an invocation of arraycall.
     By coincidence, certain other forms work as array-ref, for  instance (apply
     f l)  where f  turns out  to be an  array.  The  second element,  value, is
     evaluated and stored into the specified cell of the array.  store evaluates
     its second "argument" before its first "argument".

        Examples:

        (store (data i j) (plus i j))

        (store (sine-values (fix (*$ x 100.0)))
               (sin x))

        (store (arraycall fixnum az i j) 43)


arraydims           SUBR 1 arg

     (arraydims x),  where x  is an array-pointer  or an  atomic symbol  with an
     array property, returns a list of  the type and bounds of the  array.  Thus
     if A was defined by (array A t 10 20),

        (arraydims 'A) => (t 10 20)

December 5, 1977                     ∪2-8.                             Page 2-95
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


fillarray           SUBR 2 args

     (fillarray a l) fills the array  a with consecutive items from the  list l.
     If the array is too short to  contain all the items in the list,  the extra
     items are ignored.  If the list is too short to fill up the array, the last
     element of  the list is  used to fill  each of the  remaining cells  in the
     array.

     (fillarray x y) fills the array x  from the contents of the array y.   If y
     is bigger than x, the extra elements are ignored.  If y is smaller  than x,
     the rest  of x is  unchanged.  x and  y must be  atomic symbols  which have
     array properties, or  array-pointers.  The two arrays  must be of  the same
     type, and they may not be readtables or obarrays.

     The list-into-array case of fillarray could have been defined by:

        (defun fillarray (a x)
            (do ((x x (or (cdr x) x))
                 (n 0 (1+ n))
                 (hbound (cadr (arraydims a))))
                ((= n hbound))
              (store (a n) (car x))
             )
            a)

     An  extension to  the above  definition is  that fillarray  will  work with
     arrays of more  than one dimension, filling  the array in  row-major order.
     fillarray returns its first argument.


listarray           LSUBR 1 or 2 args

     (listarray array-name) takes the elements of the array specified  by array-
     name and returns them as the elements of a list.  The length of the list is
     the size of the array and the elements are present in the list in  the same
     order as they are stored  in the array, starting with the  zero'th element.
     If the array has more than one dimension row-major order is used.

     (listarray  array-name n)  is the  same, except  that at  most the  first n
     elements will be listed.

     array-name  may be  an array-pointer  or an  atomic symbol  with  an array-
     property.

Page 2-96                            ∪2-8.                      December 5, 1977
**DRAFT**                            Arrays                            **DRAFT**


   Number arrays  may be efficiently  saved in the  file system and  restored by
using the functions loadarrays and dumparrays.


loadarrays          SUBR 1 arg

     (loadarrays file-spec) reloads the arrays  in the file, and returns  a list
     of 3-lists, of the form:

                         ( (newname oldname size) ...)

     newname  is a  gensym'ed atom,  which is  the name  of the  reloaded array.
     (newname ought to be an array-pointer, but this function was defined before
     array-pointers were invented.)  oldname is the name  the array had  when it
     was dumped.  size is the number of elements in the array.


dumparrays          SUBR 2 args

     (dumparrays (array1 array2 ...) file-spec) dumps the listed arrays into the
     specified file.  The arrays must be fixnum or flonum arrays.

     In both of  the above, the file-spec  argument is dependent on  the system.
     In ITS or DEC-10 Lisp, the file-spec is a list of zero to four items, as in
     uread, and the same defaults  apply.  In Multics Lisp, the file-spec  is an
     atomic symbol or  a string which  gives the pathname  of the segment  to be
     used.   The defaults  and other  features of  the Lisp  I/O system  are not
     applied.  Only a segment may be specified, not a stream.

     As  a  special  compatibility  feature,  in  Multics  Lisp  loadarrays will
     recognize a pdp-10 dumparrays file.   (One can be moved to  Multics through
     the ARPA Network File Transfer  Protocol if the "type image"  and "bytesize
     36" commands are employed.)  The pnames will be converted to lower case and
     flonums will be converted to the H6880 machine  representation.  dumparrays
     can create a  file which pdp-10  loadarrays can read,  including upper-case
     pnames and pdp-10 format flonums, if it is invoked as follows:

     (dumparrays (array1 array2...) '(pdp10 file-spec))






December 5, 1977                     ∪2-8.                             Page 2-97
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**














































Page 2-98                            ∪2-8.                      December 5, 1977
**DRAFT**                      Mapping Functions                       **DRAFT**




9.  Mapping Functions


   Mapping is a type of iteration in which a function is successively applied to
pieces of a list.  There are several options for the way in which the  pieces of
the  list are  chosen and  for what  is done  with the  results returned  by the
applications of the function.

   For example, mapcar operates on successive elements of the list.  As  it goes
down the list, it calls the function giving it an element of the list as its one
argument:  first the car, then the cadr, then the caddr, etc.,  continuing until
the end of the list is reached.   The value returned by mapcar is a list  of the
results of  the successive  calls to  the function.   An example  of the  use of
mapcar would  be mapcar'ing the  function abs  over the list  (1 -2  -4.5 6.0e15
-4.2).  The result is (1 2 4.5 6.0e15 4.2).

   The form of a call to mapcar is

                                  (mapcar f x)

where f is the function  to be mapped and x is  the list over which it is  to be
mapped.  Thus the example given above would be written as

               (mapcar 'abs
                       '(1 -2 -4.5 6.0e15 -4.2))

This has been generalized to allow a form such as

                            (mapcar f x1 x2 ... xn)

In this case f must be a function of n arguments.  mapcar will proceed  down the
lists x1, x2, ..., xn in parallel.   The first argument to f will come  from x1,
the second from  x2, etc.  The iteration  stops as soon as  any of the  lists is
exhausted.

   There  are five  other  mapping functions  besides mapcar.   maplist  is like
mapcar except that the function is  applied to the list and successive  cdr's of
that list rather than to successive elements of the list.  map and mapc are like
maplist and mapcar respectively except  that the return value is the  first list
being  mapped over  and the  results of  the function  are ignored.   mapcan and


December 5, 1977                     ∪2-9.                             Page 2-99
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


mapcon are  like mapcar and  maplist respectively except  that they  combine the
results of the function using nconc instead of list.  That is,

      (defun mapcon (f x y)
          (apply 'nconc (maplist f x y)))

Of course, this definition is far less general than the real one.

   Sometimes a do or a straight  recursion is preferable to a map;  however, the
mapping  functions should  be used  wherever they  naturally apply  because this
increases the clarity of the code.

   Often f will be a  lambda-type functional form rather than  the atomic-symbol
name of a function.  For example,

              (mapcar '(lambda (x) (cons x something)) some-list)

   The functional argument to a  mapping function must be acceptable to  apply -
it cannot be a macro.  A fexpr or an fsubr may be acceptable however the results
will be bizarre.  For instance, mapping set works better than mapping  setq, and
mapping cond is unlikely to be useful.

   It is permissible (and often  useful) to break out of  a map by use of  a go,
return, or throw in a  lambda-type function being mapped.  This is  a relaxation
of the usual prohibition against "non-local" go's and return's.  If go or return
is used  the program may  have to  be compiled with  the (mapex  t) declaration,
depending on the implementation, so watch out!  Consider this function  which is
similar  to  and,  except that  it  works  on a  list,  instead  of  on separate
arguments.

     (defun andl (x)
      (catch
       (progn
        (mapc (function (lambda (y)
                          (or y (throw nil the-answer)) ))
              x)
        t)
       the-answer))

   Admittedly this could be better expressed as a do:




Page 2-100                           ∪2-9.                      December 5, 1977
**DRAFT**                      Mapping Functions                       **DRAFT**



     (defun andl (x)
       (do ((y x (cdr y)))
           ((null y) t)
        (or (car y) (return nil))
      ))


      Here is a table showing the relations between the six map functions.

                                    applies function to

                               |  successive  |   successive  |
                               |   sublists   |    elements   |
                ---------------+--------------+---------------+
                    its own    |              |               |
                    second     |     map      |     mapc      |
                   argument    |              |               |
                ---------------+--------------+---------------+
                  list of the  |              |               |
      returns      function    |    maplist   |    mapcar     |
                    results    |              |               |
                ---------------+--------------+---------------+
                  nconc of the |              |               |
                    function   |    mapcon    |    mapcan     |
                    results    |              |               |
                ---------------+--------------+---------------+


mapatoms            LSUBR 1 or 2 args

   (mapatoms  fn obarray)  applies the  function fn  to all  the symbols  on the
specified obarray.  If  the second argument is  omitted, the current  obarray is
used.  Note  that the obarray  argument must be  an array-pointer, not  a symbol
which names an array.  The symbol  obarray is bound to the obarray  being mapped
over during the execution of mapatoms.

   This function exists because some of the cells in an obarray contain lists of
symbols and others contain single  symbols, and user programs shouldn't  have to
know this.
Example:



December 5, 1977                     ∪2-9.                            Page 2-101
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



      (mapatoms
        (function
          (lambda (x)
            (and (sysp x)
                 (print (list x (sysp x) (args x))) ))))






































Page 2-102                           ∪2-9.                      December 5, 1977
**DRAFT**                          The System                          **DRAFT**


                              Part 3 - The System

                               Table of Contents



1.      The System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1
1.1     The Top Level Function . . . . . . . . . . . . . . . . . . . . . . . 3-1
1.2     Breakpoints  . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-5
1.3     Control Characters . . . . . . . . . . . . . . . . . . . . . . . . . 3-9
1.4     Exceptional Condition Handling . . . . . . . . . . . . . . . . . . .3-15
1.4.1   The LISP Error System  . . . . . . . . . . . . . . . . . . . . . . .3-15
1.4.2   User Interrupts  . . . . . . . . . . . . . . . . . . . . . . . . . .3-16
1.4.3   Catalogue of User Interrupt Channels . . . . . . . . . . . . . . . .3-19
1.4.4   Autoload . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3-26
1.5     Debugging  . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3-28
1.5.1   Binding, Pdl Pointers, and the Evaluator  . . . . . . . . . . . . . 3-28
1.5.2   Functions for Debugging  . . . . . . . . . . . . . . . . . . . . . .3-28
1.5.3   The Trace Package  . . . . . . . . . . . . . . . . . . . . . . . . .3-35
1.5.4   The Stepper  . . . . . . . . . . . . . . . . . . . . . . . . . . . .3-40
1.5.5   The MAR Break Feature  . . . . . . . . . . . . . . . . . . . . . . .3-56
1.6     Storage Management . . . . . . . . . . . . . . . . . . . . . . . . .3-60
1.6.1   Garbage Collection . . . . . . . . . . . . . . . . . . . . . . . . .3-60
1.6.2   Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3-61
1.6.3   Storage Control Functions  . . . . . . . . . . . . . . . . . . . . .3-64
1.6.4   Dynamic Space and Pdl Expansion  . . . . . . . . . . . . . . . . . .3-65
1.6.5   Initial Allocation . . . . . . . . . . . . . . . . . . . . . . . . .3-66
1.7     Implementing Subsystems with Maclisp . . . . . . . . . . . . . . . .3-69
1.7.1   Entering LISP  . . . . . . . . . . . . . . . . . . . . . . . . . . .3-69
1.7.2   Saving an Environment  . . . . . . . . . . . . . . . . . . . . . . .3-70
1.7.3   Gaining and Keeping Control  . . . . . . . . . . . . . . . . . . . .3-73
1.7.4   Purity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3-74
1.8     Miscellaneous Functions  . . . . . . . . . . . . . . . . . . . . . .3-79
1.8.1   The Status Functions . . . . . . . . . . . . . . . . . . . . . . . .3-79
1.8.2   Time   . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-100
1.8.3   Escaping from LISP . . . . . . . . . . . . . . . . . . . . . . . . 3-101
1.8.4   Additional Functions . . . . . . . . . . . . . . . . . . . . . . . 3-103







December 6, 1977                                                        Page 3-1
**DRAFT**                          The System                          **DRAFT**




1.  The System


1.1  The Top Level Function


   The following function is an approximation to what Maclisp does when it is at
its "top level."


































December 6, 1977                     ∪3-1.                              Page 3-1
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



      (defun standard-top-level nil
             (prog (↑q ↑w ↑r evalhook base ibase ... )
              errors          ;errors, uncaught throws, etc. come here
              ↑g              ;ctrl/G quits come here
                   (reset-bound-vars-and-restore-pdls)
                   (setq ↑q nil)
                   (setq ↑w nil)
                   (setq evalhook nil)
                   (nointerrupt nil)
                   (do-delayed-tty-and-alarmclock-interrupts)
      ;Recall that errors do (setq // errlist) so lambda-binding
      ;  errlist will work properly.  See errlist.
           (mapc (function eval) //)
           (or (status linmode)(terpri))
           (do ((eof (list nil))      ;internal variables
                (prt '* *))
               (nil)          ;do forever (until ↑g or error)
               (setq * (cond ((status toplevel)
                              (eval (status toplevel)))
                             (t (terpri)
                                (cond (prin1 (funcall prin1 prt))
                                      (t (prin1 prt)))
                                (typ 40)
                                (setq (do ((form))(nil)
                                          (setq form
                                              (cond (read
                                                     (funcall read
                                                              eof))
                                                    (t (read eof))))
                                          (or (eq form eof)
                                              (return form))
                                          (terpri)))
                                (and (null read)
                                     (atom -)
                                     (is-a-space (tyipeek))(tyi))
                                ((lambda (+) (eval -))
                                 (prog2 nil + (setq + -)))))))))

which causes a "read-eval-print loop,"  i.e. each S-expression that is  typed in
gets evaluated  and the value  is printed, then  the next S-expression  is read.
Errors and ↑g quit  to top level.  That  is they reinitialize and  then re-enter


Page 3-2                             ∪3-1.1                     December 6, 1977
**DRAFT**                          The System                          **DRAFT**


this  loop, printing  a *  (but not  destroying the  value of  the  variable *).
Notice that there  is a place in  the middle where the  user can insert  his own
special form to be evaluated, using (sstatus toplevel).  It is also  possible to
change just the reader or just  the printer by setq'ing read or prin1.   See the
sstatus function (page 3-79).

   Variables used by the top-level read-eval-print loop:

*                   VARIABLE

     Contains the  last S-expression  printed out  by the  read-eval-print loop,
     that is, the value of the last  form typed in.  This is true even  after an
     error return to top level, allowing  one to refer to the value  printed out
     before the aborted computation.

+                   VARIABLE

     Contains the last S-expression typed in.  This can be used to edit it or to
     do it  over again.   (Notice how +  is bound  in the  read-eval-print loop.
     This causes + to receive  the correct value even if the  evaluation aborts,
     since an error or ↑g quit will undo the binding.)

-                   VARIABLE

     Contains the  current S-expression  typed in.   This can  be used  by user-
     written error handlers.  It can't be usefully accessed by expressions typed
     in, since it is set before the expression is evaluated.

   By special  arrangement the  values of  +, *,  and -  are preserved  across a
break.  When the break is first entered these have the values for the  last top-
level operation,  during the break  they behave  the same as  at top  level, and
after the break returns they are restored to the values for the top  level loop.
(See break, page 3-5).

/                   VARIABLE

     / is used to temporarily hold the value of errlist when an error returns to
     top level.   This is  so that  lambda-binding errlist  will have  an effect
     (assuming no one lambda-binds /).  Note that / must be typed in as // since
     the slash character is special to the LISP reader.




December 6, 1977                     ∪3-1.1                             Page 3-3
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


errlist             VARIABLE

     The value of errlist  is a list of  forms which are evaluated  when control
     returns to top level either because  of an error or when an  environment is
     initially  started.  It  doesn't apply  if the  environment started  up was
     saved  using (suspend).   This  feature is  used to  provide  special error
     handling for subsystems written in LISP.

     The symbol errlist  is evaluated to  get the list  of forms in  the binding
     context in which the error occurred, but the forms themselves are evaluated
     in the top-level binding context.

     Example:

        ((lambda (errlist)
                 (putprop 'foo 'bar 'baz)
                 (hack)
                 (remprop 'foo 'baz))
         (cons '(remprop 'foo 'baz)
               errlist))


     The property list of foo will be properly restored even if  the computation
     (hack) is aborted.




















Page 3-4                             ∪3-1.1                     December 6, 1977
**DRAFT**                          The System                          **DRAFT**


1.2  Breakpoints


   Breakpoints are a mechanism to allow the user to gain control at any point in
a program.  Use of the function break causes a read-eval-print loop,  similar to
the one at top  level, to be entered.  (This  is also called a break  loop.) The
user may evaluate any S-expressions, inspect the bindings of variables, and exit
from the break in several ways.   Normal execution then proceeds.  (See ,  , and
page 3-31)

   This mechanism can  be used to permit  human intervention when  an unexpected
condition occurs.  It is used in this way by the Maclisp error system.   See the
section on Exceptional  Condition Handling, page 3-15.   A break loop  makes the
full power of the LISP interpreter available for debugging.


break               FSUBR

     (break tag pred) evaluates pred, but not tag.  If the value of pred  is not
     nil, the state of  the I/O system is saved,  ";bkpt tag" is typed  out, and
     control  returns to  the terminal.   We say  that a  "break loop"  has been
     entered.  tag may be  any object.  It is used  only as a message  typed out
     (using princ)  to identify  the break.  It  is not  evaluated.  If  pred is
     omitted, t is  assumed.  Thus (break tag)  is equivalent to (break  tag t).
     (break tag nil) returns nil, and produces no action whatsoever.

     A break loop is a read-eval-print loop similar to top level.  break does an
     errset so that errors cannot cause an abnormal return from the break.  A ↑x
     quit, which causes an ordinary error, will thus return to the break loop if
     used to  interrupt a  computation started in  the break  loop.  A  ↑g quit,
     however, returns back  to LISP top  level, resetting the  environment using
     the errlist, as described above.

     Two forms, $P and (return x), may be typed in a break loop.  If $P is typed
     in, break returns nil and execution continues.  This "$P" is <dollar>  P in
     the Multics implementation, but <altmode> P in the  PDP-10 implementations,
     followed of course in either case by a <space> or <newline>  as appropriate
     (see (status linmode)).  (An atom other than $P can be used to perform this
     function  by changing  the  value of  $P  to another  (non-nil)  atom.  The
     initial value of $P is always $P).

     If (return x) is typed in, break evaluates x and returns that value.  If as


December 6, 1977                     ∪3-1.2                             Page 3-5
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     a  result  of  the  evaluation  of a  typed-in  form,  (throw  x  break) is
     evaluated,  break  returns  x  as its  value.   (Notice  the  distinction -
     executing a form (return x) does not return from break unless it  was typed
     directly at the break loop.)  See return, .

     When break returns, the state of the I/O system is restored.

     An approximate LISP definition of  what break does follows.  Note  that the
     user program can modify this by using (sstatus breaklevel).



































Page 3-6                             ∪3-1.2                     December 6, 1977
**DRAFT**                          The System                          **DRAFT**



(defun break fexpr (x)
       (*break (eval (cadr x)) (car x)) ;note argument reversal

(declare (special ↑q ↑w evalhook * + -))

(defun *break (breakp breakid)
 (and breakp
  (do ((↑q nil) (↑w nil) (evalhook nil) (terpri t) (* *) (+ +) (- -))
      ()                    ;bind key variables
   (terpri msgfiles)        ;msgfiles arguments
   (princ '|;bkpt | msgfiles)   ;  used for
   (princ breakid msgfiles)     ;  Newio only
   (terpri msgfiles)
   (setq + -)               ;last form typed
   (return
    (prog2 nil
     (catch
      (do () (nil)          ;do forever (until throw)
        (errset
          (do ((eof (list nil)) (form))
              (nil)
           (cond ((status breaklevel)
                  (eval (status breaklevel)))
                 (t (setq form (cond (read (funcall read eof))
                                     (t (read eof))))
                    (and (null read)
                         (atom form)
                         (is-a-space (tyipeek))
                         (tyi))
                    (cond ((eq form eof) (terpri))
                          ((and $P (eq form '$P)) (throw nil break))
                          ((eq (car form) 'return)
                           (throw (eval (cadr form)) break))
                          (t (setq - form)
                             (print
                              (setq * ((lambda (+) (eval form))
                                       (prog2 nil + (setq + -)))))
                             (terpri))))))))
     break)
     (terpri)
  ))))


December 6, 1977                     ∪3-1.2                             Page 3-7
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   The arguments  to break  are a breakpoint  identification and  (optionally) a
break  switch.  If  the break  switch evaluates  to nil,  then nil  is returned.
Otherwise,  the variables  ↑q, ↑w,  evalhook and  terpri are  bound to  nil, the
variables *, +, and - are bound to their current values, and the  message ";bkpt
<breakid>" is printed.  A read-eval-print loop similar to the top level  loop is
then entered.  This break loop is surrounded by an errset.  Errors or  typing ↑x
merely cause the break loop to be re-entered.  The value of  (status breaklevel)
serves a function similar to that of (status toplevel) in the top level loop.

   As each form is read in the default break loop, there are four cases:

     1.  End of file.  For console input this merely indicates rubout beyond the
        number of input characters.  Whether input is from console or elsewhere,
        the (terpri) is done and the reader is entered.

     2.  The  form is the  atom $P  or eq to  the non-nil value  of $P.   nil is
        returned from the break.

     3.  The form is (return  value).  The form value is evaluated  and returned
        from the break.

     4.  Otherwise the form is evaluated and the result printed out in  a manner
        analogous to the  top level read-eval-print  loop.  The variables  +, -,
        and * are updated appropriately.  (Recall, however, that they were bound
        on entry to *break, and so will be restored eventually.)

   The way to return  from a break is  to do a throw  with a tag of  break; this
will return from the catch which surrounds the break loop.  This is how  cases 2
and 3 return their values; case 4 may also cause a return from the break.















Page 3-8                             ∪3-1.2                     December 6, 1977
**DRAFT**                          The System                          **DRAFT**


1.3  Control Characters


   LISP can be directed to take certain actions by entering "control characters"
from the terminal.  The  difference between control characters and  normal input
is that control characters take effect as soon as they are entered, while normal
input only takes effect when LISP asks for it, by use of functions such as read,
or by being in the top level read-eval-print loop or in a break loop.

   Control  characters can  be  typed in  from  the terminal  according  to some
procedure that depends on the  implementation.  A program can mimic  the effects
of the various  control characters by  directly calling the  function associated
with the particular control key (see below).

   Although control characters are usually processed as soon as they  are typed,
they may be delayed if there is  a garbage collection in progress or LISP  is in
(nointerrupt tty) mode - see the nointerrupt function (page 3-19).

                    Entering Control Characters in ITS LISP

   In the ITS implementation of Maclisp, control characters are entered by means
of the "CTRL" key  on the terminal.  For  example, CTRL/g is entered  by holding
down "CTRL" and  striking the "G" key.   Control characters normally echo  as an
uparrow or circumflex followed by the character.

   In  Newio, any  character at  all may  be made  an interrupt  character.  See
(sstatus tty) and (sstatus ttyint).

   Interrupt characters  are also  read as  part of  the terminal  input stream.
Normally they are marked in the readtable as "worthless" characters.

                  Entering Control Characters in TOPS-10 LISP

   Most control characters may be entered in the same way as in ITS LISP if LISP
is currently read'ing from the terminal.  If a LISP program is actively running,
it is necessary to first gain  its attention by typing the CTRL/c  character one
or two times, thereby returning to the monitor.  The monitor command REENTER may
then be used to re-enter the  LISP.  LISP will print "?↑" and read  a character,
which may be a control character or a character whose "control" meaning is to be
used.  Thus typing either  ↑g or "G" will cause  the ↑g interrupt to  occur.  If
LISP is not ready to take an interrupt, there may be a delay before the  "?↑" is
printed.

