Minor cleanups for overfull hboxes.
[bpt/emacs.git] / lispref / compile.texi
1 @c -*-texinfo-*-
2 @c This is part of the GNU Emacs Lisp Reference Manual.
3 @c Copyright (C) 1990, 1991, 1992, 1993, 1994 Free Software Foundation, Inc.
4 @c See the file elisp.texi for copying conditions.
5 @setfilename ../info/compile
6 @node Byte Compilation, Debugging, Loading, Top
7 @chapter Byte Compilation
8 @cindex byte-code
9 @cindex compilation
10
11 GNU Emacs Lisp has a @dfn{compiler} that translates functions written
12 in Lisp into a special representation called @dfn{byte-code} that can be
13 executed more efficiently. The compiler replaces Lisp function
14 definitions with byte-code. When a byte-code function is called, its
15 definition is evaluated by the @dfn{byte-code interpreter}.
16
17 Because the byte-compiled code is evaluated by the byte-code
18 interpreter, instead of being executed directly by the machine's
19 hardware (as true compiled code is), byte-code is completely
20 transportable from machine to machine without recompilation. It is not,
21 however, as fast as true compiled code.
22
23 In general, any version of Emacs can run byte-compiled code produced
24 by recent earlier versions of Emacs, but the reverse is not true. In
25 particular, if you compile a program with Emacs 19.29, the compiled
26 code does not run in earlier versions.
27 @iftex
28 @xref{Docs and Compilation}.
29 @end iftex
30 Files compiled in versions before 19.29 may not work in 19.29 if they
31 contain character constants with modifier bits, because the bits were
32 renumbered in Emacs 19.29.
33
34 @xref{Compilation Errors}, for how to investigate errors occurring in
35 byte compilation.
36
37 @menu
38 * Speed of Byte-Code:: An example of speedup from byte compilation.
39 * Compilation Functions:: Byte compilation functions.
40 * Docs and Compilation:: Dynamic loading of documentation strings.
41 * Dynamic Loading:: Dynamic loading of individual functions.
42 * Eval During Compile:: Code to be evaluated when you compile.
43 * Byte-Code Objects:: The data type used for byte-compiled functions.
44 * Disassembly:: Disassembling byte-code; how to read byte-code.
45 @end menu
46
47 @node Speed of Byte-Code
48 @section Performance of Byte-Compiled Code
49
50 A byte-compiled function is not as efficient as a primitive function
51 written in C, but runs much faster than the version written in Lisp.
52 Here is an example:
53
54 @example
55 @group
56 (defun silly-loop (n)
57 "Return time before and after N iterations of a loop."
58 (let ((t1 (current-time-string)))
59 (while (> (setq n (1- n))
60 0))
61 (list t1 (current-time-string))))
62 @result{} silly-loop
63 @end group
64
65 @group
66 (silly-loop 100000)
67 @result{} ("Fri Mar 18 17:25:57 1994"
68 "Fri Mar 18 17:26:28 1994") ; @r{31 seconds}
69 @end group
70
71 @group
72 (byte-compile 'silly-loop)
73 @result{} @r{[Compiled code not shown]}
74 @end group
75
76 @group
77 (silly-loop 100000)
78 @result{} ("Fri Mar 18 17:26:52 1994"
79 "Fri Mar 18 17:26:58 1994") ; @r{6 seconds}
80 @end group
81 @end example
82
83 In this example, the interpreted code required 31 seconds to run,
84 whereas the byte-compiled code required 6 seconds. These results are
85 representative, but actual results will vary greatly.
86
87 @node Compilation Functions
88 @comment node-name, next, previous, up
89 @section The Compilation Functions
90 @cindex compilation functions
91
92 You can byte-compile an individual function or macro definition with
93 the @code{byte-compile} function. You can compile a whole file with
94 @code{byte-compile-file}, or several files with
95 @code{byte-recompile-directory} or @code{batch-byte-compile}.
96
97 The byte compiler produces error messages and warnings about each file
98 in a buffer called @samp{*Compile-Log*}. These report things in your
99 program that suggest a problem but are not necessarily erroneous.
100
101 @cindex macro compilation
102 Be careful when byte-compiling code that uses macros. Macro calls are
103 expanded when they are compiled, so the macros must already be defined
104 for proper compilation. For more details, see @ref{Compiling Macros}.
105
106 Normally, compiling a file does not evaluate the file's contents or
107 load the file. But it does execute any @code{require} calls at top
108 level in the file. One way to ensure that necessary macro definitions
109 are available during compilation is to require the file that defines
110 them (@pxref{Named Features}). To avoid loading the macro definition files
111 when someone @emph{runs} the compiled program, write
112 @code{eval-when-compile} around the @code{require} calls (@pxref{Eval
113 During Compile}).
114
115 @defun byte-compile symbol
116 This function byte-compiles the function definition of @var{symbol},
117 replacing the previous definition with the compiled one. The function
118 definition of @var{symbol} must be the actual code for the function;
119 i.e., the compiler does not follow indirection to another symbol.
