Archive-name: lisp-faq/part2
Last-Modified: Thu Nov 5 19:30:40 1992 by Mark Kantrowitz
Version: 1.27
;;; ****************************************************************
;;; Answers to Frequently Asked Questions about Lisp ***************
;;; ****************************************************************
;;; Written by Mark Kantrowitz and Barry Margolin
;;; lisp-faq-2.text -- 28371 bytes
This post contains Part 2 of the Lisp FAQ.
If you think of questions that are appropriate for this FAQ, or would
like to improve an answer, please send email to us at ········@think.com.
Topics Covered (Part 2):
[2-1] Is there a GNU-Emacs interface to Lisp?
[2-3] What is the equivalent of EXPLODE and IMPLODE in Common Lisp?
[2-4] Is Lisp inherently slower than more conventional languages such as C?
[2-5] Why does Common Lisp have "#'"?
[2-6] How do I call non-Lisp functions from Lisp?
[2-7] Can I call Lisp functions from other languages?
[2-8] I want to call a function in a package that might not exist at
compile time. How do I do this?
[2-9] What is CDR-coding?
[2-10] What is garbage collection?
[2-11] How do I save an executable image of my loaded Lisp system?
How do I run a Unix command in my Lisp?
How do I get the current directory name from within a Lisp program?
[2-12] I'm porting some code from a Symbolics Lisp machine to some
other platform, and there are strange characters in the code.
What do they mean?
[2-13] History: Where did Lisp come from?
[2-14] How do I find the argument list of a function?
How do I get the function name from a function object?
[2-15] How can I have two Lisp processes communicate via unix sockets?
Search for [#] to get to question number # quickly.
----------------------------------------------------------------
[2-1] Is there a GNU-Emacs interface to Lisp?
ILISP is a powerful GNU-Emacs interface to many dialects of
Lisp, including Lucid, Allegro, {A}KCL, IBCL, and CMU. Written by
Chris McConnell <····@cs.cmu.edu>. It is available by anonymous
ftp from katmandu.mt.cs.cmu.edu [128.2.250.68] in the directory
pub/ilisp as the file ilisp.tar.Z. If you start using it, send
Chris mail, as he maintains a mailing list of users.
Franz Inc.'s GNU-Emacs/Lisp interface includes an online Common
Lisp manual. It is available by license from Franz Inc. Contact
····@franz.com for more information. There is also a mailing list,
························@ucbarpa.berkeley.edu.
The cl-shell package provides a major mode (cl-shell-mode) for running
Common Lisp (CL) as an Emacs subprocess. It provides a general
mechanism for communication between CL and Emacs which does not rely
on extra processes, and should therefore be easily portable to any
version of CL. Features include direct (i.e., not through a temp file)
evaluation and in-package compilation of forms from lisp-mode buffers,
type-ahead and a history mechanism for the cl-shell buffer, and pop-up
help facilities for the CL functions documentation, macroexpand and
describe. Extensions for Lucid Common Lisp provide pop-up arglists
and source file editing. Other extensions are provided to allow
editing source files of CLOS or Flavors methods. Cl-shell is
available on the Lucid tape (in the goodies directory) or via
anonymous ftp from whitechapel.media.mit.edu (18.85.0.125).
Lucid includes some other Emacs-Lisp interfaces in its goodies directory.
Harlequin's LispWorks includes an Emacs-Lisp interface.
Venue's Medley has an optional EMACS Interface.
GNU-Emacs itself is available by anonymous ftp from prep.ai.mit.edu.
----------------------------------------------------------------
[2-3] What is the equivalent of EXPLODE and IMPLODE in Common Lisp?
Hopefully, the only reason you need to do this is as part of trying to port
some old MacLisp code to Common Lisp. These functions predated the
inclusion of strings as a first-class data type in Lisp; symbols were used
as strings, and they ere EXPLODEd to allow the individual characters to be
manipulated in a list.