*****SAILB*****

                  Entering Control Characters in SAIL LISP

   Control characters are entered by first getting the attention LISP by  typing
"<CALL>" and then "REENTER". At this  point control characters will be  accpeted
as in the  standard TOPS-10  implementation. Alternatively, one  can enable  the
<eci>I interrupt  and have  an  interrupt routine  simulate the  <CALL>  REENTER
sequence.

*****SAILE*****
December 6, 1977                     ∪3-1.3                             Page 3-9
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


                  Entering Control Characters in Multics LISP

   In the Multics implementation of Maclisp, one signals one's desire to enter a
"control" character by  hitting the "attention" key  on the terminal.   (This is
called "break,"  "interrupt," "attn," "quit,"  etc. on different  terminals.  If
Multics is being accessed through the ARPA network, an "interrupt process" (@S S
or @S I P from a TIP) signal should be transmitted.) LISP responds by typing out
"CTRL/".  Now you may type one letter from the list later in this section, which
will be interpreted to have its "control" meaning.  This control  character must
be followed by a newline.

   It is  also possible to  enter "control" characters  from an  input character
stream, which may have its source at the terminal or in an exec←com, without the
use of the "attention" key.  The desired control character is prefixed by a \036
character.  If two of these prefix characters occur together, one \036 character
is  read  and  no  "control"  action  is  performed.   Otherwise,  the character
following the \036 is processed as a control character, then reading continues.

   Control characters will be accepted  in upper or lower case.   All characters
other than those with defined meanings are rejected with an error message.  Only
one control character may be entered at a time.

      Example for Multics LISP:
              (lines containing user input are preceded by ">>>")

      >>>   (defun loop (x) (loop (add1 x)))
              loop
      >>>     (loop 0)
                      function runs for a long time,
      >>>     <ATTN>  then user hits attention button.
      >>>     CTRL/B  LISP types "CTRL/", user types "B"
      >>>     ;bkpt ↑b        system enters break loop
      >>>     x       user looks at value of x
              4067
      >>>     <ATTN>  user hits attention button again
      >>>     CTRL/G  and returns to top level
              Quit
              *
When  a "user  interrupt" is  caused, if  the interrupt  is not  enabled nothing
happens.  If the interrupt is enabled, then a user-specified function is called.
The interrupt may be enabled  by binding the appropriate symbol to  the function
to handle it, or by using the (sstatus ttyint) function (page 3-80).


Page 3-10                            ∪3-1.3                     December 6, 1977
**DRAFT**                          The System                          **DRAFT**


   In the following descriptions  of control characters, those which  can always
be processed immediately,  even during a  garbage collection or  in (nointerrupt
'tty) mode, will be indicated by an "!" for the ITS and TOPS-10 implementations,
and an "*" for the Multics Implemention.  When appropriate, equivalent LISP code
is given for producing the same result from a user program.

Control Characters that have initially defined meanings in all implementations:

  B !*    runs a break  loop with breakid "↑b"  (see break, page 3-5).   (In the
          PDP-10 Oldio implementation ↑h is used instead of ↑b.) (break ↑b).

  C !*    sets the value  of the atom ↑d  to nil, turning off  garbage collector
          messages.   Because ↑c  is trapped  by DEC-10  monitors,  this control
          character can  only be  typed using the  REENTER mechanism  and typing
          "C".  (setq ↑d nil).

  D !*    sets  the value  of the  atom ↑d  to t,  turning on  garbage collector
          messages.  (setq ↑d t)

  G       quits back to the top level of LISP, rebinding all variables  to their
          global values, resetting various system variables, and  evaluating the
          errlist forms.  This is used  to stop a running program when  there is
          no intention of  restarting it again.  (Prints  out a *; see  the "top
          level" function and the ↑g function.)  H is used instead of ↑b in some
          implementations (see above).

  Q !     sets the value  of the atom  ↑q to t,  enabling input from  the source
          selected by the  value of infile, or  selected by use of  the function
          uread.  In the PDP-10  Newio implementation, this is not  an interrupt
          character; instead it is a  macro character, and takes effect  only if
          processed by read.  (setq ↑q t).

  R !     sets  the  value  of  the  atom  ↑r  to  t,  enabling  output  to  the
          destinations selected by the value of outfiles, or selected by  use of
          the uwrite function.  (setq ↑r t).

  S !     turns off typeout until input  is read.  This is used to  suppress the
          rest  of  the  typeout from  the  current  request,  without affecting
          typeout from the next request that is typed in.  It is  implemented by
          setting ↑w to  t, then putting a  macro character in the  input stream
          which sets ↑w to nil and does a (terpri) when it is read.



December 6, 1977                     ∪3-1.3                            Page 3-11
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


  T !     sets  the  value  of the  atom  ↑r  to nil,  disabling  output  to the
          destinations that CTRL/r enables.  (setq ↑r nil).

  U       causes the current  call to read to  be restarted from  the beginning.
          (Not available in PDP-10 implementations).

  V !     sets the value of the atom ↑w to nil, enabling output to the terminal.
          (setq ↑w nil).

  W !     sets the value of the atom ↑w to t, disabling output to  the terminal.
          (setq ↑w t)  (and possible also (clear-output tyo)).

  X       causes an error which can be caught by errset.  This is a less drastic
          "quit"  than CTRL/g.   If it  is typed  within a  break loop,  it will
          return  no  further than  the  break loop,  since  break  uses errset.
          (error 'quit).

  Z !*    On ITS returns to ITS command level, i.e. DDT.  On Multics  returns to
          Multics command level. (start re-enters LISP.) On TOPS-10 goes  to DDT
          if a DDT has been loaded with LISP.

   The following control characters only exist in the Multics implementation.

  . !*    does nothing, and is used merely to speed up a slow process by causing
          an interaction.

  ? !*    asks the LISP subsystem what it is doing: running, waiting  for input,
          collecting garbage, or running with tty-interrupts masked off.

   The following  control characters only  exist in PDP-10  implementations with
the "moby I/O" capability.  (For now, this means only at the MIT AI Laboratory.)
(see (sstatus ttyscan), page 3-82).

  F       Cause graphics display slave to seize a display.

  N       Turn on display for character output.

  O       Turn off display for character output.

  Y       Cause display slave to release display.

   The  following control  characters only  work in  the  PDP-10 implementation.


Page 3-12                            ∪3-1.3                     December 6, 1977
**DRAFT**                          The System                          **DRAFT**


They are not  really interrupts, but occur  only when processed by  the terminal
input prescanner (see (sstatus ttyscan), page 3-82).

  K       redisplay the current input.  This  allows you to get a clean  copy of
          your input after rubouts have been used.  (On a display  console, LISP
          will attempt to make rubbed-out characters actually disappear from the
          screen.  ↑k is still useful if a timing error disrupts this process.)

  L       erases the screen if the terminal is a display, then does a CTRL/k.


                          Control-Character Functions


↑g                  SUBR 0 args

     Produces a quit to top level just as if a CTRL/g had been typed.


   These   functions  exist   only  in   the  PDP-10   Oldio  and   the  Multics
implementations.  They are  being phased out, so  using them in new  programs is
not recommended.


ioc                 FSUBR

     The argument to ioc is processed  as if it were a "control  character" that
     had been typed  in.  Numbers are taken  as a whole, atomic  symbols' pnames
     are  processed  character  by  character,  except  that  nil   is  ignored.
     Examples:

          (ioc 1)    causes user interrupt 1.
          (ioc vt)   switches output to the terminal.
          (ioc q)    switches input to a file.
          (ioc g)    quits back to the top level of LISP.

     If ioc returns, its value is t.







December 6, 1977                     ∪3-1.3                            Page 3-13
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


iog                 FSUBR

     iog first saves  the values of  the I/O switches ↑q,  ↑r, and ↑w.   Then it
     processes its first argument the same as ioc.  Next the remaining arguments
     to iog are evaluated, from left to right.  The values of the  variables ↑q,
     ↑r, and ↑w are  restored, and the value  of the last argument  is returned.
     Example:


               (iog vt (princ "A Message."))

     gets a message to the console  no matter what the I/O system is  doing.  It
     evaluates to "A Message."

               (iog a x1 x2 ... xn)

     can also be written

               ((lambda (↑q ↑r ↑w)
                        (ioc a)
                        x1
                        x2
                        ...
                        xn)
                nil nil nil)



















Page 3-14                            ∪3-1.3                     December 6, 1977
**DRAFT**                          The System                          **DRAFT**


1.4  Exceptional Condition Handling



1.4.1  The LISP Error System


   The  errors  detected  by  the  LISP  system  are  divided  into  two  types:
correctable and uncorrectable.  The uncorrectable errors will be explained first
since they are simpler.

   An uncorrectable  error is an  error that causes  the evaluation in  which it
occurs to be aborted.  When an uncorrectable error occurs, the first  thing that
happens is the printing of an  error message.  In Oldio, the error  message goes
to the terminal and nowhere else (unless suppressed, see errset, page  3-21), no
matter  how the  I/O switches  and variables  are set.   In Newio,  the variable
msgfiles is a list of the  files to which error messages should be  routed; this
initially is just the terminal.  The error message consists of  some explanatory
text and (usually) the object or form that caused the error.

   After the  error message has  been printed, control  is returned to  the most
recent  error-catcher.   There is  an  error-catcher at  top  level,  and error-
catchers are set up by the function errset (and break, which uses  errset).  All
variable  bindings  between the  error-catcher  and the  point  where  the error
occurred are restored.  Thus all  variables are restored to the values  they had
at top level or at the time  the errset was done, unless they were  setq'ed free
(without being bound).

   What happens next depends on how the error-catcher was set up.  At top level,
the  forms on  the errlist  are  evaluated and  the top  level loop  (or  a user
specified top level form) is re-entered.  (The symbol errlist is evaluated prior
to the above restoration of bindings, and saved in the variable /.  In  this way
the  errlist used  is the  one current  at the  time of  the error,  despite the
restoration of bindings.) If an error returns to break, it simply  re-enters its
read-eval-print loop.   In the  Multics implementation the  fact that  break has
caught an error is signalled by  ringing the bell on the terminal.  If  an error
returns to  errset, errset  returns nil  and evaluation  proceeds.  If  an error
returns to top level, the state of the world is reset and * is typed.

   The  above  description is  slightly  simplified.  The  user  can  request an
interrupt  to occur  between the  typing of  the message,  and the  unwinding of
bindings and return of  control to an error-catcher.   If the error is  going to


December 6, 1977                     ∪3-1.4                            Page 3-15
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


return to  top level,  the *rset-trap  user interrupt  is signalled.   This user
interrupt is  initially a system-supplied  break loop which  allows the  user to
examine the  values of variables  before the bindings  are restored, in  hope of
finding the cause of the error.  In (*rset t) mode a break loop is  entered, but
in (*rset nil)  mode the user  interrupt is ignored.  If  the error is  going to
return to a break or an errset, and *rset is non-nil, the errset  user interrupt
is  signalled.   The  initial  environment  contains  a  null  handler  for this
interrupt, but the user may supply a break loop or other handler.

   Correctable errors are  errors which may  be corrected by  user intervention.
If such an error is properly  corrected, evaluation will proceed as if  no error
had occurred.  If the option to correct the error is not exercised, this type of
error will be handled as if it were an uncorrectable error.

   When a correctable error occurs, a user interrupt is signalled.  See  page 3-
19  for  user  interrupt  channel assignments  for  these  errors.   The initial
environment  contains handlers  for these  errors which  print an  error message
similar to the message printed for an uncorrectable error and then enter a break
loop.

   The  argument  passed  to  the  user  interrupt  handler  is  an S-expression
describing the  error.  See section  1.4.2 for details.   If the  user interrupt
handler  is nil,  or  if it  returns  a non-list,  the  error is  treated  as an
uncorrectable error.  But  if the handler returns  a list, the first  element of
that list is used to correct the error in a way which depends on  the particular
error which occurred.

   If the most recent error-catcher is not top-level, correctable errors will be
treated  as uncorrectable  errors unless  there is  a non-null  handler  for the
errset interrupt.  This is  to prevent confusing "multiply nested"  error breaks
unless the user indicates that he  is sophisticated by setting up a  handler for
the errset interrupt.   (The errset handler itself  is only invoked if  *rset is
non-nil, however.)

   See the functions error, err, and errset.


1.4.2  User Interrupts


   LISP provides a number of "user interrupts," which are a mechanism by which a
user  procedure  may  temporarily gain  control  when  an  exceptional condition


Page 3-16                           ∪3-1.4.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


happens.  The exceptional conditions that use the user interrupt  system include
certain control characters, the alarmclock timers, asynchronous  I/O conditions,
the  garbage  collector,  and  many  of the  errors  that  are  detected  by the
interpreter or by the system  functions.  Errors detected by user  functions can
use this mechanism also.

   The user interrupts are divided  up into several channels.  Each  channel has
associated  with  it  a  service function.   If  the  service  function  is nil,
interrupts on that channel will be ignored.  If the service function is not nil,
it  is a  function which  is called  with one  argument when  the user-interrupt
occurs.   (A  few interrupt  handlers  take  more than  one  argument.   See the
specific descriptions.) The nature of the argument depends on which  channel the
interrupt is on; usually it is an S-expression which can be used to localize the
cause of  the interrupt.   Some user interrupts  use the  value returned  by the
service function to decide what to do about the cause of the interrupt.

   Interrupts  can  be either  synchronous  (e.g. errors  and  garbage collector
interrupts) or asynchronous (control characters, alarmclock, etc.).   To prevent
timing errors, asynchronous interrupts  are always run in (nointerrupt  t) mode.
A handler for an asynchronous interrupt must explicitly do (nointerrupt  nil) to
permit other  asynchronous errors  to interrupt it.   (For example,  the system-
supplied ↑b handler does this  so that control character interrupts can  be used
within the ↑b break loop.)

   The service  functions for  most user interrupts  are kept  as the  values of
symbols  with mnemonic  names.  A  list of  these symbols  begins on  page 3-19.
There are also  user interrupts for  control characters.  The  service functions
for these are declared using (sstatus ttyint).  See page 3-80.

   The initial values  for the service functions  of the various  interrupts are
provided by the  system as break  loops for some  interrupts and nil  for others
(except for some control characters).

   There  are  some  special considerations  for  user  interrupts  signalled by
correctable  error  conditions.   The  argument to  the  service  function  is a
description of the error whose exact  form is described in the catalogue  at the
end of this section.  If the  service function returns nil (or any  other atom),
the normal error procedure occurs -- control returns to the most  recent errset,
or to top level if there was no errset.  If the service function returns a list,
the first element of the list  is used to attempt recovery from the  error.  The
exact  way that  it  is used  is described  in  the catalogue.   If  recovery is
successful  execution proceeds  from  the point  where the  error  occurred.  If
recovery is unsuccessful another error is signalled.

December 6, 1977                    ∪3-1.4.2                           Page 3-17
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   Here is an  example of a  user interrupt service  function.  This is  the one
supplied  by the  system for  unbound  variable errors  when the  user  does not
specify one.  Note that the system-supplied error service functions consistently
bind  args to  the argument  supplied.  The  user can  check the  value  of this
variable to see what is wrong.  Note too that the system-supplied error handlers
restore readtable and obarray before breaking.

      (defun +internal-ubv-break (args)
             (declare (special args))
             (errprint nil msgfiles)         ;print error message
             ((lambda (readtable obarray)
                      (nointerrupt nil)
                      (break unbnd-vrbl))
              (get 'readtable 'array)
              (get 'obarray 'array)))
      (setq unbnd-vrbl '+internal-ubv-break)


alarmclock          SUBR 2 args

        alarmclock is a function for controlling timers.  It can start  and stop
     two separate  timers; one  is a  real-time timer  (which counts  seconds of
     elapsed time) and the other is a cpu-time timer (which  counts microseconds
     of machine  run time).   The first argument  to alarmclock  indicates which
     timer is being referred to:  it may be the atom time to indicate  the real-
     time timer or the atom runtime to indicate the cpu-time timer.

        The second argument to alarmclock controls what is done to  the selected
     timer.  If  it is  a non-negative number  (fixnum or  flonum) the  timer is
     started.  Thus if n is a positive fixnum or flonum,  evaluating (alarmclock
     'time n) sets the real-time timer  to go off in n seconds,  and (alarmclock
     'runtime n) sets the  cpu-time timer to go  off in n microseconds.   If the
     timer was already running the old setting is lost.  Thus at any  given time
     each timer can only  be running for one alarm,  but the two timers  can run
     simultaneously.

        If the second argument to alarmclock is not a positive number, the timer
     is shut off,  so (alarmclock x  nil) or (alarmclock x  -1) shuts off  the x
     timer.

        alarmclock returns t if it starts a timer, nil if it shuts it off.



Page 3-18                           ∪3-1.4.2                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


        When  a  timer goes  off,  the alarmclock  user  interrupt  occurs.  The
     service function  is run in  (nointerrupt t)  mode so that  it will  not be
     interrupted  while it  is performing  its service.   If it  wants  to allow
     interrupts, other timers, etc.  it can evaluate (nointerrupt nil).   In any
     case the status of the  nointerrupt flag will be restored when  the service
     function  returns.   The  argument passed  to  the  user  interrupt service
     function is  the atom time  or the atom  runtime, depending on  which timer
     went off.  See also the function nointerrupt.


nointerrupt         SUBR 1 arg

        (nointerrupt  t) shuts  off LISP  interrupts.  This  prevents alarmclock
     timers from going  off and prevents the  use of control characters  such as
     CTRL/g and CTRL/b.   Any of these interrupts  that occur are  simply saved.
     (nointerrupt t) mode is used  to protect critical code in  large subsystems
     written in LISP.  A similar  deferral technique is used by the  LISP system
     itself to protect against interrupts in the garbage collector.

        (nointerrupt 'tty) prevents  control characters (typed on  the terminal,
     or  "tty") from  causing  interrupts; however,  alarmclock  interrupts (and
     other asynchronous interrupts) are still allowed.  Any non-tty asynchronous
     interrupts which were saved will now go off.

        (nointerrupt nil) turns interrupts  back on.  Any interrupts  which were
     saved will now get processed.  This is the normal, initial state.

        The result returned from nointerrupt is the previous interrupt  status -
     nil, t, or tty.

        Example:

           ((lambda (oldstatus)
                    <protected code>
                    (nointerrupt oldstatus))
            (nointerrupt t))



1.4.3  Catalogue of User Interrupt Channels




December 6, 1977                    ∪3-1.4.2                           Page 3-19
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   Each  user  interrupt  channel  (except some  associated  with  files)  has a
variable  whose  value is  a  functional  form, the  service  function  for that
channel.  The  name of  the interrupt  channel is the  same as  the name  of the
variable.   The  following lists  the  user interrupt  channels  in alphabetical
order.  The  argument to  which the service  function is  applied and  the value
which  it  should  return  are described.   By  convention,  almost  all service
functions receive  one argument.  Some  user interrupts are  initially set  to a
system-supplied  handler which  binds  the variable  args to  this  argument and
enters a break loop.  The name of the interrupt is used as the break identifer.

   Some user interrupts ignore the value returned by the service function, while
others distinguish two cases.  If the value is atomic, the service  function was
not able  to recover from  the condition that  caused the interrupt.   LISP will
take its default  action, such as returning  control to the most  recent errset.
If  the value  is a  list, the  car of  that list  is used  to recover  from the
condition that caused the  interrupt.  It is usually a  new piece of data  to be
used in place of the  one that was being complained  about, or a new form  to be
evaluated in place of the form that erred.

   If the value of the service-function variable is nil instead of  a functional
form, the user interrupt is considered to be turned off.  The system  behaves as
if the function had run and returned nil.

   Some  user  interrupts  are  asynchronous  in  nature,  and  are  executed in
(nointerrupt t) mode to prevent timing errors.  The interrupt handler may choose
to run in (nointerrupt nil) mode, however, as the initial ↑b handler  does.  The
nointerrupt mode  is restored  after the  handler is  run.  Such  interrupts are
themselves deferred by (nointerrupt t) mode.

alarmclock          VARIABLE

The value of alarmclock is the service function for the user interrupt signalled
when a timer set  up by the alarmclock function  goes off.  The argument  is the
name  of the  timer which  went off,  time or  runtime.  The  returned  value is
ignored.   The  service function  is  executed in  (nointerrupt  t)  mode.  This
interrupt is initially turned off.


autoload            VARIABLE

The  value of  autoload is  the service  function for  the user  interrupt which
provides  automatic  loading  of program  packages  into  the  environment.  The


Page 3-20                           ∪3-1.4.3                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


argument is (function-name . autoload-property).  The returned value is ignored.
See page 3-26 for details.  This interrupt is initially set to a  function which
simply loads a file.


cli-message         VARIABLE

The value of cli-message service  handler for the user interrupt  signalled when
another job has interrupted  the LISP job via  the CLI device.  The  argument is
nil and the  returned value is  ignored.  A user handler  is expected to  open a
file on the  CLA device and  read the message from  the other job.   The service
function is  run in (nointerrupt  t) mode.  This  interrupt is  initially turned
off.  Currently, it exists only in the ITS Newio implementation.


errset              VARIABLE

The value  of errset is  the service  function for the  user interrupt  which is
signalled  when an  error is  caught by  an errset  and *rset  is  non-nil.  The
argument  is nil  and the  returned value  is ignored.   This user  interrupt is
initially off.  Turning it on affects the behavior of the error system (see page
3-16).


fail-act            VARIABLE

The value of fail-act  is the service function  for the user interrupt  which is
signalled when any of a large variety of miscellaneous error  conditions occurs.
The argument is a list whose first element is generally a symbol which describes
the type  of error  condition.  The rest  of the  list contains  various objects
related to the error.  The returned  value depends on the error.  These  are not
standardized and will not be described here.  This interrupt is initially set to
a break loop.


gc-daemon           VARIABLE

The value of gc-daemon is the  service function for the user interrupt  which is
signalled after each  garbage collection.  The argument  is a list of  items; in
the  PDP-10  implementation  each item  is of  the form  (space-name free-before
free-after size-before size-after) and in the Multics implementation,  each item
is of  the form (space-name  free-before . free-after).   The returned  value is
ignored.  This interrupt is initially turned off.

