NAME
Emacs::Lisp - Support for Perl embedded in GNU Emacs
SYNOPSIS
In Emacs
M-x perl-eval-expression RET 2+2 RET
M-x perl-eval-region RET
M-x perl-eval-buffer RET
... and more ...
In Perl
use Emacs::Lisp;
&switch_to_buffer('*scratch*');
&insert("Hello, world!\n");
setq { $cperl_font_lock = t };
&add_hook(\*find_file_hooks,
sub { &message("found a file!") });
use Emacs::Lisp qw($emacs_version $perlmacs_version);
save_excursion {
&set_buffer(&get_buffer_create("*test*"));
&insert("This is Emacs version $emacs_version,\n");
&insert("Perlmacs version $perlmacs_version.\n");
&insert("Emacs::Lisp version is $Emacs::Lisp::VERSION.\n");
};
DESCRIPTION
Emacs allows you to customize your environment using Lisp. With Perlmacs, you can use Perl, too. This module allows Perl code to call functions and access variables of Lisp.
You still need to learn some Lisp in order to understand The Elisp Manual, which is the definitive reference for Emacs programming. This document assumes a basic understanding of Emacs commands and Lisp data types. I also assume familiarity with Perl's complex data structures (described in perlref) and objects (see perlobj).
Quick Start
Paste this code into a buffer, select it, and type M-x perl-eval-region RET
:
use Emacs::Lisp;
&message ("hello!");
The string "hello!"
should appear in the minibuffer.
You can define a command like this:
use Emacs::Lisp;
sub doit { &insert("Cool, huh?\n"); }
defun (\*perltest, interactive, \&doit);
After evalling this, type M-x perltest RET
. The text will be inserted into the current buffer at point.
LISP SUPPORT FOR PERL
Perlmacs comes in the form of a patch of Emacs. The patch adds some Perl support to the emacs executable and installs it as pmacs. The Emacs::Lisp module extends the core support by wrapping it in a Perl-like interface.
Lisp code can check for Perl support using (featurep 'perl)
.
Functions
The following Lisp functions are part of Perlmacs proper and do not rely on the Emacs::Lisp module. Use C-h f <function-name> RET
within pmacs to see their doc strings.
perl-eval-expression EXPRESSION
perl-eval-region START END
perl-eval-buffer
perl-load-file NAME
perl-eval STRING &optional CONTEXT
perl-call SUB &optional CONTEXT &rest ARGS
perl-eval-and-call STRING &optional CONTEXT &rest ARGS
perl-to-lisp OBJECT
perl-wrap OBJECT
perl-value-p OBJECT
perl-eval-raw STRING &optional CONTEXT
perl-call-raw SUB &optional CONTEXT &rest ARGS
make-perl-interpreter &rest ARGV
perl-run &optional INTERPRETER
perl-destruct &optional INTERPRETER
Two Lisp variables affect the Perl interpreter and have doc strings accessible via C-h f <variable-name> RET
. They are:
perl-interpreter-args
perl-interpreter
Data Conversions
When Perl calls a Lisp function, its arguments are converted to Lisp objects, and the returned object is converted to a Perl value. Likewise, when Lisp calls Perl, the arguments are converted from Lisp to Perl and the return values are converted to Lisp.
Lisp has three scalar types.
Lisp integers, floats, and strings all become Perl scalars. A simple Perl scalar becomes either an integer, a float, or a string.
Interesting character encodings such as UTF-8 are not currently supported. I don't even know what happens to 8-bit characters during string conversion.
Lisp symbols correspond to globrefs.
Glob references in package
main
become symbols in Lisp. Underscores are swapped with hyphens in the name, since Perl prefers underscores and Lisp prefers hyphens. See "Symbols" for more information.Lisp's `nil' is equivalent to Perl's `undef' or `()'.
As an exception to the rule for symbols,
nil
in Lisp corresponds toundef
in Perl.In Lisp,
nil
is really a symbol. However, it is typically used as the boolean value false. Glob references evaluate to true in boolean context. It is much more natural to convertnil
toundef
.Arrayrefs correspond to lists.