Probably the best approximations of these are:
(defun explode (object)
(loop for char across (prin1-to-string object)
collect (intern (string char))))
(defun implode (list)
(read-from-string (coerce (mapcar #'character list) 'string)))
An alternate definition of EXPLODE which uses MAP instead of LOOP is:
(defun explode (object)
(map 'list #'(lambda (char)
(intern (string char)))
(prin1-to-string object)))
The creation of N conses of garbage to process a string of N
characters is a hideously inefficient way of doing the job. Rewrite
EXPLODE code with PRIN1-TO-STRING, or better STRING if the arguments
are symbols without funny characters. For IMPLODE, try to make its
caller use strings and try to make the result usable as a string to
avoid having to call INTERN or READ-FROM-STRING.
----------------------------------------------------------------
[2-4] Is Lisp inherently slower than more conventional languages such as C?
This is a tough question to answer, as I'm sure you expected. In many
cases, it appears to be. Lisp does not require the programmer to specify
the data type of variables, so generic arithmetic operators may have to
perform type checking at runtime in order to determine how to proceed.
However, Lisp code can also be denser (i.e. there is more expressed in a
single line) than many other languages: the Lisp expression (+ A B) is more
powerful than the C expression A+B (the Lisp version supports bignums,
rationals, and complex numbers, while the C version only supports
limited-size integers and floating point); therefore, one may claim that it
is reasonable that the Lisp version take longer than the C version (but
don't expect everyone to accept this rationalization). Solutions to this
include hardware support (e.g. processors that support type tags in data,
such as SPARC and Symbolics Lisp Machines), declarations, and specialized
variants of functions (e.g. in MacLisp, + accepts and returns only fixnums,
+$ accepts and returns only flonums, and PLUS is generic).
At one time, the MIT PDP-10 MacLisp compiler was compared to DEC's PDP-10
Fortran compiler. When appropriate declarations were supplied in the Lisp
code, the performance of compiled Lisp arithmetic rivaled that of the
Fortran code. It would hardly be fair to compare Lisp without declarations
to Fortran, since the Fortran compiler would have more information upon
which it could base its optimizations.
Since Lisp is a good language for rapid prototyping, it is easy for a
mediocre programmer (or even a good programmer, who isn't being careful) to
generate a large amount of inefficient Lisp code. A good example is the use
of APPEND to link successive lists together, instead of keeping a pointer
to the tail of the list. Often a programmer can obtain significant
speed increases by using a time/space profiler to identify the
functions which waste time (often small functions which are called
frequently) and rewriting those functions.
----------------------------------------------------------------
[2-5] Why does Common Lisp have "#'"?
#' is a macro-character which expands #'FOO to (FUNCTION FOO). Symbols in
Lisp have two bindings, one for values and one for functions, allowing them
to represent both variables and functions, depending on context. #'FOO
accesses FOO's lexical function binding in a context where the value
interpretation would normally occur. #' is also used to create lexical
closures for lambda expressions. A lexical closure is a function which when
invoked executes the body of the lambda-expression in the lexical
environment within which the closure was created. See pp. 115-117 of CLtL2
for more details.
----------------------------------------------------------------
[2-6] How do I call non-Lisp functions from Lisp?
Most Lisp implementations for systems where Lisp is not the most common
language provide a "foreign function" interface. As of now there has been
no significant standardization effort in this area. They tend to be
similar, but there are enough differences that it would be inappropriate to
try to describe them all here. In general, one uses an
implementation-dependent macro that defines a Lisp function, but instead of
supplying a body for the function, one supplies the name of a function written
in another language; the argument list portion of the definition is
generally augmented with the data types the foreign function expects and
the data type of the foreign function's return value, and the Lisp
interface function arranges to do any necessary conversions. There is also
generally a function to "load" an object file or library compiled in a
foreign language, which dynamically links the functions in the file being
loaded into the address space of the Lisp process, and connects the
interface functions to the corresponding foreign functions.