December 6, 1977                    ∪3-1.4.3                           Page 3-21
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


gc-lossage          VARIABLE

The value of gc-lossage is the service function for the user interrupt  which is
signalled when there is no more available address space or when the Time Sharing
Monitor rejects a request for more memory.  In the Multics implementation, there
is always  enough memory, so  this user interrupt  never occurs.  In  the PDP-10
implementation the argument is the name of the space that lost, and the returned
value is ignored.  This interrupt is initially set to a break loop.


gc-overflow         VARIABLE

The value of gc-overflow is the service function for the user interrupt which is
signalled when a space overflows its gcmax.  (see alloc and (status gcmax).) The
argument  is  the name  of  the space.   The  returned value  is  ignored.  This
interrupt is initially set to a break loop.


io-lossage          VARIABLE

The value of io-lossage is the service function for the user interrupt  which is
signalled when the I/O system encounters an error (for example, a file which was
being opened was not found).  The argument is a list of the name of the function
which erred  and its arguments,  which may have  been standardized  or otherwise
partially digested.  The returned value is a list of a new form to  be evaluated
in place of the call to  the function which erred.  This interrupt  is initially
set to a break loop.


machine-error       VARIABLE

The  value  of machine-error  is  the  service handler  for  the  user interrupt
signalled when some difficulty is experienced by the host machine.   The service
function receives four arguments instead of one.  The first is an  atomic symbol
indicating the type of error:

   eval             illegal machine operation
   examine          attempt to reference non-existent memory
   deposit          attempt to write into read-only memory
   oddp             parity error
The other three arguments are  fixnums which are addresses of  memory locations.
The second is the location of  the error; the third is the program  counter when


Page 3-22                           ∪3-1.4.3                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


the error occurred;  and the fourth  is the JPC (the  program counter as  of the
last jump instruction before the error occurred).  The machine-error handler may
signal a different kind of error or a ↑g quit (see the ↑g function)  if desired,
or enter  a break loop.   The subr function  (see page 3-103)  may be  useful in
decoding  the three  fixnum arguments.   If the  handler returns,  the  value is
ignored  and  the erroneous  operation  is  retried.  If  the  user  provides no
machine-error  handler (the  interrupt is  initially turned  off), the  error is
handled in the default manner for the host machine.  On ITS, this puts  the user
in DDT.  This currently exists only in the PDP-10 Newio implementation.
Example:

   (defun machine-error-handler (type loc pc jpc)
          ((lambda (args terpri)
                   (declare (special args terpri))
                   (terpri msgfiles)
                   (cond ((eq type 'examine)
                          (princ '|;REFERENCE TO NON-EXISTENT
                                 MEMORY| msgfiles))
                         ((eq type 'deposit)
                          (princ '|;WRITE INTO READ-ONLY MEMORY|
                                 msgfiles))
                         ((eq type 'eval)
                          (princ '|;ILLEGAL MACHINE OPERATION|
                                 msgfiles))
                         ((eq type 'oddp)
                          (princ '|;MEMORY PARITY ERROR|
                                 msgfiles))
                         (t (princ '|;UNKNOWN MACHINE ERROR|
                                   msgfiles)))
                   (princ '| FROM LOCATION | msgfiles)
                   (princ pc)
                   (princ '| IN FUNCTION | msgfiles)
                   (prin1 (subr pc))
                   (break machine-error))
           (list type loc pc jpc)
           t))


mar-break           VARIABLE

The value of mar-break is  the service handler for the user  interrupt signalled
when the  memory location specified  by (sstatus mar)  (see page 3-97)  has been


December 6, 1977                    ∪3-1.4.3                           Page 3-23
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


accessed in the specified manner.  The argument is nil and the returned value is
ignored.   The service  function is  run in  (nointerrupt t)  mode.   Also, LISP
implicitly performs (sstatus mar 0 nil) before running the user  interrupt; this
helps to prevent  infinite loops.  This interrupt  is initially turned  off.  It
currently exists only  in the ITS Newio  implementation. See page 3-56  for more
information on using this interrupt.


pdl-overflow        VARIABLE

The value of pdl-overflow is  the service function for the user  interrupt which
is signalled when  a pushdown list exceeds  its pdlmax.  (see alloc  and (status
pdlmax).) The  argument is  the spacename  of the  pushdown list.   The returned
value is ignored.  This interrupt is initially set to a break loop.


sys-death           VARIABLE

The value of sys-death is  the service handler for the user  interrupt signalled
when the time-sharing  system is about  to go down,  has been revived  from that
state, or  is being debugged.   The argument  is nil and  the returned  value is
ignored.  A  user handler  may wish  to examine  the result  of (status  its) to
determine the state of the system.  The service function is run  in (nointerrupt
t) mode.  This interrupt is initially turned off.  This currently exists only in
the ITS Newio implementation.


tty-return          VARIABLE

The value of tty-return is the service handler for the user  interrupt signalled
when control of the terminal is returned to the LISP job by its  superior.  This
allows LISP to determine that the display screen may have been changed  by other
jobs.   The argument  is nil  and the  returned value  is ignored.   The service
function is  run in (nointerrupt  t) mode.  This  interrupt is  initially turned
off.  This currently exists only in the ITS Newio implementation.


unbnd-vrbl          VARIABLE

The value of unbnd-vrbl is the service function for the user interrupt  which is
signalled when an attempt  is made to evaluate  an atomic symbol which  does not
have a value (an unbound variable.)  The argument is a list of the  symbol which


Page 3-24                           ∪3-1.4.3                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


could not  be evaluated.  The  returned value is  a list of  a new symbol  to be
evaluated in its place.  This interrupt is initially set to a break loop.


undf-fnctn          VARIABLE

The value of undf-fnctn is the service function for the user interrupt  which is
signalled when an attempt is made to apply an undefined function.   The argument
is a list of the functional form which could not be applied.  The returned value
is  a list  of a  new  functional form  to take  its place.   This  interrupt is
initially set to a break loop.


unseen-go-tag       VARIABLE

The value of unseen-go-tag is the service function for the user  interrupt which
is signalled when  go or throw is  used with a tag  which does not exist  in the
current prog body or in any catch, respectively.  The argument is a list  of the
erroneous tag.  The returned value is a  list of a new tag to replace  it.  This
interrupt is initially set to a break loop.


wrng-no-args        VARIABLE

The value of wrng-no-args is  the service function for the user  interrupt which
is signalled when a function is called with the wrong number of  arguments.  The
argument is a list of two items: First, a list of the function and the arguments
that were passed.  Second, the  lambda-list if the function was  interpreted, or
the same dotted pair as args returns if the function was compiled, or the atom ?
if this information could not be determined.  The returned value is a list  of a
new  form  to be  evaluated  in place  of  the losing  one.   This  interrupt is
initially set to a break loop.


wrng-type-arg       VARIABLE

The value of wrng-type-arg is the service function for the user  interrupt which
is  signalled when  an argument  is passed  to a  system function  which  is not
acceptable to that function.  The argument  is a list of the argument  which was
not accepted.  The  returned value is  a list of a  new argument to  replace it.
That is, directly an  argument, not a form to  be evaluated to get  an argument.
This interrupt is initially set to a break loop.


December 6, 1977                    ∪3-1.4.3                           Page 3-25
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


*rset-trap          VARIABLE

The value of *rset-trap is the service function for the user interrupt  which is
signalled when an error returns  control to top level, just before  the bindings
are  restored.  By  convention, the  handler for  this interrupt  should  not do
anything unless the variable  *rset is non-nil.  This  is so that the  user will
not be bothered unless he has  put LISP in debugging mode.  The argument  is nil
and  the  returned value  is  ignored.  This  interrupt  is initially  set  to a
function which enters a break loop if *rset is non-nil.

   There are  other interrupt handlers  which are associated  with I/O  files or
inferior jobs.  See eoffn, endpagefn, (sstatus ttyint), and create-job.


1.4.4  Autoload


   The autoload feature provides the  ability for a function not present  in the
environment to  be automatically  loaded in  from a  file the  first time  it is
called.  When  eval, apply, funcall,  or the version  of apply used  by compiled
LISP searches the  property list of an  atom looking for a  functional property,
and fails to find one, it looks for a property under the indicator autoload, and
it it finds one, automatic loading will occur.

   Automatic loading is performed by means of the autoload user  interrupt; thus
the user may assert  any desired degree of  control over it.  When  the autoload
property is encountered, the user interrupt handler is called with one argument,
which is  a dotted pair  whose car is  the atomic symbol  which is  the function
being autoload'ed,  and whose cdr  is the value  of the autoload  property.  The
system-supplied handler for this user interrupt could have been defined by:

      (setq autoload
            (function (lambda (x) (load (cdr x)) )))

From this one can see that the value of the autoload property should be the name
of the file which contains the definition of the function.  Note:  in  the TOPS-
10 implementations  the system  autoload handler  presently uses  fasload rather
than load because  the load function requires  the Newio feature.   This affects
the form of an autoload property.

   When the interrupt handler returns, it should have put a  functional property
on the property  list of the function  being autoloaded.  If not,  an undf-fnctn
error will occur with a message such as "function undefined after autoload."

Page 3-26                           ∪3-1.4.3                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


   Examples of setting up functions to be autoloaded:

   In the Multics implementation:

      (putprop 'foo ">udd>AutoProg>Library>foo-function" 'autoload)

   In the PDP-10 Oldio implementation:

      (putprop 'foo '(foo fasl dsk me) 'autoload)

   In the PDP-10 Newio implementation:

      (putprop 'foo '((dsk me) foo) 'autoload)

   or  (putprop 'foo  '|dsk:me;foo fasl|  'autoload) or  the Oldio  version also
works.




























December 6, 1977                    ∪3-1.4.4                           Page 3-27
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


1.5  Debugging



1.5.1  Binding, Pdl Pointers, and the Evaluator


   The Maclisp evaluator  is based on  a push down  list (pdl), or  stack, which
holds  bindings,  evaluation frames,  and  sundry internal  data.   Bindings are
values of atomic  symbols which are  saved when the  symbols are used  as lambda
variables, prog variables, or  do variables.  Evaluation frames  are constructed
when a non-atomic form is evaluated  or when apply is used.  They  correspond to
function calls.

   As the evaluator recursively evaluates a form, information is pushed onto the
pdl and later popped off.  When the *rset and nouuo flags are t this information
is sufficiently detailed  to be of use  in debugging.  (See the  variables *rset
and nouuo in the next section.)

   A position within the pdl may be named by means of a "pdl pointer",  which is
a  negative  fixnum whose  value  has meaning  to  the evaluator.   nil  is also
accepted as a pdl pointer; it means  the top of the stack, i.e. the  most recent
evaluation.  Note that this is different from nil as a binding  context pointer,
which means  the bottom of  the stack  or the outermost  evaluation.  0  is also
accepted as a pdl pointer; it  designates the frame at the bottom of  the stack.
Pdl pointers may be used  as arguments to several debugging  functions described
in the next section.  Since the fixnum value of a pdl pointer has  only internal
meaning, generally a pdl pointer  cannot be obtained from user input,  except by
the user typing in a pdl pointer chosen from a list of pdl pointers typed out at
him.  The "frame" functions described in the next section may be used  to obtain
pdl pointers.

   An  important thing  to note  about pdl  pointers is  their limited  scope of
validity.  If the  information on the  pdl which is named  by a pdl  pointer has
been popped off since the pdl pointer was created, the pdl pointer no longer has
valid meaning.


1.5.2  Functions for Debugging





Page 3-28                            ∪3-1.5                     December 6, 1977
**DRAFT**                          The System                          **DRAFT**


*rset               SUBR 1 arg

        (*rset x) sets the *rset switch to nil if x is nil, or to t if x is non-
     nil,  and returns  the value  it set  it to.   (See below).   This function
     exists primarily for user typing convenience.


*rset               SWITCH

        If  the  *rset switch  is  non-nil,  extra information  is  kept  by the
     interpreter  to  allow  the  debugging  functions,  such  as  baktrace  and
     evalframe, to work.   In addition, the  interpreter will make  extra checks
     such as  checking the number  of arguments  passed to a  subr or  lsubr and
     checking that array subscripts lie within the declared  bounds.  Generally,
     the *rset switch being on means  "I am debugging"; this is known  as "*rset
     mode".  The initial state of the switch is nil.


nouuo               SUBR 1 arg

        (nouuo t) sets the nouuo switch.

        (nouuo nil) turns off the nouuo switch. (This is the initial state.)

        nouuo returns t or nil  according to whether it turned the  nouuo switch
     on or off.   (See below.)  This function  exists primarily for  user typing
     convenience.


nouuo               SWITCH

        If the nouuo switch is on, function calls made by compiled  functions to
     compiled  functions  or  system  functions are  forced  to  go  through the
     interpreter each  time.  This aids  in debugging.  If  the nouuo  switch is
     off, which is the normal case,  compiled calls can be made to  go directly,
     which is much faster.

        The nouuo switch may be turned off at any time.  Each  compiled function
     call will only go through the interpreter once more, at which time  it will
     be linked directly.  If the compiled code has been reloaded into the system
     with the PURE option (see page 3-74) then this direct link may be unsnapped
     and the  Interpreter route re-established  by (sstatus  uuolinks).  Because


December 6, 1977                    ∪3-1.5.2                           Page 3-29
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     the PURE  option requires  an amount  of extra  space and  time, it  is not
     normally  on; thus  links  snapped in  code  loaded as  non-PURE  cannot be
     unlinked.

        The trace  package turns this  switch on when  a function is  traced, in
     order  to  ensure  that  tracing will  work  even  for  compiled functions.
     Compiled function  calls which have  been "snapped" to  go directly  do not
     push debugging information  in *rset mode and  cannot be traced.   See also
     (status uuolinks).


baktrace            LSUBR 0 to 2 args

        baktrace  displays  the  stack  of  pending  function  calls.   It gives
     detailed information only in (*rset  t) mode.  The first argument is  a pdl
     pointer, as with evalframe.  If it is omitted, nil is assumed,  which means
     start from the top of the  pdl.  The second argument is the  maximum number
     of lines to be typed; if it is omitted the entire stack is displayed.  (The
     second argument is currently permitted only in the Multics implementation.)
     The information printed by baktrace  is not the same as that  obtained with
     evalframe;  both should  be used  to get  the maximum  amount  of debugging
     information.


baklist             LSUBR 0 to 1 arg

        baklist returns a list  containing the information which  baktrace would
     print.  (This is available only on the PDP-10 implementations.)


errframe            SUBR 1 arg

        errframe returns a  list describing an error  which has been  stacked up
     because of  a user interrupt.   The list has  the form (err  pdlptr message
     bcp), where pdlptr is a number  which describes the location in the  pdl of
     the error, message is  a list of from one  to three things which,  given to
     the error  function, could have  caused this same  error, and  bcp (binding
     context pointer) is a number which can be used as a second argument to eval
     or a  third argument  to apply to  cause evaluation  using the  bindings in
     effect just before the error occurred.

        The argument to errframe  can be nil, which  means to find the  error at


Page 3-30                           ∪3-1.5.2                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


     the top of  the stack; i.e. the  most recent error.  It  can also be  a pdl
     pointer, in which  case the stack is  searched downward from  the indicated
     position.  Thus  if there  are recursive  calls to  the error  handler, the
     second error back down the stack may be found by:

                        (errframe (cadr (errframe nil)))

        The argument  to errframe may  also be a  positive number, which  is the
     negative of a pdl pointer.  This means start from the position in the stack
     marked by the pdl pointer and search upwards.

        If no error is found, errframe returns nil.


errprint            LSUBR 1 to 2 args

        errprint treats its argument the same as errframe.  The second argument,
     in  Newio implementation  only,  is the  file(s)  into which  to  print the
     information  (see  print).   The  message portion  of  the  error  frame is
     princ'ed.  errprint  returns t if  a message  was typed out  and nil  if no
     error frame was found.


evalframe           SUBR 1 arg

        The argument to evalframe is  a pdl pointer, as with errframe.   The pdl
     is searched  for an  evaluation of a  function call,  using the  same rules
     about  starting point  and direction  as errframe  uses.   evalframe always
     skips over any calls to itself that it finds in the pdl.

        The value is a list (type pdlptr form bcp), where type is eval or apply,
     pdlptr is a pdl pointer to the evaluation in the stack, suitable for use as
     an argument to  evalframe or errframe or  baktrace, form is the  form being
     evaluated or  a list  of the  name of  the function  being applied  and the
     arguments it was applied to, and bcp is a binding context pointer which can
     be used with eval to evaluate something in the binding context  just before
     the evaluation found by evalframe.

        evalframe returns nil if no evaluation can be found.

        evalframe only works in (*rset t) mode, since no extra frame information
     is saved otherwise.


December 6, 1977                    ∪3-1.5.2                           Page 3-31
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


freturn             SUBR 2 args

        (freturn p x)  returns control to the  evaluation designated by  the pdl
     pointer p, and forces it  to return x.  This "non-local-goto"  function can
     be used to do fancy recovery from errors.


evalhook            VARIABLE

        If the value of evalhook is non-null, then special things happen  in the
     evaluator.  When  a form  (even an atom)  is to  be evaluated,  evalhook is
     bound to nil and the functional form which was its value is applied  to one
     argument - the form that was trying to be evaluated.  The value  it returns
     is then returned from the  evaluator.  This feature is used by  the Stepper
     package described later in this section.

        evalhook is bound to nil by break, and setq'ed to nil by errors  that go
     back to top level  and print *.  This  provides the ability to  escape from
     this mode if something bad happens.

        In order not to impair  the efficiency of the LISP  interpreter, several
     restrictions  are imposed  on evalhook.   It only  applies to  evaluation -
     whether in  a read-eval-print loop,  internally in evaluating  arguments in
     forms, or  by explicit  use of  the function  eval.  It  does not  have any
     effect on compiled function references, on use of the function apply, or on
     the  "mapping" functions.   Also, as  a special  case, the  array reference
     which  is  the  first argument  to  store  is never  seen  by  the evalhook
     function; however, the subexpressions of the array reference  (the indices)
     will be seen.  (This special treatment avoids a problem with the  way store
     works.) Normally  the evaluator does  not check the  value of  evalhook, in
     order to  save time.  To  make it check,  you must both  be in (*rset  t) -
     debugging  -  mode,   and  have  done   (sstatus  evalhook  t).    Not  all
     implementations need  both of those,  but you should  always do both  to be
     sure.  If you  use the Stepper  package, you need  not worry; it  does this
     automatically.


evalhook            LSUBR 2 or 3 args

        (evalhook  form hook)  is a  function which  helps exploit  the evalhook
     feature.   The  form  is  evaluated  with  evalhook  lambda-bound   to  the
     functional  form  hook.   The  checking  of  evalhook  is  bypassed  in the


Page 3-32                           ∪3-1.5.2                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


     evaluation  of form  itself,  but not  in any  subsidiary  evaluations, for
     instance  of  arguments  in  the form.   This  is  like  a "one-instruction
     proceed"  in  a  machine-language debugger.   If  all  three  arguments are
     present, the second is a binding context pointer and is used as  the second
     argument to eval, and the third argument is the hook.

        Example:


           (defun hook fexpr (x)           ;called as (hook <form>)
                  ((lambda (*rset)
                           (prog2 (sstatus evalhook t)
                                           ;magic sstatus
                                  (evalhook (car x) 'hook-function)
                                           ;evaluate form
                                  (sstatus evalhook nil)))
                                           ;more magic
                   t))

           (defun hook-function (f)
                  ((lambda (terpri)
                           (declare (special terpri))
                           (princ '|form: | msgfiles)
                           (prin1 f msgfiles)
                           ((lambda (v)
                                    (terpri msgfiles)
                                    (princ '|value: | msgfiles)
                                    (prin1 v msgfiles)
                                    v)
                            (evalhook f 'hook-function)))
                   t))
                                           ;this is how to eval the
                                           ; form so as to hook
                                           ; sub-forms


        The following output might be seen from (hook (cons (car '(a . b)) 'c):

                form: (cons (car (quote (a . b))) (quote c))
                form: (car (quote (a . b)))
                form: (quote (a . b))
                value: (a . b)
                value: a

December 6, 1977                    ∪3-1.5.2                           Page 3-33
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


                form: (quote c)
                value: c
                value: (a . c)
                (a . c)


The following functions only exist in the Multics implementation.

baktrace1           LSUBR 0 to 2 args

        baktrace1 is the same  as baktrace except that binding  context pointers
     suitable for use with eval and apply are displayed along with  the function
     names.


baktrace2           LSUBR 0 to 2 args

        baktrace2 is the  same as baktrace1  except that pdl  pointers, suitable
     for use with baktrace and evalframe, are displayed along with  the function
     names and binding context pointers.
























Page 3-34                           ∪3-1.5.2                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


1.5.3  The Trace Package


   The LISP trace package provides the ability to perform various actions at the
time a  function is called  and at the  time it returns.   This can be  used for
traditional tracing or for more sophisticated debugging actions.

   The trace  package is  not part of  the initial  environment; however,  it is
automatically  loaded in  on the  first reference  to the  function  trace. (See
autoload.)

   The LISP trace package consists of three main functions, trace,  untrace, and
remtrace, all of which are fexprs.

   A call to trace has the following form:
                              (trace trace←specs)

A trace←spec in turn is either an  atom (the name of the function to  be traced)
or a list:
                      (function-name option1 option2 ...)

where the options are as follows:

break pred      causes  a break  after  printing the  entry trace  (if  any) but
                before applying  the traced  function to  its arguments,  if and
                only if pred evaluates to non-nil.

cond pred       causes trace information to be printed for function entry and/or
                exit if and only if pred evaluates to non-nil.

wherein fn      causes  the function  to  be traced  only when  called  from the
                specified function fn.  The user can give several trace specs to
                trace,  all  specifying  the same  function  but  with different
                wherein options,  so that  the function  is traced  in different
                ways when called from  different functions.  Note that if  fn is
                already being  traced itself, the  wherein option  probably will
                not work as desired.  (Then again, it might.)  Note that  fn may
                not be a compiled function.

argpdl pdl      specifies an atom pdl  whose value trace initially sets  to nil.
                A  list of  the current  recursion level  for the  function, the
                function's name, and a list of the arguments is cons'ed onto the


December 6, 1977                    ∪3-1.5.3                           Page 3-35
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


                pdl when the function is  entered, and cdr'ed back off  when the
                function is exited.  The pdl can be inspected from a breakpoint,
                for example, and  used to determine  the very recent  history of
                the function.  This option  can be used with or  without printed
                trace output.  Each  function can be given  its own pdl,  or one
                pdl may serve several functions.

entry list      specifies a list of arbitrary S-expressions whose values  are to
                be  printed  along  with the  usual  entry-trace.   The  list of
                resultant values, when printed, is preceded by a \\  to separate
                it from the other information.

exit list       similar  to entry,  but specifies  expressions whose  values are
                printed with the exit-trace.  Again, the list of  values printed
                is preceded by \\.

arg                     that the function's arguments, resultant value, both, or
                specify
value                    are to  be traced.   If not  specified, the  default is
                neither
both                   Any "options" following one of these four are  assumed to
                both.
nil                arbitrary  S-expressions whose  values are  to be  printed on
                be
                both  entry  and  exit  to the  function.   However,  if  arg is
                specified, the values are  printed only on entry, and  if value,
                only on exit.  Note that since arg, value, both, and nil swallow
                all  following expressions  for this  purpose, whichever  one is
                used  should  be the  last  option specified.   Any  such values
                printed will  be preceded  by a  // and  will follow  any values
                specified by entry or exit options.

   If the variable arglist is used in any of the expressions given for the cond,
break, entry, or  exit options, or  after the arg,  value, both, or  nil option,
when those expressions are evaluated the value of arglist will effectively  be a
list of the arguments given to the traced function.  Thus

                    (trace (foo break (null (car arglist))))

would cause a break in foo if and only if the first argument to foo is nil.

   Similarly, the variable  fnvalue will effectively  be the resulting  value of
the traced  function.  For obvious  reasons, this should  only be used  with the
exit option.

   The trace specifications may be "factored."  For example,


Page 3-36                           ∪3-1.5.3                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


                     (trace ((foo bar) wherein baz value))

is equivalent to

            (trace (foo wherein baz value) (bar wherein baz value))

   All output printed by trace can be ground into an indented,  readable format,
by simply setting the variable  sprinter to t.  Setting sprinter to  nil changes
the output back  to use the  ordinary print function,  which is faster  and uses
less storage but is less readable for large list structures.