Lists are a central data structure in Lisp. To make it as easy as possible to pass lists to Lisp functions that require them, Perl array references are converted Lisp lists. For example, the Perl expression such as
["x", ["y", 1]]
is converted to
'("x" ("y" 1))
in Lisp.
Arrayref refs correspond to vectors.
Adding
\
to an arrayref makea it an arrayref ref, which becomes a vector in Lisp. For example,\[1, 2, undef]
becomes[1 2 nil]
.Conversion uses "deep" copying by default.
Conversion of lists and vectors to arrayrefs and arrayref refs is recursive by default. Changes made by Lisp to a list will not affect the Perl array of which it is a copy, nor will changes to a Perl array affect a Lisp list. See "BUGS" about converting cyclic structures.
There are ways to make "shallow" copies.
A shallow copy simply wraps a Perl scalar in a Lisp object or vice versa. Wrapped Perl values appear as a Lisp objects of type
perl-value
. Wrapped Lisp values appear in Perl as objects of classEmacs::Lisp::Object
. See "CAVEATS" for issues relating to wrapped data.Where a data type has no natural equivalent in the other language, shallow copying is the default. Examples include Perl hashrefs and Lisp buffer objects.
In Perl, the
lisp
function wraps its argument in a Lisp object. This allows Perl arrays to be passed by reference to Lisp functions. (Of course, the value returned bylisp
is really a Perl value wrapped in a Lisp object wrapped in a Perl object.)An Emacs::Lisp::Object's
to_perl
method performs a deep copy (if the argument is Lisp data) or unwraps its argument (if it is Perl data).Lisp functions called through package Emacs::Lisp convert their return values using deep copying. The same functions are accessible through Emacs::Lisp::Object, which does shallow conversion and always returns Emacs::Lisp::Object objects.
These examples show how the data wrapping functions work:
$x = lisp [1, 2, 3]; print ref($x); # "Emacs::Lisp::Object" print ref($x->to_perl); # "ARRAY" print @{&list(2, 3)}; # "23" $x = Emacs::Lisp::Object::list(2, 3); print ref($x); # "Emacs::Lisp::Object" print @{$x->to_perl}; # "23"
Scripts
Perlmacs can run Perl programs. By default, Perlmacs is installed under two names, pmacs and perlmacs. Which name is used to invoke the program determines how it parses its command line.
If perlmacs is used (or, more precisely, any name containing "perl"), it behaves like Perl. For example,
$ perlmacs script.pl
runs the Perl program script.pl.
When invoked as pmacs, it behaves like Emacs. Example:
$ pmacs file.txt
This begins an editing session with file.txt in the current buffer.
The first command line argument can override the invocation name. If it is --emacs, Emacs takes control. If it is --perl, the program runs in Perl mode.
The Emacs module (that is, the Perl module named "Emacs") includes support for starting an editing session from within a Perlmacs script. See Emacs.
PERL SUPPORT FOR LISP
The Emacs::Lisp module allows Perl programs to invoke Lisp functions and handle Lisp variables as if they were Perl subs and variables.
The directive use Emacs::Lisp;
causes any use of a function not defined in Perl to invoke the Lisp function of the same name (with hyphens in place of underscores). For example, this writes a message to the standard error stream (in Perl mode) or displays it in the minibuffer:
&message ("this is a test");
Functions
This code calls the hypothetical Lisp function foo-bar
with arguments 4
and t
.
&foo_bar(4, t);
The Lisp syntax for the same call would be
(foo-bar 4 t)
The ampersand (&
) in the Perl example is not required, but it is needed for functions, such as read
, eval
, and print
, which are Perl keywords. Using it with Emacs::Lisp is a good habit, so the examples in this document include it.