If you need to do this, see the manual for your language implementation for
full details. In particular, be on the lookout for restrictions on the
data types that may be passed. You may also need to know details about the
linkage conventions that are used on your system; for instance, many C
implementations prepend an underscore onto the names of C functions when
generating the assembler output (this allows them to use names without
initial underscores internally as labels without worrying about conflicts),
and the foreign function interface may require you to specify this form
explicitly.
Franz Allegro Common Lisp's "Foreign Function Call Facility" is
described in chapter 10 of the documentation. Calling Lisp Functions
from C is treated in section 10.8.2. The foreign function interface in
Macintosh Common Lisp is similar. The foreign function interface for
KCL is described in chapter 10 of the KCL Report. The foreign function
interfaces for Lucid on the Vax and Lucid on the Sun4 are
incompatible. Lucid's interface is described in chapter 5 of the
Advanced User's Guide.
----------------------------------------------------------------
[2-7] Can I call Lisp functions from other languages?
In implementations that provide a foreign function interface as described
above, there is also usually a "callback" mechanism. The programmer may
associate a foreign language function name with a Lisp function. When a
foreign object file or library is loaded into the Lisp address space, it is
linked with these callback functions. As with foreign functions, the
programmer must supply the argument and result data types so that Lisp may
perform conversions at the interface. Note that in such foreign function
interfaces Lisp is often left "in control" of things like memory
allocation, I/O channels, and startup code (this is a major nuisance
for lots of people).
----------------------------------------------------------------
[2-8] I want to call a function in a package that might not exist at
compile time. How do I do this?
Use (funcall (find-symbol "SYMBOL-NAME" :pkg-name) ...).
----------------------------------------------------------------
[2-9] What is CDR-coding?
CDR-coding is a space-saving way to store lists in memory. It is normally
only used in Lisp implementations that run on processors that are
specialized for Lisp, as it is difficult to implement efficiently
in software. In normal list structure, each element of the
list is represented as a CONS cell, which is basically two pointers (the
CAR and CDR); the CAR points to the element of the list, while the CDR
points to the next CONS cell in the list or NIL. CDR-coding takes
advantage of the fact that most CDR cells point to another CONS, and
further that the entire list is often allocated at once (e.g. by a call to
LIST). Instead of using two pointers to implement each CONS cell, the CAR
cell contains a pointer and a two-bit "CDR code". The CDR code may contain
one of three values: CDR-NORMAL, CDR-NEXT, and CDR-NIL. If the code is
CDR-NORMAL, this cell is the first half of an ordinary CONS cell pair, and
the next cell in memory contains the CDR pointer as described above. If
the CDR code is CDR-NEXT, the next cell in memory contains the next CAR
cell; in other words, the CDR pointer is implicitly thisaddress+1, where
thisaddress is the memory address of the CAR cell. If the CDR code is
CDR-NIL, then this cell is the last element of the list; the CDR pointer is
implicitly a reference to the object NIL. When a list is constructed
incrementally using CONS, a chain of ordinary pairs is created; however,
when a list is constructed in one step using LIST or MAKE-LIST, a block of
memory can be allocated for all the CAR cells, and their CDR codes all set
to CDR-NEXT (except the last, which is CDR-NIL), and the list will only
take half as much storage (because all the CDR pointers are implicit).
If this were all there were to it, it would not be difficult to implement
in software on ordinary processors; it would add a small amount of overhead
to the CDR function, but the reduction in paging might make up for it. The
problem arises when a program uses RPLACD on a CONS cell that has a CDR
code of CDR-NEXT or CDR-NIL. Normally RPLACD simply stores into the CDR
cell of a CONS, but in this case there is no CDR cell -- its contents are
implicitly specified by the CDR code, and the word that would normally
contain the CDR pointer contains the next CONS cell (in the CDR-NEXT case)
to which other data structures may have pointers, or the first word of some
other object (in the CDR-NIL case). When CDR-coding is used, the
implementation must also provide automatic "forwarding pointers"; an
ordinary CONS cell is allocated, the CAR of the original cell is copied
into its CAR, the value being RPLACD'ed is stored into its CDR, and the old
CAR cell is replaced with a forwarding pointer to the new CONS cell.