      Examples of the use of trace:

      (1) To trace function foo, printing both arguments on entry and  result on
   exit:
                                  (trace foo)

      or (trace (foo)) or (trace (foo both)).

      (2) To  trace function foo  only when called  from function bar,  and then
   only if (cdr x) is nil:
                 (trace (foo wherein bar cond (null (cdr x))))

      or (trace (foo cond (null (cdr x)) wherein bar))

   As this example shows, the  order of the options makes no  difference, except
   for arg, value, both, or nil, which must be last.

      (3) To trace function quux, printing the resultant value on exiting but no
   arguments on entry, printing  the value of (car  x) on entry, of  foo1, foo2,
   and (foo3 bar) on exit, and of zxcvbnm and (qwerty shrdlu) on both  entry and
   exit:

               (trace (quux entry ((car x)) exit (foo1 foo2 (foo3 bar))
                            both zxcvbnm (qwerty shrdlu)))

      (4)  To trace  function foo  only when  called by  functions bar  and baz,
   printing args on entry and result  on exit, printing the value of  (quux barf
   barph) on exit from foo  when called by baz only, and  conditionally breaking
   when called by bar if a equals b:



December 6, 1977                    ∪3-1.5.3                           Page 3-37
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



               (trace (foo wherein bar break (equal a b))
                      (foo wherein baz exit ((quux barf barph))))

      (5) To trace  functions phoo and fu,  never printing anything  for either,
   but saving all arguments for both on a common pdl called foopdl, and breaking
   inside phoo if x is nil:

               (trace (phoo argpdl foopdl break (null x) cond nil nil)
                      (fu argpdl foopdl cond nil nil))

      The "cond nil"  prevents anything at all  from being printed.   The second
   nil in each  trace spec specifies  that no args or  value are to  be printed;
   although the cond nil would prevent the printout anyway, specifying  this too
   prevents trace from even setting up the mechanisms to do this.

      trace returns as its  value a list of  names of all functions  traced; for
   any functions traced with the wherein option, say (trace (foo  wherein bar)),
   instead of  returning just  foo it returns  a 3-list  (foo wherein  bar).  If
   trace finds a trace spec it  doesn't like, instead of the function's  name it
   returns a list whose car is ?  and whose cdr is an error message.   The error
   messages are:

(? wherein foo)  trace couldn't find  an expr, fexpr, or macro property  for the
                function specified by the wherein option.

(? argpdl foo)  The  item following the argpdl  option was not a  non-nil atomic
                symbol.

(? foo  not function)  Indicates  that the function  specified to be  traced was
                non-atomic, or  had no  functional property.   (Valid functional
                properties are expr, fexpr, subr, fsubr, lsubr, and macro.)

(? foo)         foo is not a valid option.

   Thus a use of trace such as

                  (trace (foo wherein (nil)) (bar argpdl nil))

would return, without setting up any traces,

                       ((? wherein (nil)) (? argpdl nil))


Page 3-38                           ∪3-1.5.3                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


   If you  attempt to specify  to trace a  function already being  traced, trace
calls untrace before setting up the  new trace.  If an error occurs,  causing (?
something) to be returned, the function for which the error occurred may  or may
not have been untraced.  Beware!

   It is possible to call trace with no arguments.  (trace) evaluates to  a list
of all the functions currently being traced.

   untrace is used to undo the  effects of trace and restore functions  to their
normal, untraced state.  The argument to untrace for a given function  should be
what trace returned for  it; i.e. if trace  returned foo, use (untrace  foo); if
trace returned (foo wherein bar) use (untrace (foo wherein bar)).   untrace will
take multiple specifications, e.g. (untrace foo quux (bar wherein  baz) fuphoo).
Calling untrace  with no  arguments will untrace  all functions  currently being
traced.

   remtrace,  oddly enough,  expunges  the entire  trace package.   It  takes no
arguments.


























December 6, 1977                    ∪3-1.5.3                           Page 3-39
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


1.5.4  The Stepper


   The  LISP  "stepping" package  is  intended  to give  the  LISP  programmer a
facility analogous to  the Instruction Step mode  of running a  machine language
program.

                                The Rich Stepper

   The Rich stepper package  provides a simple, small debugging  capability.  It
is available as a library program in the ITS implementation.

↓_How to Use the STEP Facility_↓


   The package contains two compiled functions which are loaded by

                         (fasload step fasl dsk liblsp)

The user interface is through the function (fexpr) step, which sets  switches to
put the LISP interpreter in and out of "stepping" mode.  The basic commands are:


              (step t)        ;Turn on stepping mode.
              (step nil)      ;Turn off stepping mode.

These commands are usually typed at top level, and will take  effect immediately
(i.e. the next S-expression typed  in will be evaluated in stepping  mode). Also
↑g, in addition to returning to top level, turns off stepping mode.

   In stepping mode, the LISP  evaluator will print out each S-expression  to be
evaluated before  evaluation, and the  returned value after  evaluation, calling
itself recursively to display the stepped evaulation of each argument, if the S-
expression  is  a  function  call.   After  displaying  each  S-expression,  the
evaluator will wait before evaluation for a command character from the console:


     <space>             Continue stepping recursively.

     <rubout>            Show returned value from this level only,  and continue
                         stepping upward.



Page 3-40                           ∪3-1.5.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


     <cr> or <tab>       Turn  off  stepping  mode.  (but   continue  evaluation
                         without stepping).

     P                   Redisplay current form in full (i.e.  without prinlevel
                         or prinlength)

     B                   Get breakpoint; proceed with $P

     M                   See advanced features under Stepping Macro Expansions.

↓_More Advanced Features_↓


   Selectively turning on step:

                              (step foo1 foo2 ...)

If this command is typed  at top level, stepping will not  commence immediately,
but rather when the evaluator  first encounters a S-expression whose car  is one
of  foo1, foo2,  etc.   This form  will then  display  at the  console,  and the
evaluator will be in stepping mode waiting for a command character.

   Stepping Macro Expansions:

   If the stepper is proceeded with a <space>, it will not step the execution of
macro expansions, but will rather just show the result of the macro of expansion
and wait for another command.

   To see the execution of the macro expansion itself, proceed the  stepper with
an M instead of a <space>.

   Using step with breakpoints:

   The above description applies to turning stepping on and off globally  at top
level.  More detail is necessary to use step flexibly in and out  of breakpoints
(e.g. together with trace).

   If stepping is turned on by (step t) at top level, the evaluator will  NOT be
in  stepping  mode  within  a  breakpoint loop.   If  you  wish  to  use stepped
evaluation  within  a break  loop  you must  turn  it on  locally  by  (step t).
Conversely, if stepping was  not turned on at top  level and it is turned  on by
(step t) in a break loop, it will  NOT be on when return is made from  the break
loop by $P.

December 6, 1977                    ∪3-1.5.4                           Page 3-41
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   However, executing  (step nil)  inside a  break loop  will turn  off stepping
globally, i.e. within the break loop, and after return has be made by $P.

   The most useful feature is the following, however:

     (step)              Command at top level has no immediate effect.

   After (step) has been executed at top level, a subsequent (step t)  inside of
a break loop will  have the effect of turning  on stepping mode both  inside the
break loop and globally,  i.e. the evaluator will start  to step as soon  as the
return is made  from the break  loop by $P.  Thus,  for instance, one  could set
trace  to break  at  some special  place,  and then  use  the break  to  turn on
stepping.

   prinlevel and prinlength:

   In the present version, for convenience, prinlevel and prinlength are lambda-
bound  inside the  hooking function  to 3  and 5  respectively.  These  could be
changed by editing the expr code and recompiling.

   When the P command is used, prinlevel and prinlength are temporarily bound to
nil, and the toplevel printer (the value of atom prin1) is used to redisplay the
current form.

   Overhead of Stepping:

   If stepping mode has been turned off by ↑g, the execution overhead  of having
the stepping packing in your LISP is exactly nil.

   If stepping mode has been turned off by (step nil, every call to  eval incurs
a small  overhead--several machine instructions,  corresponding to  the compiled
code for a simple cond and one function pushdown.

   From an overhead point of view,  running with (step) entered at top  level is
the same as running with (step nil).

   Stopping stepping by responding  <tab> incurs the same continued  overhead as
(step nil).

   Running with (step foo1  foo2 ...) can be  more expensive, since a  member of
the car of the  current form into the list  (foo1 foo2 ...) is required  at each
call to eval.


Page 3-42                           ∪3-1.5.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


   In terms of memory requirements, the total compiled stepping package occupies
about 423 words of binary program storage.

   Interaction with debug and trace:

   No special interactions of the  step package with debug, trace, or  any other
system packages are known.


                            The Morgenstern Stepper

   The  Morgenstern   Stepper  package   provides  debugging   capabilities  for
interpreted LISP programs  that are comparable  to the capabilities  provided by
DDT for assembler code.  These capabilities include:


1) Single stepping through the evaluation  of a function and over or  into other
    interpreted functions, when  called, on a  selective basis as  determined by
    the user.  Each such form and its resulting value may be displayed.

2) Dynamic breakpointing on one  or more of the following conditions:   the form
    or atom about to be evaluated matches a pattern you provide; the  form being
    evaluated involves a specified function; a given atomic symbol  evaluates to
    a given value; a given atomic symbol  is to be bound in a prog,  either type
    of  do,  or an  eval'd  lambda-expression; or  upon  an  arbitrary condition
    specified by a predicate written as LISP code.

3)  Returning a  different  value for  a  given S-expression.   This  allows for
    changing the action  that would be selected  by conditionals in  the program
    and/or by  go's in a  prog or  do.  You can  also go to  any tag  inside the
    current prog.

4) These capabilities may be requested when the program is initially  started by
    a top-level form, or they may be initiated at any other point in  the course
    of execution - either from  the terminal while in a breakpoint,  or directly
    by the program.


   The stepper may be invoked initially  by using the function mev as  one would
use eval of one argument; e.g.  (mev '(fcn arg1 arg2)).  From a breakpoint or in
a program, the stepper may be turned on by invoking (hkstart) with no arguments.
It may be turned off by the q command described below, or of course by  a CTRL/g


December 6, 1977                    ∪3-1.5.4                           Page 3-43
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


break.  After mev evaluates its argument, it returns the value and turns off the
stepper.  Note that in  the above example the form  given as an argument  to mev
was quoted.  If, say, the value of f was the S-expression (fcn arg1  arg2), then
one could use (mev f) instead.

   At  any  point during  the  stepping, one  may  inspect the  values  of other
variables, and  even reapply mev  to any form.   This may be  done in  either of
three ways.  Each command will be prompted for by //, usually following the last
form printed out.  Any S-expression that  is typed which is not recognized  as a
command will be eval'd (within an errset to catch errors).  Alternatively, the e
command to eval any  expression, or the h command  to get a nice type  of CTRL/h
break.  (This is really a CTRL/b break, but it used to be CTRL/h so  the command
happens to be called h.)

   in the its  implementation each command must  be followed by a  space (unless
the command  is a  list).  in the  multics implementation  each command  must be
followed by a newline.  actually, this depends not on the implementation  but on
(status linmode).  each form and result which is printed out will be followed by
#number  indicating the  relative level  of evaluation  (i.e. stack  depth since
invocation).

   the primary commands are:

d   (mnemonic for  down) go  down to  the next  deeper level  of  evaluation and
   display the first  form there before  evaluating it.  e.g.  if the form  is a
   function call, this will display the first argument of the function if it has
   arguments in the  call; otherwise it will  display the first  s-expression of
   the body of the function.  it then prompts for the next command.

e  (eval)  can be used  to evaluate  an arbitrary expression.   it starts  a new
   line, waits for you to type the expression, then eval's it within  an errset,
   and prints the result.  this  is comparable to just typing the  expression or
   atom after the //, but cannot  be confused with a command, and the  format is
   nicer.

h  (control-h) enters  a break loop, and  when $p'ed displays the  current form.
   within the  break, one can  inspect the values  of variables, etc.,  and even
   reapply mev to any form.

n  (next) Display the next form at this level, without showing or inspecting the
   evaluation of the lower levels of the current form.  The value of the current
   form is displayed first.  If you  wish a condition to be tested for  at lower
   levels, use nn instead.

Page 3-44                           ∪3-1.5.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


nn Like n but slower since it inspects the lower levels.  Use instead of  n when
   testing for a condition.

u  (up) Go up to the next  higher level of evaluation and show the next  form at
   that  level.  The  form(s)  at the  current  and lower  levels  are evaluated
   without  display.   As  an  example  of its  use,  after  you  have  seen the
   evaluation of the arguments to a function, the next form to be  evaluated, if
   the function  is being  interpreted, will  be the  first S-expression  of the
   function; to avoid seeing how  the function is evaluated internally,  you can
   type u.  Note that the lower  levels are not inspected - thus if  a condition
   is to be tested for at these levels, use uu.

(u num)  If  num is positive  (or zero), forms  are not inspected  nor displayed
   until level num  is returned to.   If negative, it  goes up (abs  num) levels
   relative to the current level.  Thus (u -1) is equivalent to u.

uu Like u but slower.  Use if testing for a condition.

(uu num) Like (u num) but slower.  Use if testing for a condition.

q  (quit) Exit from the stepper.

s  (show  or display  mode)  For  datapoints and  other display  terminals, this
   gives  a nice  easily  read output  of  selected levels  that  constitute the
   context  of the  current evaluation.   Specifically, it  selects  the current
   level for sprint'ing (pretty printing)  as a "header", and as you  go deeper,
   the local context  is abbreviate-printed under  this header, and  the current
   output will be sprint'ed.  s may be used as often as you like.   Headers will
   automatically  be popped  when you  return.  The  command (s  num)  selects a
   particular  level  as a  header.   It and  the  command sn  and  several user
   settable parameters are described in the more detailed section below.

(= s-exp)   The S-expression  is substituted  for the  current form  and another
   command is prompted for (i.e. you can  step into or over the new form  if you
   want to).  When the resulting value is returned it will be as if the original
   form had yielded that value.  For example, you can change the  apparent truth
   or falsity of predicates  or bypass a (go  label), as well as  just returning
   different values for an S-expression.

(cond ...)  Tests  for conditions prior to  evaluation of each future  form, and
   when satisfied will print a  message, display the form, and wait  for another
   command (which may of course be h for a break).  The argument to this cond is


December 6, 1977                    ∪3-1.5.4                           Page 3-45
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   an  arbitrary S-expression  or symbol  which is  evaluated like  a predicate.
   This is similar to the cond feature of the trace package.

   In specifying the predicate, the  form about to be evaluated may  be obtained
   as the value of the variable %%form.  The expression (hooklevel)  returns the
   relative level of evaluation.  More than one predicate may be given, in which
   case they are or'ed together,  except when two arguments form a  special test
   as described in the more  detailed section below.  The condition  will remain
   active  at all  levels that  are inspected  by the  stepper  until explicitly
   turned off by (cond nil).

   (matchf ...) is a function which will pattern match against the current form.
   It may be used in the predicate of the cond.  (Also see its related use  as a
   command.)  The argument  to matchf is compared  to %%form element  by element
   from left to right, and  succeeds when each element of the  pattern succeeds.
   Of  importance, the  pattern need  not include  the entire  form.   * matches
   anything.   The  procedure is  applied  recursively to  sublists,  unless the
   sublist is  of the  form (#  ...) in  which case  # is  bound to  the current
   element of %%form and  the cdr (not cadr) of  the #-list is evaluated  as the
   test on that element.  Except in  this case, atoms and lists should  be given
   as in the original code  since they are not evaluated.  Some  simple examples
   are:

      (matchf xyz) succeeds if the atom xyz is about to be evaluated.

      (matchf (setq alpha)) succeeds if the atom alpha is about to be setq'd.

      (matchf  (putprop name  * 'source))   succeeds if  the property  source is
         about to be  putprop'd on the  atom pointed to  by (i.e. the  value of)
         name.

      (matchf (setq (#  member # '(alpha beta  s3)))) succeeds if  either alpha,
         beta, or s3 is about to be setq'd.

      (matchf (rplacd * '(* 9))) matches (rplacd (last urlist) '(2 9 4)).

      (matchf ((# member # '(foo  bar)))) succeeds if a function call  to either
         foo or bar is about to  be evaluated (more precisely if the car  of the
         form about to be evaluated is either foo or bar).

nil (cond nil) turns the condition off and saves the current non-nil condition.



Page 3-46                           ∪3-1.5.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


(cond) When no argument is given,  the last non-nil condition (which is  the old
   property of  %%cond) is established  as the current  condition (which  is the
   value of %%cond).  (If the previous condition was not nil then it is saved as
   the old property, thus allowing for alternation of two conditions.)

(matchf ...) is equivalent to (cond (matchf ...)), see above.

↓_The following functions are useful in connection with the stepper_↓.

   (hkstart) will initiate stepping when encountered in a program or  typed from
a breakpoint.  (hkstop) will act like the q command to turn off  stepping. (Also
see below for more info.)

   (mbak) is  a function to  be used like  the LISP system's  (baklist).  (mbak)
strips out from the result of (baklist) those functions that have to do with the
stepper.

   The remainder  of this section  is a complete  list of the  Stepper commands,
which can be used for reference.

   Commands which are not lists must be followed by a space.  You can use rubout
before completing the command (and its space if necessary).   Alternatively, you
may abort the command before completing it by doing a CTRL/x break.

   Any S-expression that you type which  is not recognized as a command  will be
evaluated (within an errset to catch errors).  Thus you can evaluate any atom or
do any function call simply by  typing it following the prompting // as  long as
it is not interpretable  as one of the commands  below (or nil).  Note  that you
can actually go to  a tag within your prog  simply by typing (go tag)  after the
//.  To evaluate a form which  looks like a command, type (or form)  to evaluate
it, e.g. (or a) evaluates the atom a.  If you want you can even  write functions
which know about the stepper and treat them as commands.

a    (all) Automatically  displays all forms and  values seen by the  stepper at
     all levels.  Typing a space  at any time thereafter will cause  the stepper
     to leave this mode and prompt  for a new command.  If you want  the stepper
     to wait for a command after each form, you can use the d command.

     Commands a  ad (a -)  c and cc  pause after each  new form is  displayed if
     %%ac-sleep is non-nil.  Its value is used as the sleep time in seconds.

ad   (all down) Automatically displays  all forms and values encountered  by the


December 6, 1977                    ∪3-1.5.4                           Page 3-47
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     stepper in evaluating the current  form (i.e. at deeper levels).   Typing a
     space prior  to completion will  cause the stepper  to leave this  mode and
     prompt  for a  new command.   (Also  see d.)   Sleeps after  each  form, as
     described under the a command.

(a lev)  Automatically displays all forms and values at the indicated  level and
     lower (deeper) levels, turning itself  off when evaluation pops to  a level
     with a  smaller level  number.   Typing  a space  prior to  completion will
     cause the stepper to leave this  mode and prompt for a new  command.  (Also
     see d.)  Sleeps after each form, as described under the a command.

b    Sets a breakpoint  to occur after evaluation  of the current form.   At the
     break, the value to be returned is the value of %%value, and may be changed
     by setq'ing this variable.  The  form that yielded this value is  the value
     of %%form.  Type $P to proceed from the breakpoint.  If you prefer that the
     system wait rather than break see the wtif command.  (b operates  by adding
     the current hooklevel to  %%breaklist.)  You can get automatic  breaking at
     all levels by using (retcond t) or conditional breaking as  described below
     for the (retcond ...) command.

c    (current) Automatically displays all  forms and values at just  the current
     level.   Typing a  space at  any  time during  the display  will  cause the
     stepper to leave this mode and prompt for a new command.  The  stepper does
     not inspect the forms of lower levels - thus if a condition is to be tested
     for at these levels, use  cc.   Sleeps after each form, as  described under
     the a command.

cc   Like c, but inspects the lower levels.

ctog  Flips the %%condnotallow toggle which is initially t, meaning do not allow
     c, m, n, or u commands if a condition is being tested for.  nil means allow
     these anyway.

(cond ...)  Tests for conditions  prior to evaluation of each future  form.  For
     pattern matching against the form  using the matchf function and  for other
     information see the description of (cond ...) above.

special tests for (cond ...) :
     To  aid  the  specification  of common  tests,  the  following  "flags" are
     provided - the same effects could be obtained by inspecting %%form  in your
     own predicate given to cond.  If the first argument to the cond is from the
     set (form  formq bind bindq  atomval atomvalq fcn  fcnq and andq)  then the


Page 3-48                           ∪3-1.5.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


     second argument is  used to derive a  test.  This process is  repeated with
     the remaining arguments,  if any.  The  resulting tests, together  with any
     remaining  arguments not  satisfying  this process,  are  effectively or'ed
     together to  derive the  overall condition  (except for  the and  andq flag
     special  tests which  are and'ed).   The arguments  are not  evaluated when
     typed but  are evaluated each  time the condition  is tested.   These flags
     each may be used more than once.

The meanings of these flags are:


   andq             The next  argument is and'ed  with the remaining  tests, and
                    must  yield  a  non-nil  value  for  the  remainder  of  the
                    condition to  succeed.  (See  the comments  for cond  in the
                    "complete list of commands" below regarding the use  of side
                    effects)

   atomvalq         The  next argument  should be  a list  of two  elements, the
                    first  an (unquoted)  name of  an atom,  and the  second the
                    value of this atom for the test to succeed.

   bindq            Watch for the following (unquoted) atomic symbol to be bound
                    in  a  prog, or  in  either  type of  do,  or  an explicitly
                    evaluated  lambda (as  distinct  from an  applied  lambda or
                    function call).

   fcnq  Watch for the following (unquoted) function name to be seen by  eval as
                    the car  of the  form about to  be evaluated.   (This cannot
                    check for applied or mapped function calls).

   formq  The following (unquoted) S-expression is to be watched for.  E.g. used
                    to  check  when  a  particular  variable  is  about   to  be
                    evaluated.
               These evaluate their argument  each time the condition  is tested
     and
               in order to get the  desired S-expression or atom name,  and then
     bind
               perform  like  their "q"  counterparts.   These  are particularly
     fcn
               useful if the  flag's argument is the  value of a  variable.  (Be
     form
               sure not  to change the  variable's value accidentally  while the
     atomval
               condition remains in effect.)

   As a simple example, (cond fcnq rplacd) will check and stop when the function
                    rplacd is about to be used  (i.e. when it is the the  car of
                    the form to be evaluated).

December 6, 1977                    ∪3-1.5.4                           Page 3-49
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     The  commands c,  m, n,  and  u do  not inspect  all levels,  and  thus the
     condition cannot be tested for at these levels.  You can use cc, nn, mm, or
     uu instead, or  use the ctog  command.  Naturally, condition  testing slows
     the speed  of execution  at levels that  are inspected  by the  stepper but
     which you do not have displayed.

     If you choose to, you can have your predicates produce side-effects such as
     recording information  of value  to you or  setting states  for use  by the
     condition later.  You can  use the and, andq  flags (more than once  if you
     like) to have the expressions executed even upon success, so long  as these
     flags appear first in the condition.  Other conditions are evaluated in the
     order of appearance until the first success is found.

d    (down) Displays the next  level down (as described above also).   Note that
     if the form is an atom, the effect is the same as the n command.   Hence if
     you want the  stepper to display  every form and value,  but to wait  for a
     command after each form, just keep using the d command.

e    (eval) Can be used to evaluate an arbitrary S-expression.  It starts  a new
     line, waits for you to type the expression, evaluates it within  an errset,
     and prints the  result.  Comparable to just  typing the expression  or atom
     after the  //, but cannot  be confused  with a command,  and the  format is
     nicer.