If you don't want an AUTOLOAD
sub to affect your namespace, you may either put parentheses after "use Emacs::Lisp
" or import to a different package, and use qualified function names. For example:
use Emacs::Lisp ();
Emacs::Lisp::message("hello\n");
{package L; use Emacs::Lisp;}
L::message("goodbye\n");
Symbols
Many Lisp functions take arguments that may be, or are required to be, symbols. In Lisp, a symbol is a kind of name, but does not have the same type as a string. Lisp programs typically use the quote
operator to specify a symbol. For example, this Lisp code refers to the beep
symbol:
(run-at-time nil 1 'beep)
Perlmacs uses glob references of package main to specify symbols. A literal globref begins with a backslash followed by an asterisk, so the last example would be written as
&run_at_time(undef, 1, \*beep);
in Perl. (You may want to do &cancel_function_timers(\*beep)
soon after trying this example.)
Only globs from package main may be used as Lisp symbols, so code that is compiled in another package must use the form \*::sym
rather than \*sym
.
When comparing the returned values of Lisp functions to each other and to symbols, it is best to use the Lisp eq
function instead of Perl's equality operators.
### PREFERRED
if (&eq(&type_of($x), \*::cons)) { ... }
### PROBABLY OK
if (&type_of($x) eq \*cons) { ... }
if (&type_of($x) == \*cons) { ... }
Variables
In Lisp, variables play a role akin to that of Perl scalar variables. A variable may hold a number, a string, or a reference to any type of complex Lisp data structure. (They are not called references in Lisp, but rather "objects".)
You can create a Perl alias for any reasonably named Lisp variable by saying use Emacs::Lisp qw($varname);
. Thereafter, assignment to $varname
will update the Lisp value. Changes made to the variable in Lisp will be reflected in Perl when $varname
is used in expressions.
This example saves and replaces the value of the Lisp variable inhibit-eol-conversion
:
use Emacs::Lisp qw($inhibit_eol_conversion);
$old_val = $inhibit_eol_conversion;
$inhibit_eol_conversion = 1;
This sort of thing could be accomplished in Lisp as follows:
(setq old-val inhibit-eol-conversion)
(setq inhibit-eol-conversion 1)
(but you would probably rather use let
instead, for which there is still no convenient Emacs::Lisp equivalent). See also the setq
function below.
Property Lists
Lisp symbols all have an associated object called a plist, for "property list". The plist is an object just like any other, but it is typically used in a way vaguely resembling Perl's hashes.
Plists are not used nearly as often as Lisp functions and variables. If you are new to Lisp, you can probably skip this section.
A plist is different from a Perl hash. Lookups are not based on string equality as with Perl, but rather on Lisp object equality of the eq
variety. For this reason, it is best to stick to the Lisp convention of using only symbols as keys. (See "Symbols".)
Emacs::Lisp provides a shorthand notation for getting and setting plist elements. If you say "use Emacs::Lisp qw(%any_name)
", then subsequent access to the elements of %any_name
will get or set the corresponding properties of the Lisp symbol any-name
.
For example, the following Perl and Lisp fragments are more or less equivalent:
# Perl fragment
use Emacs::Lisp qw(%booboo %upcase_region);
$booboo{\*error_conditions} = [\*booboo, \*error];
$can_upcase = ! $upcase_region{\*disabled};
; Lisp fragment
(put 'booboo 'error-conditions '(booboo error))
(setq can-upcase (not (get 'upcase-region 'disabled)))
See also the setq
function below.
Macros
Lisp macros, such as setq
and defun
, do not work the same way functions do, although they are invoked using the function syntax. (Here you see the vast philosophical chasm separating Perl from Lisp. While Perl might have five syntaxes to mean the same thing, Lisp has one syntax with two meanings!)
Some macros are equivalent to Perl operators, such as if
and while
. Others have meanings peculiar to Lisp. A few macros are implemented in Emacs::Lisp. They are described below. If you try to call a macro that has not been implemented, you will get an error message which may propose an alternative.