Whenever CAR or CDR is performed on a CONS, it must check whether the
location contains a forwarding pointer. This overhead on both CAR and CDR,
coupled with the overhead on CDR to check for CDR codes, is generally
enough that using CDR codes on conventional hardware is infeasible.
There is some evidence that CDR-coding doesn't really save very much
memory, because most lists aren't constructed at once, or RPLACD is done on
them enough that they don't stay contiguous. At best this technique can
save 50% of the space occupied by CONS cells. However, the savings probably
depends to some extent upon the amount of support the implementation
provides for creating CDR-coded lists. For instance, many system functions
on Symbolics Lisp Machines that operate on lists have a :LOCALIZE option;
when :LOCALIZE T is specified, the list is first modified and then copied
to a new, CDR-coded block, with all the old cells replaced with forwarding
pointers. The next time the garbage collector runs, all the forwarding
pointers will be spliced out. Thus, at a cost of a temporary increase in
memory usage, overall memory usage is generally reduced because more lists
may be CDR-coded. There may also be some benefit in improved paging
performance due to increased locality as well (putting a list into
CDR-coded form makes all the "cells" contiguous). Nevertheless, modern
Lisps tend to use lists much less frequently, with a much heavier
reliance upon code, strings, and vectors (structures).
----------------------------------------------------------------
[2-10] What is garbage collection?
Garbage Collection (GC) refers to the automatic storage allocation
mechanisms present in many Lisps. There are several kinds of storage
allocation algorithms, but most fall within two main classes:
1. Stop and Copy. Systems which copy active objects from "old"
storage to "new" storage and then recycle the old storage.
2. Mark and Sweep. Systems which link together storage
used by discarded objects.
Generational scavenging garbage collection (aka emphemeral GC) is a
variation in which memory is allocated in layers, with tenured
(long-lived) objects in the older layers. Rather than doing a full GC
of all of memory every time more room is needed, only the last few
layers are GCed during an ephemeral GC, taking much less time.
Short-lived objects are quickly recycled, and full GCs are then much
less frequent. It is most often used to improve the performance of
stop and copy garbage collectors. It is possible to implement
ephemeral GC in mark and sweep systems, just much more difficult.
Stop and copy garbage collection provides simpler storage allocation,
avoids fragmentation of memory (intermixing of free storage with used
storage). Copying, however, consumes more of the address space, since up to
half the space must be kept available for copying all the active objects.
This makes stop and copy GC impractical for systems with a small address
space or without virtual memory. Also, copying an object requires that you
track down all the pointers to an object and update them to reflect the new
address, while in a non-copying system you need only keep one pointer to an
object, since its location will not change. It is also more difficult to
explicitly return storage to free space in a copying system.
Garbage collection is not part of the Common Lisp standard. Most Lisps
provide a function ROOM which provides human-readable information about the
state of storage usage. In many Lisps, (gc) invokes an ephemeral garbage
collection, and (gc t) a full garbage collection.
----------------------------------------------------------------
[2-11] How do I save an executable image of my loaded Lisp system?
How do I run a Unix command in my Lisp?
How do I get the current directory name from within a Lisp program?