(=  S-exp)  Replaces  the current  form with  the given  S-expression,  and then
     prompts  for another  command, as  described above.   If two  arguments are
     given,  then this  expression will  not be  treated as  a  stepper command,
     rather it will be evaluated (see comments at top of this section).

h                   CTRL/h break is  executed.  The current form  is redisplayed
     when $P is typed.  The form  about to be evaluated is the value  of %%form.
     Within the break, one can  inspect the values of variables, etc.,  and even
     reapply mev to any form.


k                   (kill)  Does not evaluate  the current form nor  display any
     value.  This is  good for  avoiding side  effects if  restepping  through a
     program again.  Equivalent to (= nil) followed by m command.


lr                  (last result) A complete rather than abbreviated printout of
     the last result is given. (See (p - -) for further information.)


Page 3-50                           ∪3-1.5.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


m                   Next,  like n  but the  result of  the current  form  is not
     displayed.  If a condition is to be tested for at lower levels, use mm.


(matchf ...)  is equivalent to  (cond (matchf ...)), see the the  description of
     (cond ...) above.


mm                  Next, like  nn but  the result  of the  current form  is not
     displayed.


n                   (next) Displays the value  of the current form  and displays
     the next form,  then awaits the next  command.  Does not inspect  the lower
     levels.   If a  condition  is to  be tested  for  at lower  levels,  use nn
     instead.


nn                  Like n but inspects the lower levels.


o                   (old)  Does (mev 'last form).  This is useful for seeing how
     a form produced an unexpected value when you went over it with n or nn.  If
     reevaluating the form can produce  side effects be careful.  Can  be exited
     from by the xx command.  (The old form is the value of %%oldform.)


ol                  (old,  at  current  level)  Does  (mev  'last  form  at this
     level).  Behaves  like o.   Useful to see  the form  (at this  level) which
     produced the current value  - rather than the  last form printed out,  as o
     would yield.  (The old form  used here can be obtained by  (get %%hooklevel
     'oldform).)


po(in)(print)  Redisplays the current form.  This is useful if you wish to clear
     the screen first with control-L.  Gives typical abbreviated display (see (p
     -  -)), except  has somewhat  different effect  if in  display mode  (see s
     command).  (For  hackers of special  data structures, e.g.  "owl", printing
     will be done with the function which is the value of the atom prin1 if non-
     nil - as also applies to top-level in LISP.  This value of prin1 is checked
     only in the mev function.   Moreover, unless you request LISP not  to "snap
     links" in compiled code, you may have to reload the stepper  after changing
     prin1.

December 6, 1977                    ∪3-1.5.4                           Page 3-51
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


pp (full print) Gives a complete printout of the current form.

ppp  (even better  printout)  Pretty-prints  the current  form using  the sprint
     function.  Uses a lot of screen  in general, and so will turn  on pagepause
     for you.

(p -  -) Resets the  parameters for the  abbreviated printout used  for results,
     forms and the p command.  The first parameter is the prinlevel,  the second
     is the prinlength; both must be given.  If nil is given instead of a number
     no abbreviating is  done with respect to  that parameter; thus (p  nil nil)
     turns  off   abbreviation.   (The  current   settings  are  the   value  of
     %%hookprin.)

q  (quit) Exits  from the stepper.   Previously requested breaks  and conditions
     are disabled, and any non-nil  conditions are saved on the old  property of
     the condition name.  (Control-G also exits as usual.)

s (show or display mode)  For datapoints and other display terminals, this gives
     a nice easily read output of selected levels that constitute the context of
     the current  evaluation.  Specifically,  it selects  the current  level for
     sprinting  as  a "header",  and  as you  go  deeper, the  local  context is
     abbreviate-printed  under  this  header, and  the  current  output  will be
     sprinted.  s may be used as often as you like.   Headers will automatically
     be popped when  you return.  All sprinting  is done with pagepause  on.  If
     control-X is  typed during sprinting,  that expression will  be redisplayed
     using  abbreviated-printing  instead.  When  in  this display  mode,  the p
     command will clear the screen from the last form down, unless  preceeded by
     control-L (or if wrap-around occurred),  in which case the screen  is fully
     redisplayed.  Also see (s arg) for more information and options.

(s arg)  If arg is positive, this selects the form at that level as the "header"
     for s(how) mode.   If negative, it  uses the form  at arg levels  above the
     current  one.  If  arg is  nil,  display mode  is turned  off  (headers are
     remembered though).   (s t)  just turns  display mode  on if  currently off
     using the previously remembered headers  if still applicable; but if  it is
     already on,  this pops the  stack of headers  by one (normally  headers are
     automatically popped when  the level is  returned from).  All  sprinting is
     done  with pagepause  on.   If control-X  is typed  during  sprinting, that
     expression will  be redisplayed  using abbreviated-printing  instead.  Also
     see the sn command.

Several parameters  are user settable  from their defaults.   %%lowerdisplay and


Page 3-52                           ∪3-1.5.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


     %%lowerdisplay-min  control the  maximum and  minimum number  of  levels to
     display  below  the  header  (defaults  of  5  and  2).   This  is  done in
     abbreviate-printed form using %%shortprin which is a list of  the prinlevel
     and prinlength (defaults 3 and 3).  Sprinting of forms and results  will be
     abbreviate-sprinted  by  the  msprint  function  if  the  flatsize  of  the
     expression exceeds %%flatsize-max  (default about 450).  The  prinlevel and
     prinlength for the latter are  the list which is the value  of %%sprintabbr
     (default is (7  8)).  If %%flatsize is  nil, full sprinting will  always be
     used;  (if  negative,  abbreviate-sprinting will  always  be  used  so that
     infinite  printing  circular structures  will  sprint  and abbreviate-print
     finitely.  To  turn off  sprinting of results  setq %%result-sprint  to nil
     (default  t).  If  %%mdistitle is  neither  nil nor  a number,  it  will be
     evaluated just after the screen  is cleared, allowing printing of  a title.
     If it is a number,  that number of blank lines  will be left at the  top of
     the screen (also see sviewmsg function below).  If the partial  clearing of
     the screen bothers your eyes, setq'ing %%eyestrain1 to a number  of seconds
     (e.g. 0.5 to 2.0) will slow down the new display depending on the number of
     lines cleared.

sn  Just for  s(how) display  mode.  It  prevents clearing  of the  screen after
     prompting for another command, but  only until the next prompting  // after
     that.  Useful if you want a result to remain displayed a little longer.  If
     you want to prevent clearing of the screen for more than a couple of times,
     use (s nil), then do (s t) when you want to resume display mode.

(retcond  ...)  Tests  for conditions  just after  each form  is  evaluated, and
     breaks when  such condition is  satisfied.  At the  break, the value  to be
     returned  is the  value of  %%value, and  may be  changed by  setq'ing this
     variable.  The form that yielded  this value is the value of  %%form.  Type
     $P to  proceed from the  breakpoint.  The conditions  are specified  as for
     (cond ...).  Note that (retcond t)  will give you a break as each  level is
     popped (returned from),  including levels above  the one where  the request
     was  made.  (retcond  nil)  disables the  retcond.  If  you  prefer waiting
     rather than breaking see the wtif command.

Two additional flags are available:

   valueq  The  test (equal %%value  next-argument) is performed  as if  it were
   and'ed with the remaining predicates in the condition.

   value  Like valueq but the test is (equal %%value (eval next-argument)).  The
   overall condition is maintained on  the value of the atom %%retcond,  and the


December 6, 1977                    ∪3-1.5.4                           Page 3-53
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   previous non-nil condition is on the old property of this atom.  If  you want
   both  cond and  retcond conditions  to be  the same  you can  (setq %%retcond
   %%cond).  The value and valueq predicates will be ignored in a (cond ...).

u  (up)  Go up  to next  higher level.   Current and  lower levels  are executed
     without display.  The lower levels are not inspected - thus if  a condition
     is to be tested for at these levels, use uu.  This can be used to  skip the
     display  of  a  function's  internal  evaluation  after  having   seen  the
     arguments, as described in the previous section.

(u num)  If  num is positive  (or zero), forms  are not inspected  nor displayed
     until that level  number is reached.  If  negative, it goes up  this number
     (absolute value) of levels relative  to the current level.  Thus (u  -1) is
     equivalent to u .

uu Like u, but  also inspects lower levels.  Use  if you have a condition  to be
     tested.

(uu num) Like (u  num) but slower.  Use if  testing for a condition.   Note that
     (uu  -999) effectively  means  that you  won't  see any  levels  unless the
     condition in a cond or retcond is satisfied.

wtal  (wait-all)   Flips  a  toggle  which when  on  causes  a  pause  after the
     evaluation of every  form, but before that  value is returned.   The system
     waits for an input character.  Typing y(es), b(reak), or h  (for control-h)
     followed by space will cause a break as would the b command.  Typing just a
     space, or any  other character followed by  a space, will proceed  from the
     pause.  Default is off.

wtif  (wait-if)   Flips a  toggle which  when on  causes requests  by the  b and
     (retcond ...) commands to result in a pause rather than a break.  The pause
     is like that  of the wtal command,  and may be proceeded  by a space;  or a
     break initiated by typing y, b, or h followed by a space.  Default is off.

xx Does a control-X type of  LISP quit.  (A control-X typed after the  // prompt
     will be caught by  an errset.  The xx  command is executed outside  of that
     errset.)

The following other facilities exist:

(gethklevel  num)   This  function  returns  the  S-expression  that  is  on the
     execution stack of the stepper at the given level number (see hkshow).  Can


Page 3-54                           ∪3-1.5.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


     be used to get an unsprinted unabbreviated display of the form or to record
     or process the form as you desire, including reapplication of mev to  it in
     the current context.

(hkshow  num)   This function  will  display  previous forms  which  are  on the
     execution stack, as seen by  the stepper while it has been  activated.  The
     previous  num of  levels are  shown,  with the  current form  last.   If no
     argument is given,  then all levels are  shown.  The display is  done under
     the control of prinlevel and prinlength  which are settable by the (p  - -)
     command.  Of course this function can also be used as if it were  a command
     by typing it after the prompting //.

(hksprint num)  This function will sprint the form on the level whose  number is
     given as the argument.  Can also be used as a command.

(hkstart)  Use this function to invoke or reinvoke the stepper from a breakpoint
     or  from  a program  as  described above.   If  used within  a  break, type
     (hkstart) by itself rather than within another S-expression or function, as
     it has to climb the stack from the point of invocation.  If an  argument is
     given  to this  fexpr,  it will  be  evaluated just  prior  to establishing
     stepping, with ↑w bound  to nil, so that  you can print out  information if
     called from a program.

(It is possible for the invocation of the stepper by this method to have limited
     scope  under  some  circumstances.   Such  a  boundary  would  be  a second
     breakpoint higher on the stack or a previously terminated invocation of the
     stepper that  is still  on the stack.   Also if  the program  was initially
     started without  mev, and stepping  is retained thoughout  the rest  of the
     execution, stepping may also remain for forms typed at top level -  to stop
     this just do control-G (or use the q command) .)

(hkstop)  This function  turns off the stepper  whenever executed - in  the same
     manner as the q command would.

hooklist  is  an atom whose  value is  inspected before each  attempt to  read a
     command from the console.   If hooklist is non-nil,  it is assumed to  be a
     list of commands to the stepper - each is printed out when used and treated
     as if it came  from your typein.  hooklist  is also examined at  each level
     that is inspected by the stepper  even if no command reading is  done (e.g.
     nn or uu modes).

(mbak)   This function  gives (baklist)  but without  the stepper  functions, as
     described above.

December 6, 1977                    ∪3-1.5.4                           Page 3-55
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


(mev top-form)  This function initiates stepping and otherwise acts like eval of
     one argument, as described above.

(msprint form)   Gives abbreviated sprinting  of the form.   A second  and third
     numeric argument specify the effective prinlevel and prinlength  here, else
     a list of two numbers found as the value of msprint are used.   The current
     implementation is somewhat slow as  the regular sprint does not  respond to
     standard abbreviating.

(sviewmsg  lineno toeval)   Useful in  conjunction with  s(how) mode.   Puts the
     cursor at the  lineno and evaluates the  second argument, then  returns the
     cursor to its original position.  lineno = 0 means top; if  negative counts
     from bottom, with -1 the  bottom line.  Typically have %%mdistitle  (see (s
     -) command) be a number to  skip lines on top, and use sviewmsg  to display
     your debugging information up there.

   If you really want  specialized processing in particular situations,  you can
inspect  and/or  change %%form  in  a (cond  ...)  predicate, and  %%value  in a
(retcond  ...).  If  %%nohookflag  is t,  form  and value  printout  and command
reading (except from a non-nil hooklist) is inhibited until it is reset  to nil.
Normal command processing is invoked by (%%mhookcom) with %%nohookflag  bound to
nil.  Also described above are %%breaklist, %%cond, %%retcond, and %%hookprin.


1.5.5  The MAR Break Feature


   This feature is currently available only in the ITS implementation.

   The  MAR break  feature takes  advantage of  a hardware  feature  (the Memory
Address Register  break) which  interrupts whenever a  given memory  location is
accessed in a specified  way.  It allows the  LISP user to specify  an interrupt
function to run  whenever a variable  or list cell  is modified.  The  user must
first  "arm"  the interrupt  by  saying (sstatus  mar  cond loc).   cond  is the
condition on which to interrupt:

          0         Turn off the mar feature.
          1         Interrupt on instruction fetch.
          2         Interrupt on write (modification).
          3         Interrupt on all references.
                    (numbers are octal)

On a KL-10 processor, additional conditions are available:

Page 3-56                           ∪3-1.5.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


          10        Interrupt on data read.
          11        Interrupt on data read or instruction fetch.
          12        Interrupt on data read or write.
          13        Interrupt on instruction fetch or write.

loc is any s-expression; that cell is the one monitored.


   Example:

              (setq foo (list 'a 'b))
              (sstatus mar 2 foo)
will interrupt if the list cell in foo is ever rplaca'd or rplacd'd.

   An example of the use of the mar-break interrupt:
























	




December 6, 1977                    ∪3-1.5.5                           Page 3-57
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



      (defun mar-tracer (x)
             ((lambda (val)
                      (sstatus mar 2
                               (get the-mar-variable 'value))
                      (nointerrupt nil)       ;let endpagfn interrupts in
                      (terpri msgfiles)
                      (princ '|Now the variable | msgfiles)
                      (prin1 the-mar-variable msgfiles)
                      (princ '| has the value| msgfiles)
                      (prin1 val msgfiles))
              (symeval the-mar-variable)))

      (setq mar-break mar-tracer)

      (defun mar fexpr (x)
             (cond ((null x)(sstatus mar 0 nil))
                   (t (setq the-mar-variable (car x))
                                     ;make sure the variable has a value cell
                      (or (boundp the-mar-variable)
                          (set the-mar-variable nil))
                      (sstatus mar 2 (get the-mar-variable 'value)))))
      (mar quux)
      (setq quux 5)
      ;Now the variable quux has the value 5
      5
      (do ((quux 0 (+ quux 1))) ((= quux 2))
          (hack quux))
      ;Now the variable quux has the value 0
      ;Now the variable quux has the value 1
      ;Now the variable quux has the value 2
      ;Now the variable quux has the value 5
      nil

Notice that quux is altered by the  do loop, and also by the restoration  of the
old value  5.  This  example is  for a KA-10  processor.  On  a KA-10,  the user
interrupt  is  always run  after  the location  has  been changed.   On  a KL-10
processor, the interrupt occurs just before a modification or access rather than
just after.

   The mar break feature is sometimes used by DDT to debug the LISP  system.  As
long as DDT and LISP  do not both try to use  the mar break feature on  the same


Page 3-58                           ∪3-1.5.5                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


LISP at the same time, there should be no problem.  (sstatus mar 0 nil) releases
the mar break feature entirely for use by DDT.

   The suspend function will  attempt to save and  restore the state of  the mar
break feature. If you don't want an armed mar break to persist beyond a  call to
suspend, turn it off first with (sstatus mar 0 nil).

   When a CTRL/g quit (or the ↑g  function) forces a quit back to top  level, it
disables  the mar  break before  unwinding variable  bindings and  re-enables it
afterwards.  This  is because during  a CTRL/g quit  LISP may not  be in  a good
state for running user interrupt functions.

































December 6, 1977                    ∪3-1.5.5                           Page 3-59
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


1.6  Storage Management


   In Maclisp  storage for  programs and  data is  automatically managed  by the
system.  The casual  user need not concern  himself with storage  management and
need  not  read  this section.   However,  the  user who  is  curious  about the
implementation or who has to construct a subsystem on top of Maclisp may need to
be concerned with how the  internal storage management routines work and  how to
control their general  functioning.  In no case  is it necessary to  control the
exact step by step operations of storage management, but a variety  of functions
are provided to set the  general policy followed by the LISP  storage management
procedures.


1.6.1  Garbage Collection


   Garbage  collection  is the  mechanism  which LISP  uses  to  control storage
allocation.  Whenever LISP feels that too much storage is being used,  a garbage
collection  is  initiated.  The  garbage  collector traces  through  all  the S-
expressions which  can be reached  by car'ing and  cdr'ing from  internal atomic
symbols' values and property  lists, from forms and temporary  results currently
being used by the evaluator, from data used by compiled code, and from the saved
values of bound variables.   All the data which it  finds in this way  is "good"
data,  in that  it is  possible for  it to  be used  again.  Everything  else is
garbage,  which  can never  again  be used  for  anything because  it  cannot be
accessed, so the storage used by it is reclaimed and reused for creating  new S-
expressions.


gc                  FSUBR

        (gc) causes a garbage collection and returns nil.


gctwa               FSUBR

        gctwa  is used  to control  the garbage  collection of  "truly worthless
     atoms," which are atomic symbols which have no value and no properties, and
     which are not referenced by any list structure, other than the obarray (the
     current obarray if there is more than one).



Page 3-60                            ∪3-1.6                     December 6, 1977
**DRAFT**                          The System                          **DRAFT**


        (gctwa) causes truly worthless atoms  to be removed on the  next garbage
     collection.

        (gctwa t)  causes truly worthless  atoms to be  removed on  each garbage
     collection from now on.  Note:  gctwa does not evaluate its argument.

        (gctwa nil) causes this continual removal of truly worthless atoms to be
     shut  off, but  it  does not  affect  whether the  next  garbage collection
     removes twa's.

        The  value returned  by  gctwa is  a fixnum  which  is 0  if  no garbage
     collection of  truly worthless atoms  will be  done, 1 if  twa's are  to be
     gc'ed on the next  garbage collection, 10 if twa's  are to be gc'ed  on all
     garbage collections, or 11 if  both. (These numbers are octal;  the decimal
     values are 0, 1, 8, 9.)


↑d                  SWITCH

        If  the  value  of  ↑d  is  non-nil,  the  garbage  collector  prints an
     informative  message  each time  garbage  collection occurs.   (In  a Newio
     implementation, this message is output  to the files in msgfiles,  see page
     3-15) In the PDP-10 implementation,  it also prints a message when  a space
     is expanded without  first doing a complete  garbage collection, or  when a
     file or inferior job is closed because a file object was garbage collected.
     See also (status gcwho).


See also the user interrupts gc-daemon, gc-overflow, and gc-lossage.


1.6.2  Spaces


   In  Maclisp  the  storage  used for  LISP  objects  is  divided  into several
conceptual subdivisions, called spaces.  Each space contains a different type of
object.  Allocation  proceeds separately  in the  different spaces,  but garbage
collection of  all spaces  occurs together since  an object  in one  space could
contain a pointer to an object in any other space.

   For example, in the PDP-10 implementation, the spaces are as follows:

  LIST    Conses (dotted pairs) and lists.

December 6, 1977                    ∪3-1.6.1                           Page 3-61
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


  FIXNUM  Fixnums.

  FLONUM  Flonums.

  BIGNUM  Bignum headers.  Bignums also occupy fixnum and list space.

  SYMBOL  Atomic symbols.

  HUNK4   Hunks of various sizes.   PDP-10 implementations without hunks  do not
          have these spaces.

  HUNK8

  HUNK16

   ...

  ARRAY   "Special array cells."

  REGPDL  The "regular" pushdown list.

  SPECPDL The "special" pushdown list, used in binding.

  FXPDL   The fixnum pushdown list, used for temporary numeric values.

  FLPDL   The flonum pushdown list, used for temporary numeric values.

  Binary Program Space  used to hold arrays and compiled code.

  pure LIST, pure FIXNUM, pure FLONUM, pure BIGNUM, pure HUNG4, ...
          These  spaces  are  used  to  store  "pure"  (read-only)  data  of the
          indicated  types.  This  is  a feature  used to  make  subsystems more
          efficient.  See page 3-69.

   In the Multics implementation, the spaces are:

  list              Conses (dotted pairs),  lists, atomic symbols,  bignums, and
               strings.

  Static Storage    Arrays, files, and linkage to compiled code.

  markedpdl         A pushdown list of LISP objects.


Page 3-62                           ∪3-1.6.2                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


  unmarkedpdl       A pushdown list of machine data, not LISP objects.

   Note:  in the  Multics implementation there is  no space for  numbers because
numbers are stored in such a way that they do not take up any extra room.

   The precise spaces available in  a given implementation can be  determined by
using (status spcnames), (status purspcnames), and (status pdlnames).

   Associated  with each  space is  information determining  when an  attempt to
allocate in that space should cause a garbage collection.  The idea is  that one
should allocate for quite a while in  a space, and then decide that it  is worth
the trouble  of doing an  expensive garbage collection  in order to  prevent the
space from using too many bits of actual storage.

   The exact nature  of this information varies  with the space.  In  a pushdown
list  (pdl) space,  all information  must be  stored contiguously,  so  the only
parameter of  interest is how  big the pdl  is.  This can  be measured  in three
ways, so there are three parameters associated with a pdl:

  pdlsize The number of words of valid data in the pdl at the moment.

  pdlmax  The size to  which the pdl may  grow before intervention  is required.
          This is used to detect infinite recursion.

  pdlroom The size beyond which the pdl may not grow no matter what.

   A space such as a list space has three parameters, called the  gcsize, gcmax,
and gcmin.  These are in machine-dependent units of "words".  The gcsize  is the
expected size of the space; as objects are allocated in the space, it  will grow
without garbage collection until it reaches this size.  When it gets  above this
size  garbage collection  will occasionally  be required,  under control  of the
other two parameters.

   The gcmax is the maximum size to which the space should grow; if it gets this
big garbage collections may occur  quite frequently in an attempt to  prevent it
from growing bigger.

   The  gcmin  specifies  the  minimum amount  of  free  space  after  a garbage
collection.  It may be either a  fixnum, which specifies the number of  words to
be free, or a flonum, which specifies the fraction of the space to be free.  The
exact  interpretation of  this  depends on  the implementation.   In  the PDP-10
implementation, which uses free storage lists, the gcmin is the number  of words


December 6, 1977                    ∪3-1.6.2                           Page 3-63
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


which must be on the free storage list after a garbage collection.  If there are
not this many, the  space is grown, except if  its size approaches gcmax  it may
not be grown by  the full amount.  In  the Multics implementation, which  uses a
compacting garbage collector, the criterion for garbage collection is not when a
free list is exhausted but when the space reaches a certain size.  This  size is
the maximum of gcsize and the sum of the size after compactification  plus gcmin
(if it  is a fixnum)  or the size  after compactification times  1/(1-gcmin) (if
gcmin is a flonum.) The effect of this is to allow the same amount of allocation
between garbage collections as there would be in the PDP-10  implementation with
the same gcmin.