- catch SYMBOL,CODE
-
Evaluate CODE in a Lisp
catch
construct. At any point during CODE's execution, thethrow
function may be used to return control to the end of thecatch
block. For example:$x = catch \*::out, sub { $y = 1; &throw(\*::out, 16); $y = 2; }; print $x; # prints 16 print $y; # prints 1
Some Perl constructs have functionality similar to
throw
; for example,return
andlast
. However, they do not work with catches in Lisp code. - defun SYMBOL,DOCSTRING,SPEC,CODE
- defun SYMBOL,DOCSTRING,CODE
- defun SYMBOL,SPEC,CODE
- defun SYMBOL,CODE
-
Make CODE callable as the Lisp function SYMBOL. This is Lisp's version of Perl's
sub
keyword. A function defined in this way becomes visible to Lisp code.defun
is useful for defining Emacs commands. Commands are functions that the user can invoke by typingM-x <function-name>
. A command may be bound to a key or sequence of keystrokes. See the Emacs documentation for specifics.When defining a command, you must specify the interactive nature of the command. There are various codes to indicate that the command acts on the current region, a file name to be read from the minibuffer, etc. Please see The Elisp Manual for details.
Emacs::Lisp's
defun
uses a SPEC returned by theinteractive
function to specify a command's interactivity. If no SPEC is given, the function will still be callable by Lisp, but it will not be available to the user viaM-x <function-name> RET
and cannot be bound to a sequence of keystrokes. Even commands that do not request information from the user need an interactive spec. See "interactive".This example creates a command,
reverse-region-words
, that replaces a region of text with the same text after reversing the order of words. To be user-friendly, we'll provide a documentation string, which will be accessible through the Emacs help system (C-h f reverse-region-words RET
).use Emacs::Lisp; defun (\*reverse_region_words, "Reverse the order of the words in the region.", interactive("r"), sub { my ($start, $end) = @_; my $text = &buffer_substring($start, $end); $text = join('', reverse split (/(\s+)/, $text)); &delete_region($start, $end); &insert($text); });
If you try this example and invoke the help system, you may notice something not quite right in the message. It reads as follows:
reverse-region-words is an interactive Lisp function. (reverse-region-words &optional START END &rest ARGS) Reverse the order of the words in the region.
Notice the part about "&optional" and "&rest". This means that Lisp thinks the function accepts any number of arguments. It knows the names of the first two because of the assignment "
my ($start, $end) = @_
".But our function only works if it receives two args. Specifying a prototype documents this:
sub ($$) { my ($start, $end) = @_; ... } reverse-region-words is an interactive Lisp function. (reverse-region-words START END)
- interactive SPEC
- interactive
-
Used to generate the third (or, in the absence of a doc string, the second) argument to
defun
. This determines how a command's arguments are obtained.What distinguishes a "command" from an ordinary function in Emacs is the presence of an
interactive
specifier in thedefun
expression.SPEC may be a string, as described in The Elisp Manual, or a reference to code which returns the argument list. If no spec is given, the command runs without user input.
- save_excursion BLOCK
-
Execute BLOCK within a Lisp
save-excursion
construct. This restores the current buffer and other settings to their original values after the code has completed. See The Elisp Manual for details. - setq BLOCK
-
BLOCK is searched for assignments of either of these forms:
$var = EXPR; $hash{$key} = EXPR;
Every such
$var
and%hash
is imported from the Emacs::Lisp module as if you had said, "use Emacs::Lisp qw($var %hash)
". Afterwards, BLOCK is executed. This is a convenient way to assign to variables, for example in customization code.This code
use Emacs::Lisp; setq { $A = 2*$foo[5]; $B{\*foo} = "more than $A"; };
would have exactly the same effect as this:
use Emacs::Lisp qw(:DEFAULT $A %B); $A = 2*$foo[5]; $B{\*foo} = "more than $A";
The following, which does not tie or import any variables, has the same effect on Lisp as the above:
use Emacs::Lisp (); Emacs::Lisp::set( \*A, 2*$foo[5] ); Emacs::Lisp::put( \*B, \*foo, "more than " . &Emacs::Lisp::symbol_value( \*A ));
BUGS
These are some of the known bugs in Perlmacs and Emacs::Lisp. See also the file BUGS in the Perlmacs distribution. If you find other bugs, please check that you have the latest version, and email me.
Deep copy of cyclic data crashes.