There is no standard for dumping a Lisp image. Here are the
commands from some lisp implementations:
Lucid: DISKSAVE
Symbolics: Save World [CP command]
CMU CL: SAVE-LISP
Franz Allegro: EXCL:DUMPLISP (documented)
SAVE-IMAGE (undocumented)
Medley: IL:SYSOUT or IL:MAKESYS
MCL: SAVE-APPLICATION <pathname>
&key :toplevel-function :creator :excise-compiler
:size :resources :init-file :clear-clos-caches
KCL: (si:save-system "saved_kcl")
There is no standard for running a Unix shell command from Lisp,
especially since not all Lisps run on top of Unix. Here are the
commands from some Lisp implementations:
Allegro: EXCL:RUN-SHELL-COMMAND
Lucid: RUN-PROGRAM (name
&key input output
error-output (wait t) arguments
(if-input-does-not-exist :error)
(if-output-exists :error)
(if-error-output-exists :error))
KCL: SYSTEM
For example, (system "ls -l").
You can also try RUN-PROCESS and EXCLP, but they
don't work with all versions of KCL.
There's no standard function for finding the current directory from
within a Lisp program, since not all Lisp environments have the
concept of a current directory. Here are the commands from some Lisp
implementations:
Lucid: working-directory (which is also SETFable)
pwd and cd also work
Allegro: current-directory (use excl:chdir to change it)
CMU CL: default-directory
Allegro also uses the variable *default-pathname-defaults* to resolve
relative pathnames, maintaining it as the current working directory.
So evaluating (truename "./") in Allegro (and on certain other
systems) will return a pathname for the current directory. Likewise,
in some VMS systems evaluating (truename "[]") will return a pathname
for the current directory.
----------------------------------------------------------------
[2-12] I'm porting some code from a Symbolics Lisp machine to some
other platform, and there are strange characters in the code.
What do they mean?
The Symbolics Zetalisp character set includes the following
characters not present in other Lisps:
^] >= greater than or equal to
^\ <= less than or equal to
^Z != not equal to
^^ == equivalent to
^E not
^G pi
^L +/- plus/minus
^H lambda
^F epsilon
^W <--> left/right arrow
^X <-- left arrow
^Y --> right arrow
Other special characters to look out for are the font-change characters,
which are represented as a ^F followed by a digit or asterisk. A digit
means to push font #N onto the stack; an asterisk means to pop the most
recent font from the stack. You can clean up the code by replacing "\^F."
with "".
----------------------------------------------------------------
[2-13] History: Where did Lisp come from?
John McCarthy developed the basics behind Lisp during the 1956 Dartmouth
Summer Research Project on Artificial Intelligence. He intended it as an
algebraic LISt Processing (hence the name) language for artificial
intelligence work. Early implementations included the IBM 704, the IBM
7090, the DEC PDP-1, the DEC PDP-6 and the DEC PDP-10. The PDP-6 and
PDP-10 had 18-bit addresses and 36-bit words, allowing a CONS cell to
be stored in one word, with single instructions to extract the CAR and
CDR parts. The early PDP machines had a small address space, which
limited the size of Lisp programs.
Milestones in the development of Lisp:
1956 Dartmouth Summer Research Project on AI.
1960-65 Lisp1.5 is the primary dialect of Lisp.
1964- Development of BBNLisp at BBN.
late 60s Lisp1.5 diverges into two main dialects:
Interlisp (originally BBNLisp) and MacLisp.
early 70s Development of special-purpose computers known as Lisp
Machines, designed specificly to run Lisp programs.
Xerox D-series Lisp Machines run Interlisp-D.
Early MIT Lisp Machines run Lisp Machine Lisp
(an extension of MacLisp).
1969 Anthony Hearn and Martin Griss define Standard Lisp to
port REDUCE, a symbolic algebra system, to a variety
of architectures.
late 70s Macsyma group at MIT developed NIL (New Implementation
of Lisp), a Lisp for the VAX.
Stanford and Lawrence Livermore National Laboratory
develop S-1 Lisp for the Mark IIA supercomputer.
Franz Lisp (dialect of MacLisp) runs on stock-hardware
Unix machines.