   Note that these controls over the sizes of spaces are somewhat inexact, since
there is rounding.  For instance, the PDP-10 implementation  presently allocates
memory to spaces in blocks of 512. words.  The Multics  implementation allocates
at least  16384. words  between garbage collections  and presently  controls the
size of pushdown lists in blocks of 16. words.

   Some spaces,  such as Binary  Program Space in  the PDP-10  implementation or
Static storage in the Multics implementation are not subject to detailed control
by the user.  The management  of these spaces is entirely  automatic.  Generally
these are spaces where the rate of allocation is fairly placid and most objects,
once allocated, are used forever  and never freed.  Hence the exact  policy used
for storage management in these spaces is not too important.


1.6.3  Storage Control Functions


alloc               SUBR 1 arg

        The alloc  function is  used to  examine and  set parameters  of various
     spaces  having  to do  with  storage management.   To  set  parameters, the
     argument to alloc should be  a list containing an even number  of elements.
     The first  element of  a pair is  the name  of a space,  and the  second is
     either a  fixnum or a  3-list.  A  fixnum specifies the  pdlmax (for  a pdl
     space) or the gcsize and gcmax (for other spaces.) A 3-list specifies, from
     left to right, the gcsize, gcmax, and gcmin.  nil means "don't  change this
     parameter."  Otherwise  a fixnum  must  be supplied,  except  in  the third
     element (the gcmin), where a flonum is acceptable.  A 3-list cannot be used
     with a pdl space.

        An example of this use of alloc, in the PDP-10 implementation:


Page 3-64                           ∪3-1.6.2                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**



          (alloc '(list (30000. 5000. 0.25)
                   fixnum (4000. 7000. nil)
                   symbol 6000.
                   regpdl 2000.))

     or, in the Multics implementation:

          (alloc '(list (30000. nil 0.3)
                   markedpdl 5000.
                   unmarkedpdl 5000.))

        alloc may  also be  called with  an argument  of t,  which causes  it to
     return a list of  all the spaces and their  parameters.  This list is  in a
     form such that it  could be given back to  alloc at some later time  to set
     the parameters back to what they are now.


   See page 3-89  for some status  functions which are  related to the  topic of
storage spaces.


1.6.4  Dynamic Space and Pdl Expansion


   There are several user  interrupts generated by the storage  management.  See
section page 3-16 for a description of user interrupts.  The gc-daemon interrupt
occurs  after each  garbage collection.   The argument  passed to  the gc-daemon
interrupt  handler  is  a  list  of  spaces  and  their  sizes.   In  the PDP-10
implementation, the items on the  list are of the form:  (space-name free-before
free-after size-before  size-after), where  space is  the name  of a  space, and
free-before indicates the number of cells free before the garbage collection and
free-after indicates the number of cells free afterwards.  The last  two numbers
are the size of the space (see (status spcsize)) before and after the  GC.  (The
sizes are in PDP-10 words.) In the Multics implementation, the items are  of the
form:   (space before  .  after).  In  the Multics  implementation,  where "free
cells" is  a meaningless concept,  only the difference  of these two  numbers is
significant; it represents the amount of compaction achieved.

   The gc-lossage interrupt  occurs if the garbage  collector tries to  expand a
space but fails because, for example, the operating system will not give  it any
more storage.  The argument passed to the interrupt service function is the name


December 6, 1977                    ∪3-1.6.3                           Page 3-65
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


of the space that lost.  If the interrupt handler returns, the value is ignored,
and another garbage collection is attempted.

   The pdl-overflow interrupt is  signalled when some pushdown list  exceeds its
pdlmax.  The  pdlmax is increased  slightly so that  the interrupt  handler will
have room to run.  The argument passed to the interrupt function is the  name of
the pdl  that overflowed.   If the interrupt  function uses  too much  pdl, this
interrupt will occur again.  If this happens enough times, the pdlmax will reach
the pdlroom, there will be no room  in the pdl to take a user interrupt,  and an
uncorrectable error will occur.

   The  interrupt  function  can  decide  to  terminate  the   computation  that
overflowed the pdl, for example by doing (↑g) or a throw, or it can increase the
pdlmax by using alloc or  (sstatus pdlmax) and then continue the  computation by
returning.  Note that,  unlike most other  user interrupts, if  the pdl-overflow
interrupt  function returns  nil  (or the  ";bkpt pdl-overflow"  is  $P'ed), the
computation is continued as if the pdl overflow had not occurred.

   The gc-overflow interrupt occurs when  some space (other than a  pdl) exceeds
its gcmax.  This gives  the user a chance to  decide that the size of  the space
should be increased  and the computation continued,  or that something  is wrong
and the computation should be terminated.  The argument passed to  the interrupt
handler  is the  name  of the  space  that overflowed.   The  interrupt handling
function will be able to run  because the garbage collector makes sure  that the
space is sufficiently large before signalling the interrupt, even if  this makes
it become  somewhat larger than  its gcmax.  This  interrupt is similar  to pdl-
overflow in that if  the interrupt handler function  returns at all, even  if it
returns nil, the interrupted computation proceeds.  To terminate the computation
an explicit (↑g) or throw must be done.


1.6.5  Initial Allocation


   The PDP-10 implementations  of Maclisp run on  a machine with  a limited-size
address space.  Consequently the allocation of portions of this address space to
different  uses,  such  as  LISP storage  spaces,  becomes  important.   This is
particularly true of the DEC-10 implementations, which cannot take  advantage of
paging.

   When LISP is first entered, it goes through a dialogue with the user known as
"allocation."  Normally the  dialogue simply consists  of the user  declining to


Page 3-66                           ∪3-1.6.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


specify anything,  in which  case LISP  chooses suitable  defaults.  If  a large
problem is  to be worked  on, the defaults  may be inappropriate  and it  may be
necessary  to  explicitly allocate  a  larger  amount of  storage.   It  is also
possible for the user's replies to come from a file.

   If LISP is called with a command line from DDT, for example

                            :LISP INDEX LOADER COM:

it reads the indicated  file in the same way  that it would read  .LISP. (INIT).
See below.

   On the other hand,  if LISP is called  without a command line,  it identifies
itself and asks
                                     ALLOC?

*****SAILB*****

   Suitable responses are Y, N, and CNTRL/q (in the SAIL implementation q and  Q
are synonymous with CTRL/q).   There are other obscure  characters which can  be
used as  replies to  this question,  but  these three  are sufficient  for  most
purposes.  ("?" causes  a list of  suitable responses to  be printed out.)   "N"
means that you do  not want to  specify allocation.  You  will get the  default.
CTRL/q means to read  your initialization file (see  below.) "Y" means that  you
wish to go through the following sequence of questions and answers.

*****SAILE*****

   LISP types out the names of  various spaces and their sizes.  If the  name of
the space is preceded by "#", then it cannot be expanded once allocated  by this
dialogue.   After each  question  you may  enter altmode,  which  terminates the
dialogue and gives the remaining parameters default values, or space, which goes
on to  the next question.   Before your altmode  or space you  may put  a number
which is  the size you  want that space  to be, instead  of the number  that was
printed.  CTRL/g restarts the dialogue with the "ALLOC?" question.

****SAILB*****

   If you reply with a  CTRL/q, it means to  read your initialization file.   In
the ITS implementation,  this is  either udir;  .LISP.  (INIT)  or (INIT);  udir
.LISP., where udir is your master sname.  In the TOPS-10 implementation, this is
LISP.INI in the directory you are logged in to.  In the Multics  implementation,
this is hd>start←up.LISP, where  hd is your home  directory.  Since the  Multics
implementation doesn't have the  allocation dialogue, this  file is always  read
when the LISP command  is given with no  arguments. In the SAIL  implementation,
the file will be LISP.INI in your aliased directory. Moreover, LISP.INI must  be
in /N E format (no directory page).

*****SAILE*****

   The first form in  the file should be a  comment which is used to  answer the
questions.  Note that supplying  nonexistent space names in the  comment doesn't
hurt, so you can use the same comment for different implementations.  An example
of the form of this comment is:

December 6, 1977                    ∪3-1.6.5                           Page 3-67
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



               (comment fixnum 5000 symbol 4000 flonum 2000
                        bignum 1400 regpdl 5000 hunk8 30000)

   The  remaining forms  in the  file are  simply read  and evaluated  using the
standard read-eval-print loop.






































Page 3-68                           ∪3-1.6.5                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


1.7  Implementing Subsystems with Maclisp



1.7.1  Entering LISP


   A subsystem  is an entity  that exists in  most time-sharing systems.   It is
normally a complete world  which the user enters  by typing a command.   He then
has  whatever facilities  the  subsystem offers.   A subsystem  can  be oriented
toward programming, as  the Maclisp subsystem itself  is, or it can  be oriented
toward a particular application, for instance compiling LISP programs, operating
machinery, or solving differential equations.

   Maclisp  is  frequently  used  as  a  base  on  which  to  build  subsystems.
Consequently it has  been equipped with a  number of mechanisms which  allow the
subsystem writer to gain complete control over the operation of Maclisp, make it
possible to  hide the vagaries  of Maclisp  from the naive  user of  a different
subsystem, and provide  increased efficiency in  memory and processor  usage for
heavily-used subsystems.

                           in the ITS implementation

   LISP may be entered by the :LISP (or LISP↑K) command.  The environment set up
by this command is the  standard initial environment.  LISP now goes  through an
allocation dialogue and optionally reads your .LISP. (INIT) file.  See page 3-64
for information on this.

   LISP  may be  entered by  the command  :LISP name1  name2 dev  dir  (or :LISP
dev:dir;name1 name2).  In this case the file dev:dir;name1 name2 is read  in the
same way as  a .LISP. (INIT) file.   This can be used  to start up  a subsystem.
The device defaults  to DSK, the directory  defaults to your current  DDT master
sname (..MSNAME), and the file names default to .LISP. (INIT).

   It is also possible to build a  subsystem in a LISP, then save it as  TS FOO.
The  :FOO  (or FOO↑K)  command  will  then enter  the  subsystem,  bypassing the
allocation dialogue.

                          in the DEC-10 implementation

   LISP  may  be  entered by  the  monitor  command R  LISP  or  RUN  LISP.  The
allocation dialogue (see  page 3-64) is entered.   Optionally a LISP.INI  in the


December 6, 1977                     ∪3-1.7                            Page 3-69
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


user's  directory  may  be  read.   If the  DEC-10  monitor  allows  the  use of
additional arguments on the R command  line, then the particular name of  a file
may be specified  (but not a directory).   The extension defaults to  INI.  This
file is read in the same way as a LISP.INI file.  This can be used to start up a
subsystem.

   As in the ITS  implementation, a subsystem can  be saved and then  invoked by
the appropriate R or RUN command.

*****SAILB*****

                         in the SAIL implementation

   Maclisp is run at SAIL by issuing the monitor command R MACLSP. In  addition,
one can specify an initilization file by saying, for instance: R MACLSP;FOO.  In
this case FOO.INI  on the aliased  directory will  be read. FOO.INI  must be  in
standard .INI format (i.e. must begin with a COMMENT statement and must be in /N
E format.)

*****SAILE*****
                         in the Multics implementation

   The  Maclisp subsystem  is entered  by issuing  the LISP  command  at Multics
command level.   If LISP is  called with  no arguments, a  copy of  the standard
initial environment containing  all the system  functions and variables  is made
the current environment.   If the LISP command  is issued with an  argument, the
argument concatenated  with ".sv.LISP"  is the pathname  of a  saved environment
which  is copied  into the  current environment.   This saved  environment would
contain some subsystem, which will receive control.  Additional arguments to the
LISP command in this case are actually arguments to the subsystem.

   Often one constructs  a trivial command for  getting into a  subsystem, which
simply calls the LISP command with the right arguments.

   For  instance,  the  LISP  compiler  subsystem  may  be  entered  through the
LISP←compiler  command,  which  calls  LISP  with  the  pathname  of  the  saved
environment  containing  the  LISP  compiler  as  the  first  argument,  and the
arguments to the LISP←compiler command as the remaining arguments.

   When the standard initial  environment (i.e. the ordinary  Maclisp subsystem)
is  entered, it  checks for  a segment  named start←up.LISP  in the  user's home
directory.  If such a  segment exists, it is  read in, using the  load function.
This facility allows users to "customize" LISP.



1.7.2  Saving an Environment


   A subsystem is constructed by  the following procedure.  One starts  with the
ordinary Maclisp  subsystem, and  defines a number  of function  definitions and
variable values.  This creates  an environment which is capable  of implementing
the desired subsystem.  This environment is then saved in a file,  and necessary


Page 3-70                           ∪3-1.7.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


mechanisms are set  up so that an  operating-system command can invoke  LISP and
cause  it  to  set  up  the environment  saved  in  the  file.   When  the saved
environment  is invoked  control is  passed to  the functions  in it  which then
proceed to do the business of the subsystem.

   The  exact  way of  saving  the environment  differs  from  implementation to
implementation.  In the Multics implementation there is a function called save:

save                FSUBR

        (save  foo)  saves  the  current  LISP  environment  in  a   file  named
     foo.sv.LISP in the  working directory.  foo  is not evaluated.   The saving
     operation destroys the working copy of the environment, so when the save is
     complete LISP returns to Multics command level.

        All  variable  values,  file  objects,  array  contents,   and  function
     definitions (and other properties) are saved, but the contents of  the push
     down lists, including previous values of bound variables, cannot  be saved,
     so  save should  only be  used from  top level.  (See also  the  section on
     Gaining and Keeping Control, below)


   In the PDP-10 implementations there is a function called suspend:

suspend             LSUBR 0 to 2 args

        suspend puts  LISP in a  state such  that it can  be :PDUMP'ed  (ITS) or
     SSAVE'd  (DEC-10)  and  later  restarted.  When  the  saved  core  image is
     restarted,  everything will  be  the same  as  it was  when  suspended, and
     control will return from the invocation of suspend.

        suspend may be used at any point in a computation, with  the restriction
     that no  I/O devices  other than  the terminal may  be in  use.  If  an I/O
     device other  than the  terminal is  in use,  a fail-act  correctable error
     occurs, indicating the offending device(s).

        In the ITS  implementation, care is taken  so that all  subsystems saved
     with suspend from  the same version  of LISP will  share the pure  pages of
     LISP.  In addition,  all invocations of  a particular subsystem  will share
     the  pure pages  peculiar to  that subsystem.   Declaration of  data  to be
     placed in pure pages is described in a later section.



December 6, 1977                    ∪3-1.7.2                           Page 3-71
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


        In  the TOPS-10  implementation, care  is taken  so that  all  LISPs and
     subsystems saved from LISP will share the same high segment.

        After suspend has prepared  the LISP core-image for dumping,  it returns
     control to the operating system so that it can be dumped.

        If suspend is  given an argument, that  argument is explodec'ed  and the
     resulting character  string is  passed back  to the  operating system  as a
     command.  (See cline and valret.)  (On ITS this is done with a  .VALUE;  on
     SAIL this is done with a PTLOAD.  On TOPS-10 the argument is ignored.)

        The SAIL version uses the  second argument, if present, as a  file name.
     The high segment is dumped under this name.  The low segment should then be
     saved with  SAVE, not SSAVE.   This is used  to create subsystems  that can
     share the same high segment containing code specific to the subsystem.

        Commonly one will write a setup routine for a subsystem like this:

     (progn
           (terpri)
           (princ 'options:)
           ... read in options ...
           (terpri)
           (princ 'loading)
           ... load in files of functions ...
           (gc)
           (sstatus gctime 0)
           (suspend)
           (start-the-subsystem)
     ))

        The subsystem's  environment is now  ready for  dumping.  Alternatively,
     one might write (on ITS)

     (suspend '|:PDUMP DSK:FOODIR; TS NSUBSYS/
     :$ALL DONE$/
     |)

     which will do the dump itself  and print a message when done.  On  SAIL one
     might write




Page 3-72                           ∪3-1.7.2                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**



     (suspend '|SAVE SYS:FOO.SAV|
              '(FOO SHR SYS))

     See also the valret function.


1.7.3  Gaining and Keeping Control


   In the Multics implementation, when a saved environment is restarted it looks
like  an  error  that returns  to  top  level.  The  forms  in  the  errlist are
evaluated.   These  forms  should  do whatever  is  necessary  to  start  up the
subsystem.  The  arguments to  the command  which invoked  the subsystem  may be
obtained via (status arg) or (status jcl).

   In  the  PDP-10  implementations,  when  a  saved  environment  is  restarted
execution continues  from the  point where  suspend was  called.  The  next form
evaluated should  do whatever is  necessary to start  up the subsystem.   (It is
also possible to cause the same return to top level on startup as in the Multics
implementation  by  using valret  instead  of suspend.)   The  arguments  of the
command line which invoked the subsystem may be obtained via (status jcl).

   If the subsystem wants to hide the underlying Maclisp from the user, it has a
number of facilities available.   By setting up its own  user-interrupt handlers
it can handle any LISP errors which occur itself.  In Newio  implementations, it
can alter, augment,  or abolish standard  interrupt control characters.   It can
replace the Maclisp interpretive interaction loop with its own by using (sstatus
toplevel) and  (sstatus breaklevel).   It can also  provide a  totally different
interaction loop  by not  returning control  to the  LISP top  level when  it is
started,  but instead  retaining control  in its  own functions  which  read and
respond to user input.

   It is possible for a subsystem to retain the trappings of Maclisp  but change
the way things read and print.   Macro characters and the readtable can  be used
to change  the way input  is parsed; alternatively,  setq'ing the  variable read
will redefine the system  reader function (in the PDP-10  implementations).  All
output  by Maclisp  (with the  exception of  character-string messages)  is done
through the function  prin1, and the subsystem  may redefine this  function.  In
the  Multics  implementation  one  simply  redefines  it,  but  in   the  PDP-10
implementations the  variable prin1 must  be bound to  the function which  is to
substitute for prin1.


December 6, 1977                    ∪3-1.7.2                           Page 3-73
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   Some subsystems don't do any of this, but simply consist of  standard Maclisp
augmented by some additional  functions which may be  used in forms typed  in at
top level.


1.7.4  Purity


   In  the  PDP-10  implementations,  there  are  some  facilities  which  allow
subsystems to  put their non-changing  data, function definitions,  binary code,
etc.  into pure  pages.  This  decreases  the load  on memory  by  sharing pages
between multiple users of the same subsystem.

   There are some extra storage spaces which are used to store pure (unchanging)
LISP objects.  These are the  pure list, pure fixnum, pure flonum,  pure bignum,
and pure hunk spaces.

purcopy             SUBR 1 arg

        This function makes and returns a copy of its argument in  pure storage.
     This is  primarily of use  in the creation  of large sharable  systems like
     MACSYMA.  In  implementations other than  PDP-10 implementations  with pure
     spaces, purcopy simply returns its argument.


        There  are  a number  of  features  which control  how  binary  code and
     constants are purified when a compiled program is loaded into LISP.


bporg               VARIABLE

        The value of bporg should always be a fixnum, whose value is the address
     of the  first unused word  of binary program  space.  This  value generally
     should not  be altered by  the user, but  only examined.  bporg  is updated
     whenever binary code is loaded by lap or fasload.


bpend               VARIABLE

        This  variable should  also  always have  a  fixnum as  its  value; this
     indicates the last available word of binary program space.  This is updated
     by many internal  LISP routines, such as  the garbage collector,  the array
     allocator, and lap and fasload.

Page 3-74                           ∪3-1.7.3                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


pagebporg           SUBR no args

        Causes the variable bporg to be adjusted upwards so as to lie on  a page
     boundary.   This  is principally  useful  on ITS  in  conjunction  with the
     function purify.  pagebporg returns the new value of bporg.


getsp               LSUBR 1 to 2 args

        (getsp n flag)  ensures that (- bpend  bporg) is at least  n, allocating
     more memory  if necessary.  If  flag is non-nil,  then the memory  added is
     marked as being potentially purifiable by purify.  If flag is  omitted, the
     value of pure is used.  This is generally used by clever subsystems loaders
     to expand  binary program space  quickly so that  fasload will  not require
     several garbage collections to do the  same thing.  It can also be  used by
     programs which create and destroy many arrays.  See also noret.


noret               VARIABLE

        Normally the  garbage collector will  return memory to  the time-sharing
     system if (- bpend bporg) is very large, but setting noret non-nil prevents
     this.  This is useful in conjunction with getsp.


purify              SUBR 3 args

        The first two arguments to purify should be fixnums, delimiting  a range
     of memory within the LISP system.  The third argument is a flag.  If  it is
     nil,  then the  pages covered  by the  specified range  of memory  are made
     impure, i.e.  writable.   If it is  t, then the  pages are made  pure, i.e.
     read-only and sharable.  If it is bporg, then the pages are made  pure, but
     in addition some work is done to  make sure that no UUO on those  pages may
     ever be "clobbered".  (See pure and purclobrl) This option should always be
     used if the  pages involved contain binary  code loaded by lap  or fasload.
     Presently purify does nothing in the TOPS-10 implementation; it is intended
     primarily for producing systems built  on LISP, such as MACSYMA, in  such a
     way that pure  pages can be shared  between users.  Example:  the following
     function might be used to produce a sharable system on ITS:





December 6, 1977                    ∪3-1.7.4                           Page 3-75
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



           (defun superdump ()
                  (setq lopage (pagebporg))     ;save low page address
                  (setq pure 3)                 ;specifies pure code
                  (setq lopage (+ lopage 6000)) ;allow for area 1
                                                   ; (ITS page = 2000 words)
                  (fasload funny fasl)          ;load up system
                  (fasload weird fasl)
                  (uread some lap)
                       ...
                  (sstatus toplevel             ;set up top level for system
                           '(top-handler))
                  (setq hipage (pagebporg))     ;save high page address
                  (purify lopage (1- hipage) 'bporg)   ;purify pages
                  (suspend '|:pdump sys:ts super↑M|)  ;tell ddt to dump
                  (terpri)                      ;stuff for system startup
                  (princ 'welcome/ to/ supersystem/!)
                  (terpri))


pure                VARIABLE

        This variable, initially nil, should be made non-nil by the  user before
     loading binary code which is to  be made pure.  It signals lap  and fasload
     to be circumspect about any UUO's in the code, because pure UUO's cannot be
     clobbered to be PUSHJ's or  JRST's.  lap solves this problem  by clobbering
     the UUO immediately  if the referenced function  is already defined  and is
     itself a subr  rather than an expr;  otherwise the UUO is  made permanently
     unclobberable (i.e. CALL is converted to CALLF, etc.).

        fasload  is somewhat  more  clever: it  too  tries to  clobber  each UUO
     immediately, but  if it  can't it  puts the address  of the  UUO on  a list
     called purclobrl, which is checked at the end of each call to  fasload, and
     each UUO on the list is clobbered at that time, if the appropriate function
     had been loaded by  that call to fasload.   If the function never  does get
     defined, then purify will also check purclobrl and convert each UUO  to its
     permanently unclobberable form.