The default data conversion operation does not check for recursive structures and will run out of memory if function args or return values reference parts of themselves. I don't think this will be too hard to fix.
Within Lisp code, everything defaults to package `main'.
It would perhaps be best to give the Lisp evaluation environment the notion of a "current package" such as Perl has.
Perl's `local()' doesn't have the effect of Lisp's `let'.
It should. At least, there should be an easy way to make a local binding of a Lisp variable in Perl.
A crash is likely if Perl code modifies the scalar value in an `Emacs::Lisp::Object' blessed reference or explicity calls DESTROY on it.
Don't do that.
CAVEATS
Circular data structures are bad.
See "Two-Phased Garbage Collection" in perlobj. Lisp data structures may be recursive (contain references to themselves) without the danger of a memory leak, because Lisp uses a periodic-mark-and-sweep garbage collector.
However, if a recursive structure involves any Perl references, it may never be destroyable.
For best results, Perl code should handle mainly Perl data, and Lisp code should handle mainly Lisp data.
Cross-language references incur a slight overhead.
For the benefit of Lisp's garbage collection, all Perl data that is referenced by Lisp participates in mark-and-sweep. For the benefit of Perl's garbage collection, all Lisp objects that are referenced by Perl maintain a (kind of) reference count.
A chain of Perl -> Lisp -> ... -> Perl references may take several garbage collection cycles to be freed. (At least, that is my theory. I haven't verified it experimentally.) It is therefore probably best to keep the number and complexity of such references to a minimum.
(To the extent that the Perl-to-Emacs interface is independent of the Lispish implementation of Emacs, these performance issues are fixable in principle by reimplementing Emacs' internals.)
TO DO
Provide XSubs for common, non-evalling functions.
There is substantial overhead in calling an arbitrary Lisp function, because care must be taken to restore the Perl interpreter's state when Lisp performs a non-local jump out of the function call. This can be avoided in the case of functions like cons, null, bufferp, car, eq, symbol-value, etc., for which a simple check can determine whether a jump will occur.
Special forms: unwind-protect, let, defmacro, defvar.
Make a way to get a tied filehandle that reads a buffer.
Improve perl-eval-buffer, perl-load-file, et al.
ACKNOWLEDGMENTS
These are among the giants on whose shoulders we stand:
- Larry Wall and the Perl 5 Porters.
- Richard Stallman and the GNU project.
- John McCarthy, inventor of Lisp.
-
Lisp is really nice. Perl is just more comfy in some cases.
- Dennis Ritchie, inventor of C.
-
Both Perl and Emacs are written in C. The existence of Perlmacs kind of rests on that fact.
-
ExtUtils::Embed is a cornerstone of Perlmacs.
-
Perlmacs is largely a product of GDB and gud-mode.
This list is incomplete.
Thank you, Di Zhao <dzhao@primeon.com>, for braving the alphas and showing me what can be done with Perlmacs. If not for you, I would still be wondering whether it could possibly have any use.
Personal thanks to Nate Patwardhan, who sparked my early interest in Perl--and shared his .emacs with me--during our NFIC days. Nate also introduced me to GDB under gud-mode.
Thanks also to Ilya Zakharevich for (1) encouraging me in my first contribution to the Perl development effort (an xsubpp patch), and (2) a comment in cperl-mode.el about changing Emacs C source. If not for that comment, I may never have realized that it is even possible for mortals to change Emacs C source. ;-)
COPYRIGHT
Copyright (C) 1998,1999,2000 by John Tobey, jtobey@john-edwin-tobey.org. All rights reserved.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; see the file COPYING. If not, write to the
Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston,
MA 02111-1307 USA
Please note: The GNU Emacs license (which is the GNU General Public License or "GPL") requires that all extensions and code designed specifically for use with Emacs be distributable under the same license. According to Richard Stallman, this includes dynamically linked code. Refer to the file COPYING and the Emacs documentaion for full details.
SEE ALSO
perl, perlref, perlobj, Emacs, emacs, and The Elisp Manual (available where you got the Emacs source, or from ftp://ftp.gnu.org/pub/gnu/emacs/).