Gerald J. Sussman and Guy L. Steele developed Scheme,
a simple dialect of Lisp with lexical scoping and
lexical closures, continuations as first-class objects,
and a simplified syntax (i.e., only one binding per symbol).
Advent of object-oriented programming concepts in Lisp.
Flavors was developed at MIT for the Lisp machine,
and LOOPS (Lisp Object Oriented Programming System) was
developed at Xerox.
early 80s Development of SPICE-Lisp at CMU, a dialect of MacLisp
designed to run on the Scientific Personal Integrated
Computing Environment (SPICE) workstation.
1980 First biannual ACM Lisp and Functional Programming Conf.
1981 PSL (Portable Standard Lisp) runs on a variety of platforms.
1981+ Lisp Machines from Xerox, LMI (Lisp Machines Inc)
and Symbolics available commercially.
April 1981 Grass roots definition of Common Lisp as a description
of the common aspects of the family of languages (Lisp
Machine Lisp, MacLisp, NIL, S-1 Lisp, Spice Lisp, Scheme).
1984 Publication of CLtL1. Common Lisp becomes a de facto
standard.
1986 X3J13 forms to produce a draft for an ANSI Common Lisp
standard.
1987 Lisp Pointers commences publication.
1990 Steele publishes CLtL2 which offers a snapshot of
work in progress by X3J13. (Unlike CLtL1, CLtL2
was NOT an output of the standards process and was
not intended to become a de facto standard. Read
the Second Edition Preface for further explanation
of this important issue.) Includes CLOS,
conditions, pretty printing and iteration facilities.
1992 X3J13 creates a draft proposed American National
Standard for Common Lisp. This document is the
first official successor to CLtL1.
[Note: This summary is based primarily upon the History section of the
draft ANSI specification. More detail and references can be obtained from
that document. See [1-5] for information on obtaining a copy.]
----------------------------------------------------------------
[2-14] How do I find the argument list of a function?
How do I get the function name from a function object?
There is no standard way to find the argument list of a function,
since implementations are not required to save this information.
However, many implementations do remember argument information, and
usually have a function that returns the lambda list. Here are the
commands from some Lisp implementations:
Lucid: arglist
Allegro: excl::arglist
Symbolics: arglist
CMU Common Lisp, new compiler:
#+(and :CMU :new-compiler)
(defun arglist (name)
(let* ((function (symbol-function name))
(stype (system:%primitive get-vector-subtype function)))
(when (eql stype system:%function-entry-subtype)
(cadr (system:%primitive header-ref function
system:%function-entry-type-slot)))))
If you're interested in the number of required arguments you could use
(defun required-arguments (name)
(or (position-if #'(lambda (x) (member x lambda-list-keywords))
(arglist name))
(length (arglist name))))
To extract the function name from the function object, as in
(function-name #'car) ==> 'car
use the following vendor-dependent functions:
Symbolics: (si::compiled-function-name <fn>)
Lucid: (sys::procedure-ref <fn> SYS:PROCEDURE-SYMBOL)
Allegro: (Xref::object-to-function-name <fn>)
CMU CL: (kernel:%function-header-name <fn>)
AKCL: (system::compiled-function-name <fn>)
MCL: (ccl::function-name <fn>)
----------------------------------------------------------------
[2-15] How can I have two Lisp processes communicate via unix sockets?
CLX uses Unix sockets to communicate with the X window server. Look at
the following files from the CLX distribution for a good example of
using Unix sockets from Lisp:
defsystem.lisp Lucid, AKCL, IBCL, CMU.
socket.c, sockcl.lisp AKCL, IBCL
excldep.lisp Franz Allegro CL
You will need the "socket.o" files which come with Lucid and Allegro.
To obtain CLX, see the entry for CLX in the answer to question [6-5].
See the file lisp-sockets.text in the Lisp Utilities repository
described in the answer to question [6-1].
----------------------------------------------------------------
;;; *EOF*