        If pure has a  fixnum as its value,  then fasload (but not  lap) behaves
     somewhat differently.   pure should  be the  desired size  in words  of the
     "uuolinks area";  this is rounded up to a whole number of pages.   (If pure
     is between 1  and 8, it is  the number of pages  rather than the  number of


Page 3-76                           ∪3-1.7.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


     words.)  Let the number of pages be n.  First fasload calls  pagebporg, and
     then reserves 2n pages of  binary program space, unless a previous  call to
     fasload has already reserved them (i.e. they are reserved only once).  Thus
     fasload has two sets of n pages  to work with; we shall call the  first set
     "area 1" and the second set  "area 2".  Now whenever fasload has to  load a
     clobberable UUO, it does not place it in the code being loaded,  but rather
     hashes it and places it  in area 1 if it  was not there already; a  copy is
     placed in the  same relative position in  area 2.  Then an  XCT instruction
     pointing to  the UUO  in area  1 is placed  in the  binary code.   When all
     loading has been done, area 2 may be purified, but area 1 may not.

        Now when  running the code,  the UUO's  pointed to by  the XCT's  may be
     clobbered (the PDP-10 LISP UUO  handler is clever about XCT),  provided, of
     course, that the value  of nouuo is nil, and  the code will run  faster the
     second time around  because the XCT's will  point to PUSHJ's.   However, if
     (sstatus  uuolinks) is  called, then  area 2  is copied  back into  area 1,
     effectively unclobbering all the  UUO's.  This allows the  called functions
     to be traced again, for example, or redefined as expr code.   Naturally, an
     area large  enough to  contain all  the UUO's  should be  reserved; (status
     uuolinks) (q.v.) yields information relevant to this.  As a rule  of thumb,
     the area hsould be  20% larger than the  number of functions called  by the
     loaded  code (including  LISP  functions such  as equal).   For  the DEC-10
     version, pure may  be set to  a negative value.   the magnitude is  used as
     above, while the sign controls  which segment to load into (positive  = low
     segments, negative = high segment).  A negative value also  causes uuolinks
     area 1  to go  in the low  segment, and  area 2 in  the high  segment.  For
     compatibility, a negative value means  the same as a positive value  to the
     ITS implementation.


purclobrl           VARIABLE

        Used by  fasload to keep  track of UUO's  which are potentially  but not
     immediately clobberable.


*pure               VARIABLE

        This  variable  controls  automatic  purification  of  S-expressions and
     atomic symbols.  If it is set non-nil (the initial value is nil),  then the
     following  are placed  in pure  storage spaces  instead of  regular storage
     spaces:  pnames of atomic  symbols; list, fixnum, flonum, bignum,  and hunk


December 6, 1977                    ∪3-1.7.4                           Page 3-77
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     constants used by code loaded with fasload; properties whose indicators are
     in the  list which is  the value of  the variable putprop  (initially subr,
     fsubr, and lsubr).   In the SAIL implementation,  if *pure is a  fixnum, it
     should be an estimate of  the total number of pure data  structures needed,
     including all files previously loaded and the initial LISP system pure data
     structures (currently about 6000.  words).  This causes purcopy to  use the
     high segment for pure data.  Making the estimate in *pure too  large merely
     wastes space  in the high  segment; making it  too small causes  purcopy to
     make  copies in  the low  segment  when it  runs out  of room  in  the high
     segment.  This whole feature only works if pure is a negative fixnum.


putprop             VARIABLE

        If the value of *pure is  non-nil and the third argument to  the putprop
     function is in the  list of indicators which  is the value of  the variable
     putprop, then the second argument  is passed through purcopy to  purify the
     structure.  Furthermore,  the two  cells of the  property list  are cons'ed
     from pure list  space.  Since impure cells  must precede pure cells  in the
     property list,  putprop may  not put  a new  property at  the front  of the
     property list in this case.

        The putprop  and remprop functions  know about purified  property lists.
     If necessary,  they will  copy the  property list  (but not  the properties
     themselves) into non-pure storage so that it can be modified.  This is true
     regardless of the value of  *pure.  Recall also that defprop and  defun use
     putprop.

















Page 3-78                           ∪3-1.7.4                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


1.8  Miscellaneous Functions



1.8.1  The Status Functions


status              FSUBR

        The  status special  form is  used to  get the  value of  various system
     parameters.   Its  first  argument,  not  evaluated,  is  an  atomic symbol
     indicating which of its many  functions status should perform.  The  use of
     additional  arguments  depends  on  what  the  first  argument  is.   These
     arguments may or may not be evaluated, depending on the first argument.  If
     certain  additional arguments  are omitted,  a default  value  is supplied,
     again  depending  on  what  the  first  argument  is.   The  various status
     functions are listed below.

sstatus             FSUBR

        The sstatus  special form  is used to  set the  value of  various system
     parameters.  Its arguments are similar to those of status.

These are the things that you can do with status and sstatus:

STATUS FUNCTIONS FOR I/O

tabsize   (status  tabsize) returns  the number  of character  positions assumed
          between  tab stops,  which depends  on the  implementation.  Currently
          this  is  8  in  the PDP-10  implementation  and  10.  in  the Multics
          implementation.

newline   (status newline)  returns a  fixnum which  is the  ascii code  for the
          character which marks  the end of a  line of input.  For  example, one
          might say (= (setq ch (tyi)) (status newline)) . Cureently this  is 15
          (octal) in  the PDP-10  implementation and 12  (octal) in  the Multics
          implementation.

charmode  (status charmode f) returns the value of the character-mode switch for
          the file f.  If f is t or omitted the value of tyo (the default output
          terminal) is assumed.  If  the character-mode switch is t  (the normal
          case for the terminal) output is  sent to the device as soon as  it is


December 6, 1977                     ∪3-1.8                            Page 3-79
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


          generated.  If the switch is nil (the normal case for files other than
          the  terminal) output  is  held until  a  newline is  typed,  an error
          occurs,  input is  requested,  or the  buffer becomes  full.   (In the
          Multics implementation, you  can also cause the  buffer to be  sent by
          using the force-output  function.) This provides  increased efficiency
          at the cost of not immediately seeing all output in some cases.

          (sstatus charmode x f) sets the character-mode switch of the file f (f
          may  be  t  or omitted  to  signify  the output  terminal  in  the tyo
          variable)  to x,  which  may be  nil or  t.   x and  f  are evaluated.
          Currently in the PDP-10 implementation  it is not possible to  set the
          character-mode switch;  one must  specify it  initially to  open.  See
          also (status linmode) and (status filemode).

*****SAILB*****

linmode   (status linmode) reads  the "line mode"  and (sstatus linmode  x) sets
          the "line  mode" to  x (t or  nil.) These  functions take  an optional
          extra argument, which is the file whose line mode is  being discussed.
          This defaults to  the value of tyi  (the default input  terminal).  In
          any case, this file must  be a terminal.  In some  implementations the
          "line mode" may not be changed.   If the "line mode" is t,  user input
          is buffered up a line at a time before being sent to LISP.  The input-
          editing conventions  of the  host operating system  are used.   If the
          "line  mode" is  nil, LISP  sees  each character  as it  is  typed and
          applies  its own  input editing  conventions.  This  mode  can provide
          input facilities more suited  to LISP and possibly better  handling of
          the terminal,  if it is  a type  that LISP knows  a great  deal about.
          However, it uses  more machine resources.  It  is possible for  a user
          program to take direct control of the terminal when the "line mode" is
	  nil; however, this may require knowledge of the (sstatus tty).  The
	  Multics implementation always operates with a "line mode" of t.  See
	  also the (sstatus ttyscan) function below. In the SAIL implementation,
	  the default setting is (sstatus linmode t).

*****SAILE*****

ttyint    (sstatus ttyint char  func file) turns  on a tty  interrupt character.
          When  the character  char is  typed on  the input  tty file,  the LISP
          program will be interrupted and  the function func will be  applied to
          two arguments - file and char.   If file is omitted, the value  of tyi
          (the  default  input file)  is  assumed.   The char  may  be  either a
          character object  or a  fixnum (ascii  code).  Any  of the  128. ascii
          characters may be used.  The func may be either an ordinary functional
          form  or a  fixnum,  which means  the  default system  action  for the
          character with that ascii value.  For instance a func of 7  means quit
          back to the top level of  LISP (control-G).  If func is nil,  the char
          is made non-interrupting.  All three arguments are evaluated.

Page 3-80                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


          In the ITS implementation, it may be necessary to use (sstatus tty) to
          inform ITS about non-standard interrupt characters.

          In PDP-10  implementations which support  fixnum tty input  mode (more
          than 128. characters), then when an interrupt occurs the  character is
          folded  down  to  7  bits  before  selecting  an  interrupt  function.
          However, the unfolded  character is passed  as the second  argument to
          the interrupt function, so  that the function can filter  out unwanted
          characters.  If a fixnum is used for the function, then the fixnum may
          specify required or forbidden supra-ascii bits as follows:  adding bit
          n to the fixnum requires the  bit, and adding (lsh n 18.)  forbids it.
          For example, in  the ITS implementation  using 400000207 (octal)  as a
          function  allows  CONTROL-G  to  perform  a  quit,  but  not  the "pi"
          character of  CONTROL-META-G, since 200  requires the CONTROL  bit and
          400000000 forbids the META bit.

          Example:  a  standard  subsystem convention  is  to  use  ↑↑ (control-
               uparrow,  ascii  036)  as a  quit  character  which  restarts the
               subsystem.

             (sstatus ttyint 36
                      (function (lambda (f c)
                                        (sstatus toplevel '(restart-subsystem))
                                        (do nil ((or (= (listen f) 0)
                                                     (= (tyi f) 36))))
                                        (↑g))))
             (defun restart-subsystem ()
                    (setq errlist nil)
                    (sstatus toplevel nil)
                    (nointerrupt nil)
                    (initialize-the-world)
                    (enter-subsystem))

          (status ttyint char file) returns func, the interrupt function for the
          character char on the tty file.  file may be omitted.  It  defaults to
          t.

ttycons   (sstatus ttycons tty1 tty2) binds  two tty files into a  console.  One
          should be an input tty and  the other an output tty.  If tty1  or tty2
          is t,  it will be  taken as the  appropriate direction of  the default
          terminal (determined by the variables tyi and tyo).  The  binding into
          a console  is basically  used for purposes  of echoing.   In addition,


December 6, 1977                    ∪3-1.8.1                           Page 3-81
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


          interrupt  characters typed  on an  input tty  file should  affect the
          output on its corresponding output tty file, not on some other tty.

          If one argument is a tty and the other is nil, a  ttycons relationship
          involving that tty will be broken.

          To put echoing at the bottom of the screen, distinct from output,

               (sstatus ttycons t
                                (open 'tty: '(echo out tty)))

          which conses tty input with a  new tty output channel which is  set to
          go to the echo area at the bottom of the screen.

          (status ttycons tty1) returns the other tty file which is bound into a
          console with tty1, or nil if there is none.

filemode  (status filemode  file) =>  (open-mode-list .  internal-cruft).  open-
          mode-list  is  a  suitable  second  argument  for  the  open function.
          internal-cruft is a list of implementation-dependent information which
          may sometimes be needed by special programs.

          The following  symbols may appear  in internal-cruft.  (These  are the
          standardized ones; additional symbols may appear at the  discretion of
          the implementation.)

          cursorpos      This file (an output  tty) has the ability  to position
                         its cursor anywhere on the screen.

          filepos        The  filepos function  can be  used to  access randomly
                         within the file.

          rubout         This  file   (an  output   tty)  has   selective  erase
                         capability.  (cursorpos 'x) will work.

          sail           This  file  (an  output  tty)  has  the  so-called SAIL
                         character set.  This is an extension of ascii  which is
                         related to but not necessarily identical to  the actual
                         character set used at SAIL.

          If file is a closed file object, (status filemode) merely  returns nil
          instead of giving an error.


Page 3-82                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


ttyscan   (sstatus ttyscan func file) allows the user to supply a function which
          performs initial processing of  terminal input.  func is  a functional
          form, and file must be a tty input file.  If it is omitted,  the value
          of tyi (the  default input terminal)  is assumed.  Both  arguments are
          evaluated.

          When LISP wants  to take input from  file, it first calls  the prescan
          function, which is  supposed to gobble down  a complete unit  of input
          (for instance an S-expression)  and return a list of  characters.  The
          prescanner supplied  automatically by  LISP when a  tty input  file is
          first opened  counts parentheses and  does fancy rubout  processing on
          display terminals.  It also implements the CTRL/k, CTRL/l,  and CTRL/u
          characters  (as appropriate  to  the implementation)  which  allow the
          complete  input to  be  redisplayed or  cancelled, and  takes  care of
          force-feed  characters.   A  user-written  prescanner   might  provide
          additional features  such as  super-parentheses, name  recognition and
          completion, or fancy editing.

          The prescan function func is applied to three arguments: the file, the
          name of the input function on whose behalf it is acting (read, readch,
          or readline), and a fixnum which, in the case of read, is the count of
          the number of unmatched left-parentheses.  It is supposed to  return a
          list  of fixnums,  which represent  characters.  The  prescan function
          should read  the input  with tyi, since  it and  tyipeek are  the only
          input functions which doesn't call the prescan.

          If the pre-scan function returns nil, an eof condition occurs  for the
          input file.  This is the standard way to signal over-rubout.

          There  is  a  function  called rubout  to  assist  the  pre-scanner in
          processing rubouts.  It is described on page .

          It is a good idea for the prescan function to lambda-bind echofiles to
          nil so that characters do not  appear in the echo files twice,  and so
          that the echo files reflect the "clean" input after rubout processing.
          The system-supplied prescan function does this.

          (status ttyscan file) returns the file's func.

          This feature does not presently exist in the Multics implementation.




December 6, 1977                    ∪3-1.8.1                           Page 3-83
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


The following status functions exist only in the ITS implementation

ttysize   (status ttysize f) returns the  height and width of the  terminal open
          as the file f.  If f is omitted, the value of tyo (the  default output
          terminal) is used.  The result is a dotted pair of fixnums "(ttyheight
          . ttywidth)" (cf.  cursorpos).  If the terminal is a  printing console
          instead of a display, then the height will be a very large number such
          as 200000000000 or so.

ttytype   (status ttytype f) returns the  type of the terminal open as  the file
          f.  If f is omitted, the value of tyo (the default output terminal) is
          used.   The result  is a  fixnum taken  from the  TCTYP  (not TTYTYP!)
          variable in ITS:

           0       Printing terminal.
           1       Good Datapoint.
           2       Bad Datapoint ("loser").
           3       Imlac.
           4       Tektronix.
           5       PDP-11 ("Knight") TV.
           6       Memorex.
           7       Software terminal.
          10       Terminet.
          11       Display using standard ascii display codes.
          12       Datamedia.


          This status  call is meant  only for esoteric  purposes.  To  find out
          whether you can position  the cursor or erase characters,  use (status
          filemode).

tty       (status  tty f)  returns a  list of  three fixnums  which  control the
          behavior of the terminal open as  file f.  If f is omitted,  the value
          of tyi (the default  input terminal) is used.   (sstatus tty x y  z f)
          sets these three fixnums to x, y,  and z.  If z is omitted, it  is not
          changed.   If  f is  omitted,  the  value of  tyi  (the  default input
          terminal) is used.   The three fixnums  are the ITS  variables TTYST1,
          TTYST2, and TTYSTS.  Their meaning is as follows:

          TTYST1 and TTYST2  are divided into twelve  groups of six  bits.  Each
          group  controls  a  certain subset  of  the  ascii  characters.  These
          groups, in left-to-right order within TTYST1 and TTYST2, are:


Page 3-84                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


                    (0) ↑@-↑F ↑K ↑L ↑N-↑R ↑T-↑Z ↑\-↑←
                    (1) A-Z a-z (letters)
                    (2) 0-9     (digits)
                    (3) ! " α# $ % & ' , . : ; ? @ \ ` | ~
                    (4) + * - / = ↑ ← (arithmetic operators)
                    (5) < > ( ) [ ] α (parentheses)
                    (6) ↑G ↑S
                    (7) ↑I ↑J (tab, linefeed)
                    (8) altmode (ascii 33)
                    (9) ↑M (carriage return)
                    (10) rubout (ascii 177)
                    (11) space, ↑H (backspace)

          40   Echo the character when read by LISP (normal mode).
          20    Echo  the character  when  typed (this  may  disappear  from ITS
               soon!).
          10   Echo in "image mode" rather than "ascii mode"
           4   Convert lower case to upper case (applicable only to letters; not
               used by LISP).
           2   Activation characters.  If this bit is not set, then ITS will not
               schedule LISP to run when  you type a character within  the group
               even  if  LISP  was waiting  for  terminal  input.   For example,
               letters and digits normally  have this bit off, since  typing one
               cannot  terminate  an  S-expression, and  there  is  no  point in
               "activating" LISP until an S-expression terminator like  space or
               ")" has been  typed.  An exception to  this rule is that  for tyi
               and readch LISP asks ITS to activate on any  character.  Normally
               only  groups  (0),  (5),  (6),  (7),  (9),  (10),  and  (11)  are
               activators.  If you alter  the readtable, you may wish  to change
               the activator groups.
           1   Interrupt characters.  If this  bit is not set, then even  if you
               have used (sstatus  ttyint) to set  up an interrupt  function you
               will not get  the interrupt since ITS  won't tell LISP  about it.
               Initially only groups (0) and (6) cause interrupts (these include
               all control-characters except altmode and format  effectors).  If
               you give  @ an  interrupt function, for  example, you  should set
               this bit for group (3).
               On a keyboard  with supra-ascii characters,  one may wish  to use
               meta-characters for  interrupts.  Suppose we  want meta-D  to run
               the function foo.   Then we would turn  on the interrupt  bit for
               group 1 (letters) and say:



December 6, 1977                    ∪3-1.8.1                           Page 3-85
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



             (sstatus ttyint 104    ;letter D
                      (function (lambda (f c)
                                (and (= c 504)  ;meta bit
                                     (foo))))   ; is 400
                                       - filter out plain "D"

          The third variable, TTYSTS, has these bits of interrest:
          40000←22 (%TSFCO) If set, output on this terminal uses the supra-ascii
                    convention of an "alpha" prefix for "control" and "beta" for
                    "meta".
          20000←22 (%TSALT) If set, ascii  codes 175 and 176 are  ∪not converted
                    to 33 on input.
          10000←22 (%TSROL) If set, terminal is in scroll mode.
           4000←22  (%TSSAI)  If set,  echoing  and ascii  output  use  the SAIL
                    character set; otherwise,  control characters are  output in
                    the form "↑X".
             10←22 (%TSNOE) If set, defer echoing of typed characters.
              2←22 (%TSSII) If set, terminal uses "super-image input" mode.  Not
                    even ↑Z and ↑← can take their usual effect for ITS.

          Notice that some of these bits affect output even though  the argument
          is an input file; they  affect the output terminal which  is logically
          related by ITS to the input terminal.  Normally LISP or the  user will
          arrange for (status ttycons) to reflect this relationship.


STATUS FUNCTIONS FOR THE OLD I/O SYSTEM

uread     (status uread) returns a 4-list for the current uread input source, or
          nil if uread is not being done.

          (sstatus uread --args--) is the same as (uread --args--)

uwrite    (status  uwrite)  returns the  2-list  for the  current  uwrite output
          destination.

          (sstatus uwrite --args--) is the same as (uwrite --args--)

crunit    (status crunit) returns a 2-list of the current unit; i.e.  device and
          directory.



Page 3-86                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


          (sstatus crunit device directory) sets the current default  device and
          directory for uread, etc.  The arguments are not evaluated.

crfile    (status crfile) returns a 2-list giving the file names for the current
          file in the "uread" I/O system.

          (sstatus crfile name1 name2)  sets the current default file  names for
          uread, etc.  The arguments are not evaluated.

STATUS FUNCTIONS FOR THE READER

   See section  13.6.2 for  a description  of how  the parameters  controlled by
these functions are used.  All  of these parameters are kept in  the readtables,
and the status functions below deal with the readtable which is the value of the
variable readtable.  Note: in the following, c represents an argument specifying
a character.  If c is non-atomic it is evaluated, and the value must be a fixnum
which is the ascii  code for a character.  If  c is atomic it is  not evaluated,
and it may be a fixnum or a character object.

chtran    (status chtran c) gets  the character translation table entry  for the
          character c.  This is the ascii code of a character substituted  for c
          when it appears in a pname being read in.  This feature is used in the
          PDP-10 implementations to translate lower-case input to upper case.

          (sstatus chtran  c k)  sets c's  character translation  to k.   f2k is
          always evaluated.  See the setsyntax function.

syntax    (status syntax c)  returns the syntax bits  for the character c,  as a
          fixnum.

          (sstatus syntax c m)  sets c's syntax bits  to m.  m is  evaluated and
          returned.  The setsyntax function is usually a better way to  do this,
          however.

Note that  in the  above two  sstatus calls,  if c  is a  macro character  it is
changed back to its standard syntax and chtran before the requested operation is
performed.  However, if in the standard  readtable c is a macro (i.e. '  and ;),
instead of being changed to its standard syntax and chtran its syntax is  set to
502 (slashified extended alphabetic) and its chtran is set to itself.

macro     (status macro c) returns nil if c is not a macro character.  If c is a
          macro character it returns a list of the macro character  function and


December 6, 1977                    ∪3-1.8.1                           Page 3-87
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


          the type,  which is nil  for normal macros  and splicing  for splicing
          macros.

          (sstatus macro c f) makes c a macro character which calls the function
          f with no  arguments.  f is evaluated.   A fourth argument  to sstatus
          may be  supplied.  It  is not evaluated.   If it  is an  atomic symbol
          whose pname begins with s, c  is made a splicing macro.  If f  is nil,
          instead of  c being  made a macro-character,  c's macro  abilities are
          taken away  and c becomes  an ordinary  extended-alphabetic character.
          The setsyntax function is generally a better way to do this, however.

+         (status +) gets the value of the + switch (t or nil) in the readtable.
          This switch is normally nil.  If it is t, atomic symbols more than one
          character long beginning with a + or a - are interpreted as numbers by
          the reader even if they contain letters.  This allows the use of input
          bases greater than ten.  See ibase.

          (sstatus + x) sets the + switch  to t or nil depending on x,  which is
          evaluated.  The new value of the + switch is returned.

ttyread   (status  ttyread)  returns the  value  of the  ttyread  switch  in the
          readtable.  At present  this is not used  for anything in  the Multics
          implementation.  In the PDP-10 implementation it controls  whether tty
          "force feed" characters are  used.  See section 13.6.2 for  details of
          force feed characters.

          (sstatus ttyread x) sets the  ttyread switch to t or nil  depending on
          x, which is evaluated.  Again,  file defaults to t.  The new  value of
          the switch is returned.

STATUS FUNCTIONS FOR THE PRINTER

terpri    (status terpri file) returns the value (t or nil) of the terpri switch
          for the file, which defaults  to t.  This switch is normally  nil.  If
          it is t, the  output functions such as  print and tyo will  not output
          any extra newlines  when lines longer than  linel are typed  out.  See
          also the terpri variable.

          (sstatus terpri x file) sets the terpri switch.





Page 3-88                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


←         (status ←  ) returns  the value  (t or nil)  of the  ← switch  for the
          current readtable, which  defaults to t.  If  this switch is t,  the ←
          format for octal fixnums with lots of trailing zeroes is not used.

          (sstatus ← x file) sets the ← switch to the value of x, t or nil.

abbreviate(status  abbreviate) returns  the value  of the  abbreviation control.
          See section 13.7 for a description of the abbreviation  control.  Note
          that the abbreviation control is kept in the current readtable.

          (sstatus abbreviate n) sets the abbreviation control to n.

          (sstatus abbreviate nil) turns off abbreviation.

          (sstatus abbreviate t) turns on a maximal amount of abbreviation.

STATUS FUNCTIONS FOR THE GARBAGE COLLECTOR

gctime    (status  gctime) returns  the  number of  microseconds  spent garbage-
          collecting.

          (sstatus gctime  n) resets  the gctime  counter to  n and  returns the
          previous value of the gctime counter.

          It is a good idea to  do (gc) and then (sstatus gctime 0)  just before
          dumping out a subsystem.  See also (status gcwho).

spcnames  (status  spcnames) returns  a  list of  the  names of  all  the spaces
          available in the LISP being  used.  These are the names  acceptable to
          the alloc function and to the following status functions which require
          a space argument.

spcsize   (status spcsize space) returns  the actual, current size of  space, in
          words.  space is evaluated.

gcmax     (status gcmax space) returns the gcmax parameter for space.

          sstatus gcmax space n) sets the gcmax parameter for space to n.

          See the alloc function.

gcmin     (status gcmin space) returns the gcmin parameter for space.


December 6, 1977                    ∪3-1.8.1                           Page 3-89
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


          sstatus gcmin space n) sets the gcmin parameter for space to n.

          See the alloc function.

gcsize    (status gcsize space) returns the gcsize parameter for space.

          sstatus gcsize space n) sets the gcsize parameter for space to n.

          See the alloc function.

purspcnames       purspcnames) returns a list of names of spaces which have pure
          (status
          versions.  See purcopy.

pursize   (status pursize space) returns the size of the pure version  of space.
          Only spaces returned by (status purspcnames have pure versions.

pdlnames  (status  pdlnames)  returns  a  list of  the  names  of  all  the pdls
          available in the LISP being  used.  These are the names  acceptable to
          the following status functions which require a pdl argument.

pdlsize   (status pdlsize pdl) returns the current number of words on a pdl.

pdlroom   (status pdlroom pdl) returns the "pdlroom" of a pdl, i.e.  the maximum
          size to which it may ever grow.

pdlmax    (status pdlmax pdl) returns the current value of the  pdlmax parameter
          of a pdl.

          (sstatus pdlmax pdl size) sets the pdlmax parameter for the  pdl space
          to size.  Both arguments are evaluated.  See also the alloc function.

   The following status function exists only in the PDP-10 implementation

memfree   (status memfree) returns the number of words of address space  not yet
          allocated  for any  purpose (i.e.  still available  for  allocation to
          various spaces).

ENVIRONMENT ENQUIRIES

Note:   the  various  enquiries related  to  time  may return  nil  in  the rare
circumstance that LISP cannot determine what time it is.



Page 3-90                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


date      (status date) returns a  3-list indicating the current date  as (last-
          two-digits-of-the-year month-number day).

dow       (status dow) returns  an interned atomic symbol  which is the  name of
          the current day of the week.

daytime   (status daytime) returns a 3-list of the 24-hour time of day  as (hour
          minute second).

system    (status system x)  returns a list of  the system properties  (that is,
          properties whose values  are supplied in  the initial LISP  system) of
          the atomic symbol x, which is evaluated.  This list may  contain subr,
          fsubr, lsubr, macro, autoload, or array if x is a system function, and
          value if this atomic symbol is a system variable.

uname     (status uname) returns  an interned atomic  symbol whose pname  is the
          user's login name.  In the ITS implementation this is just  the .UNAME
          .   In   the   Multics   implementation   this   is   in   the  format
          User.Project.Instance; the dot will be slashified if print is  used to
          display this.  In the  TOPS-10 implementation this is actually  a list
          (proj prog) representing  the project-programmer number rather  than a
          symbol.  See also (status xuname).

udir      (status udir) returns  the name of the  user's directory.  In  the ITS
          implementation this is the user's "master sname", which is usually the
          same as the user's name as returned by (status uname).  In the Multics
          implementation this is the  user's default working directory.   In the
          TOPS-10 implementation this is a list (proj prog).

LISPversion
          (status LISPversion) returns the version identification of LISP.  This
          is usually an atomic symbol.

jcl       (status jcl)  returns the "job  command line" from  DDT in  the PDP-10
          implementation.   This  is  used   to  specify  the  init   file,  but
          interpretation of the file name is terminated by altmode, and  (in the
          ITS  implementation at  least) additional  information may  follow the
          altmode.  The init file may wish to examine this information.   In the
          Multics implementation this returns the explodec'd second  argument of
          the LISP command,  or else nil  if the LISP  command did not  have two
          arguments.  If LISP was invoked by

                        LISP environment←name "foo bar"

December 6, 1977                    ∪3-1.8.1                           Page 3-91
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


          then (status jcl) => (f o o /  b a r)

          This function  is also  used by subsystems  implemented in  Maclisp to
          pick up the arguments from the command which invoked them.

   The following status functions exist only in the PDP-10 implementation.

xuname    (status xuname) in  the ITS implementation  returns the .XUNAME  as an
          interned atomic symbol.  The xuname of  a job is the name of  the user
          for whom it is  being run.  This is  normally the same as  .UNAME with
          trailing digits stripped off.  It is possible, however, to cause a job
          to  be run  with a  different xuname.   System  initializations should
          check the xuname to determine which init files to use.  In  the DEC-10
          implementation this is the same as (status uname).

jname     (status jname) returns the job name as an interned atomic symbol.  The
          jname of a job is a  unique identifier for it among all of  the user's
          jobs.  In the  ITS implementation this is  the job's .JNAME .   In the
          DEC-10 implementation this is nnnLISP, where nnn is the job  number in
          decimal.

xjname    (status xjname) in  the ITS implementation  returns the .XJNAME  as an
          interned atomic symbol.  This is normally the name of the program that
          the user invoked to create this job.  Usually this will be the same as
          the jname, but sa user can invoke several copies of a program at once,
          each having  a unique jname  but all having  the same xjname.   In the
          DEC-10 implementation this will  return the job's "program  name".  On
          Multics, it should return (status arg 0).

seglog    (status  seglog) returns  the  logarithm (base  2)  of the  size  of a
          "segment", the unit of space allocation.  (The size of a  segment need
          not be the same as the size of a memory page.)

   The following status functions only exist in the ITS implementation.

its       (status its) returns  a list of five  items (obtained from  the SSTATU
          symbolic system call):

          (1)  The  amount of  time in  seconds until  ITS is  going down,  as a
               flonum,  -1.0 if  ITS doesn't  plan to  go down,  or -2.0  if ITS
               already is down.

          (2)  a fixnum which is nonzero if ITS is being debugged.

Page 3-92                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


          (3)  the number of users logged in, as a fixnum.

          (4)  the number of memory errors the system has survived, as a fixnum.

          (5)  the time in seconds the system has been up, as a flonum.
Some of  this information  may be  of use to  a user  handler for  the sys-death
interrupt.

hactrn    (status hactrn) returns an  atomic symbol indicating what kind  of job
          is superior to  the LISP.  Current possibilities  are ddt, LISP,  t (a
          superior of unknown type), and nil (no superior).

   The following status functions exist only in the Multics implementation.

paging    (status paging)  returns a  list of the  paging-device page  reads and
          total page reads that have been caused by this process.

arg       (status  arg  n) returns  the  n+1'th argument  of  the  command which
          invoked  the  subsystem, as  an  interned atomic  symbol.   (The first
          argument,  (status arg  0),  is the  name  of the  subsystem.)  nil is
          returned if n is greater than the number of arguments to the command.

STATUS FUNCTIONS FOR THE WHO-LINE

   The following status functions exist only in the ITS implementation.

whol      (sstatus whol a  b c d) sets  four user parameters for  the "who-line"
          which appears at the bottom of some display terminals.   These control
          the print format of information specified by (sstatus  who2), (sstatus
          who3), and  (sstatus gcwho).   Each of  the four  parameters specified
          eight bits of information.  Parameters a and c must be fixnums;  b and
          d may be fixnums or atomic symbols (which represent the ascii value of
          the first character in their pnames).  Their meaning is as follows:

          a   If 1, suppress display of entire who-line.
          200
               100 If 1, suppress space between halves of who2.
                70 Mode for printing high 18. bits of who2.
                 7 Mode for printing low 18. bits of who2.

          b   If non-zero, print between halves of who2 as an ascii character.
          177
               200 If 1, print character twice.

          c   If 1, suppress space between who2 and who3.
          200

December 6, 1977                    ∪3-1.8.1                           Page 3-93
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


               177 As for a, but affects who3.

          d   As for b, but affects who3.
          377
          That is,  if the who-line  is displayed at  all, the  user information
          appears in the form:

          PPP##-QQQQ=RRRR@@+SSSS

          where:

          PPPP   result of  printing the high 18.  bits of who2 as  specified by
          is the
               a's 70 bits.

          QQQQ    bits of who2, by c's 7 bits.
          low 18.

          RRRR 18. bits of who3, by c's 70 bits.
          high

          SSSS    bits of who3, by c's 7 bits.
          low 18.

          ##   to two characters, specified by d.
          zero

          -      unless a's 100 bit is set.
          space,

          =      unless c's 200 bit is set.
          space,

          +      unless c's 100 bit is set.
          space,

          The possible print modes are:

          0  not print.
          Do

          1    in packed form:
          Date

                         774000 Year mod 100.

                           3600 Month (January = 1)

                            174 Day of month.
               Printed as mm/dd/yy.

          2    in fortieths of a second.  Printed as hh:mm:ss.t.
          Time

          3    in half-seconds, printed as hh:mm:ss.
          Time

Page 3-94                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


          4     number.
          Octal

          5       number (no "." is printed).
          Decimal

          6     sixbit characters.
          Three

          7
          Unused.
          (status who1) returns a list of four fixnums (a b c d).

who2      (sstatus who2 x) sets the who2 variable to x if it is a fixnum,  or to
          the sixbit value of the first  six characters of the pname of x  if it
          is a symbol.  (status who2 x) returns who2 as a fixnum.

who3      (sstatus who3 x) and (status who3) are analogous to those for who2.

Example:
          (all numbers in octal)
          (sstatus who1 166 0 144 254)
          (sstatus who2 'FOOBAR)
          (sstatus who3 (+ (LSH 1234 22) 3456))
would cause "FOOBAR 1234,,3456" to appear in the who-line.  The general  idea is
that the user should set up who1 just once and then keep updating who2  and who3
as appropriate.  For example, a routing that processed functions in  files might
do
             (sstatus who1 166 0 166 0)
and on opening a new file do
             (sstatus who2 (cadr (truname infile)))
             (sstatus who3 0)
and on encountering a function do
             (sstatus who3 fnname)
and so continuously display the filename and function name being processed.  One
advantage of doing  this over printing the  information is that LISP  can update
who-line information even though it doesn't have control of the console.


gcwho     (sstatus gcwho n) sets the gcwho switch to n.  Currently only  the low
          two bits are significant.

   The  1 bit  means that  during a  garbage collection  the who-line  should be
usurped to  display the message  "GC:XXXX", where "XXXX"  is the reason  for the
garbage collection.  On completion  of the garbage collection the  original user
who-line information is restored.


December 6, 1977                    ∪3-1.8.1                           Page 3-95
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


   The 2  bit means  that at  the end of  the garbage  collection the  who2 word
should be altered so  that the low 18. bits  get the result of  (status gctime),
converted to fortieths of a second, and  the high 18 bits get the result  of (//
(* (status gctime) 100.)(runtime)), which is the percentage of time spent in gc.
If the user gives 52 octal and '% as the first two arguments to  (sstatus who1),
then these quantities will be displaced in the form "nn% hh:mm:ss.t",  just like
the standard system statistics  for the LISP job.   The user can still  use who3
for his own purposes.

   Subsystems which use (sstatus gcwho) should perform

      (gc)
      (sstatus gctime 0)
before  dumping out  the subsystem,  so  that the  gc time  percentages  will be
accurate .

   (status gcwho) returns the value of the gcwho switch.

MISCELLANEOUS STATUS FUNCTIONS

evalhook  (sstatus evalhook t)  enables the evalhook feature;  (sstatus evalhook
          nil) disables it.  (status evalhook) returns the state of  the switch.
          See page 3-32 for the details of this feature.

toplevel  (status  toplevel) returns  the top-level  form, which  is continually
          evaluated when LISP  is at its  top level.  If  this is nil,  a normal
          read-eval-print loop is used.

          (sstatus toplevel x)  evaluates and returns x  and sets the  top level
          form to this value.

          For example, to  make Maclisp have an  evalquote top level  similar to
          LISP 1.5:

               (sstatus toplevel
                        '(progn (print *)
                                (apply (read) (read)) ))

          See section 12.1 for further details.

breaklevel (status breaklevel) returns the break-loop form.  (sstatus breaklevel
          x) sets this form.  See page 3-6 for how this is used.


Page 3-96                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


uuolinks  (status  uuolinks) returns  a number  which represents  the  number of
          available  slots  for  linking  between  compiled  functions.  (status
          uuolinks) returns nil if no  uuolinks pages have been set up  yet (see
          pure).  Otherwise it returns a list  of two items.  The first is  t if
          area 2 has been purified (and so no new calls may be put in the area),
          and nil otherwise.  The second is the number of slots not yet  used in
          the uuolinks areas.

          (sstatus uuolinks) causes all  links between compiled functions  to be
          "unsnapped."  This should be done whenever (nouuo t) is done to insure
          that  the  interpreter  always   gets  a  chance  to   save  debugging
          information on every function call.

divov     (status divov) returns the state of the "divide overflow"  switch.  If
          this switch is nil an attempt  to divide by zero causes an  error.  If
          the switch is t the result of a division by zero is the numerator plus
          1.

          (sstatus divov x) sets the "divide overflow" switch to x.

          In the PDP-10 implementation, divov applies only to quotient.   // and
          //$ do not detect division by zero.

mar       (sstatus  mar  cond  loc)  arms  the  mar  (memory  address  register)
          interrupt  (currently  available  only  in  the  ITS  implementation).
          (status mar) returns a list of cond and loc, or nil if the mar feature
          is not in use.   See the mar break  feature for more details  (page 3-
          56).

features  (status features) returns a list of symbols representing  the features
          implemented in the LISP being used.  The following symbols  may appear
          in this list:

               bibop         PDP-10 big-bag-of-pages memory management scheme

               lap           this LISP has a LISP Assembly Program loaded

               sort          the sorting functions  described in chapter  11 are
                              present

               edit          the  edit  function  described  in  chapter  18  is
                              present


December 6, 1977                    ∪3-1.8.1                           Page 3-97
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


               fasload       the  fasload facility  described in  chapter  14 is
                              present

               ↑f            the "moby I/O" facility is present

               bignum        the   arbitrary-precision  arithmetic   package  is
                              included in this LISP

               hunk          the hunks  data type  and its  associated functions
                              are present.

               funarg        the "fake funarg" feature (second argument to eval,
                              third to apply, etc.) is present.

               strings       character  strings   and  the  functions   on  them
                              described in chapter 8 are present

               newio         the  I/O  functions  described  in  chapter  13 are
                              included  in this  LISP;  if this  feature  is not
                              present   only   some  of   those   functions  are
                              available.

               roman         this LISP  can read and  print roman  numerals (see
                              base and ibase.

               trace         the trace package (chapter 15) is present.

               grindef       the function  definition formatter (chapter  16) is
                              present.

               grind         the file formatter (chapter 16) is present.

               compiler      this is the LISP compiler (chapter 14)  rather than
                              the interpreter (see also compiler-state).

               fastarith     the  fast-arithmetic features  of the  compiler are
                              present

               ml            this LISP is on the MathLab machine at MIT

               ai            this LISP is on the AI machine at MIT

               mc            this LISP is on the MC machine at MIT

Page 3-98                           ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


               SAIL          this LISP is running at SAIL

               H6180         this  LISP  is on  an  H6180 Multics  machine  or a
                              compatible machine such as a 68/60 or a 68/80.

               its           this LISP is on some ITS system

               Multics       this LISP is on some Multics system

               TOPS-10       this LISP is on some DEC TOPS-10 system; or on some
                              TENEX system  since the TENEX  implementation runs
                              under a TOPS-10 emulator.

          A  package being  "present" means  that it  has been  loaded  into the
          environment.  If (status  features) claims it  is not present,  it may
          still  be  available  because  it  may  be  automatically  loaded when
          required.  This does not apply to the compiler.

          (car (last (status features))) is an implementation name, such  as ITS
          or TOPS-10 or Multics.  The main idea behind this status call  is that
          an application package can  be loaded into any  Maclisp implementation
          and  can decide  what to  do on  the basis  of the  features  it finds
          available.

feature   (status  feature  foo) is  roughly  equivalent to  (memq  'foo (status
          features)), i.e. it determines whether this LISP has  the foo-feature.
          Note that foo is not evaluated.

          (sstatus feature foo) makes foo a feature.  foo is not evaluated.  For
          example, the  trace package  does (sstatus feature  trace) when  it is
          loaded.

          Example:











December 6, 1977                    ∪3-1.8.1                           Page 3-99
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**



             (cond ((status feature bignum)
                    (prog2 nil (eval (read)) (read)))  ;use first
                   (t (read) (eval (read)) ))          ;use second

             (defun factorial (n)    ;bignum version
                 (cond ((zerop n) 1)
                       ((times n (factorial (sub1 n))))
                   ))

             (defun factorial (n)    ;fixnum-only version
              (do () ((not (> n 13.)))  ;do until n <λ← 13.
                     (error "argument too big - factorial"
                            n
                            'wrng-type-arg))
                 (cond ((zerop n) 1)
                       ((* n (factorial (1- n)))) ))

nofeature (sstatus  nofeature foo)  makes  foo not  be  a feature.   foo  is not
          evaluated.

          (status nofeature foo) is equivalant to (not (status feature foo)).

status    (status status foo) returns t  if foo is a valid status  function.  If
          it is not, nil is returned.

          (status status) returns a  list of valid status functions.   The names
          are truncated to  some implementation-dependent number  of characters,
          such as 4 or 5.

sstatus   (status sstatus foo) returns t if foo is a valid sstatus function.  If
          it is not, nil is returned.

          (status sstatus) returns a  list of valid sstatus functions.   As with
          (status  status),  the  names are  truncated  to  some implementation-
          dependent number of characters, such as 4 or 5.


1.8.2  Time





Page 3-100                          ∪3-1.8.1                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


runtime             SUBR no args

          (runtime) returns as a fixnum  the number of microseconds of  cpu time
          used so far by the  process in which LISP is running.   The difference
          between two  values of (runtime)  indicates the amount  of computation
          that was done between the two calls to runtime.


time                SUBR no args

          (time) returns the time in seconds  that the system has been up,  as a
          flonum.   The difference  between  the results  of two  calls  to time
          indicates   the   amount  of   elapsed   real  time.    (In   the  ITS
          implementation, time that elapses  while the system is stopped  due to
          memory errors is not  considered "real" and not counted.   Use (status
          daytime) to measure true "real world" time.)


sleep               SUBR 1 arg

          (sleep n) causes  a real-time delay of  n seconds, then returns  n.  n
          may be a fixnum or a flonum.


See also the alarmclock function, section 1.4.3, and the date, daytime,  and dow
functions of the status special form, described in the preceding section.


1.8.3  Escaping from LISP


   It is possible to  escape temporarily from LISP  to execute a command  in the
host  operating system.   Of course,  the program  (or user)  that  supplies the
command has  to know which  operating system  it is running  under.  It  is also
possible for  LISP to  return permanently  to the  host operating  system.  This
discards the LISP environment and gives back whatever resources, such as memory,
it was using.







December 6, 1977                    ∪3-1.8.2                          Page 3-101
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


                         in the Multics implementation

cline               SUBR 1 arg

        (cline x), where x is a character string, executes the Multics command x
     and returns nil.  Example:
                              (cline "who -long")

quit                SUBR no args

        (quit) returns from the LISP command, freeing up the  temporary segments
     that were used to hold the LISP environment.


                           in the ITS implementation

valret              LSUBR 0 or 1 args

        (valret) does a .LOGOUT if LISP is a top level procedure,  and otherwise
     valrets ":VK " to DDT.

        (valret x) effectively performs an explodec on x (in practice x  is some
     strange atomic  symbol like  |:PROCEED :DISOWN  | ,  but it  may be  any S-
     expression).  If the  string of characters is  one of "$↑X.", ":KILL  ", or
     ":KILL↑M"  then  valret performs  a  "silent kill"  by  executing  a .BREAK
     16,20000; otherwise valret performs  a .VALUE, giving the  character string
     to DDT to evaluate as commands.

        Examples:

        (valret '|:PROCEED :DISOWN |)

        starts the LISP running on its own without a terminal.

        (valret '| :KILL :TECO↑M|)

        kills the LISP and starts up a TECO.

        (valret '0$N$P)

        causes DDT to print out  the contents of all non-zero locations  in LISP
     and then return to LISP.


Page 3-102                          ∪3-1.8.3                    December 6, 1977
**DRAFT**                          The System                          **DRAFT**


        If the  valret command  leaves you in  DDT (or  anywhere other  than the
     LISP), a :CONTINUE or $P command to DDT will return to the LISP, which will
     then return from the call to valret.  However, valret differs  from suspend
     in two ways:  it does not check  for open files, and it does not  alter the
     starting address of the LISP.  If you use valret to dump a  subsystem, then
     starting up the subsystem causes  a return to top level and  the consequent
     evaluation of the top-level errlist  (not the binding of errlist  in effect
     at the time of the  valret!).  Using suspend, the subsystem  startup merely
     causes a return from the suspend function.


                         in the TOPS-10 implementation

        There is currently  no way for  LISP to return  a command string  to the
     Monitor  in  the  TOPS-10 implementation.   However,  (valret)  will return
     control to the monitor so that a command may be manually typed.   Then type
     CONTINUE to resume LISP.

        The  remarks  above  about  valret  and  suspend  hold  for  the TOPS-10
     implementation also.


1.8.4  Additional Functions


subr                SUBR 1 arg

        The argument, a fixnum, is taken to be the address of a  location within
     a compiled or system function.  subr attempts to determine the name  of the
     function  by  groveling through  the  current obarray  looking  at  all the
     compiled function properties.  If it fails it returns "?".  This is used by
     baktrace, for example.   This function may be  useful to user  handlers for
     the machine-error interrupt, for example.
          The following function exists only in the ITS implementation

syscall             LSUBR 2 to 10 args

        This function allows the LISP user to directly issue ITS symbolic system
     calls.  The first argument should be
                               (+ (lsh c 18.) n)

        where n is the desired number of output values and c is the control bits


December 6, 1977                    ∪3-1.8.3                          Page 3-103
**DRAFT**                   Maclisp Reference Manual                   **DRAFT**


     for the call.  The second  argument should be an atomic symbol  whose pname
     is the name of  the ITS system call.  The  rest of the arguments  should be
     the input arguments for the call.   An argument may be a fixnum, or  a file
     object  (for  which  the  file's channel  number  is  used).   If  the call
     succeeds, a list  of n fixnums  is returned.  If  an error occurs,  the ITS
     error code is returned as a fixnum.

        Example:   (syscall 0 'scml tyi 3)
             sets the number of echo lines  for the tty to 3 and returns  nil on
             success.


































Page 3-104                          ∪3-1.8.4                    December 6, 1977