120 @code{byte-compile} returns the new, compiled definition of
121 @var{symbol}.
122
123 If @var{symbol}'s definition is a byte-code function object,
124 @code{byte-compile} does nothing and returns @code{nil}. Lisp records
125 only one function definition for any symbol, and if that is already
126 compiled, non-compiled code is not available anywhere. So there is no
127 way to ``compile the same definition again.''
128
129 @example
130 @group
131 (defun factorial (integer)
132 "Compute factorial of INTEGER."
133 (if (= 1 integer) 1
134 (* integer (factorial (1- integer)))))
135 @result{} factorial
136 @end group
137
138 @group
139 (byte-compile 'factorial)
140 @result{}
141 #[(integer)
142 "^H\301U\203^H^@@\301\207\302^H\303^HS!\"\207"
143 [integer 1 * factorial]
144 4 "Compute factorial of INTEGER."]
145 @end group
146 @end example
147
148 @noindent
149 The result is a byte-code function object. The string it contains is
150 the actual byte-code; each character in it is an instruction or an
151 operand of an instruction. The vector contains all the constants,
152 variable names and function names used by the function, except for
153 certain primitives that are coded as special instructions.
154 @end defun
155
156 @deffn Command compile-defun
157 This command reads the defun containing point, compiles it, and
158 evaluates the result. If you use this on a defun that is actually a
159 function definition, the effect is to install a compiled version of that
160 function.
161 @end deffn
162
163 @deffn Command byte-compile-file filename
164 This function compiles a file of Lisp code named @var{filename} into
165 a file of byte-code. The output file's name is made by appending
166 @samp{c} to the end of @var{filename}.
167
168 Compilation works by reading the input file one form at a time. If it
169 is a definition of a function or macro, the compiled function or macro
170 definition is written out. Other forms are batched together, then each
171 batch is compiled, and written so that its compiled code will be
172 executed when the file is read. All comments are discarded when the
173 input file is read.
174
175 This command returns @code{t}. When called interactively, it prompts
176 for the file name.
177
178 @example
179 @group
180 % ls -l push*
181 -rw-r--r-- 1 lewis 791 Oct 5 20:31 push.el
182 @end group
183
184 @group
185 (byte-compile-file "~/emacs/push.el")
186 @result{} t
187 @end group
188
189 @group
190 % ls -l push*
191 -rw-r--r-- 1 lewis 791 Oct 5 20:31 push.el
192 -rw-rw-rw- 1 lewis 638 Oct 8 20:25 push.elc
193 @end group
194 @end example
195 @end deffn
196
197 @deffn Command byte-recompile-directory directory flag
198 @cindex library compilation
199 This function recompiles every @samp{.el} file in @var{directory} that
200 needs recompilation. A file needs recompilation if a @samp{.elc} file
201 exists but is older than the @samp{.el} file.
202
203 When a @samp{.el} file has no corresponding @samp{.elc} file, then
204 @var{flag} says what to do. If it is @code{nil}, these files are
205 ignored. If it is non-@code{nil}, the user is asked whether to compile
206 each such file.
207
208 The returned value of this command is unpredictable.
209 @end deffn
210
211 @defun batch-byte-compile
212 This function runs @code{byte-compile-file} on files specified on the
213 command line. This function must be used only in a batch execution of
214 Emacs, as it kills Emacs on completion. An error in one file does not
215 prevent processing of subsequent files. (The file that gets the error
216 will not, of course, produce any compiled code.)
217
218 @example
219 % emacs -batch -f batch-byte-compile *.el
220 @end example
221 @end defun
222
223 @defun byte-code code-string data-vector max-stack
224 @cindex byte-code interpreter
225 This function actually interprets byte-code. A byte-compiled function
226 is actually defined with a body that calls @code{byte-code}. Don't call
227 this function yourself. Only the byte compiler knows how to generate
228 valid calls to this function.
229
230 In newer Emacs versions (19 and up), byte-code is usually executed as
231 part of a byte-code function object, and only rarely due to an explicit
232 call to @code{byte-code}.
233 @end defun
234
235 @node Docs and Compilation
236 @section Documentation Strings and Compilation
237 @cindex dynamic loading of documentation
238
239 Functions and variables loaded from a byte-compiled file access their
240 documentation strings dynamically from the file whenever needed. This
241 saves space within Emacs, and makes loading faster because the
242 documentation strings themselves need not be processed while loading the
243 file. Actual access to the documentation strings becomes slower as a
244 result, but this normally is not enough to bother users.
245
246 Dynamic access to documentation strings does have drawbacks:
247
248 @itemize @bullet
249 @item
250 If you delete or move the compiled file after loading it, Emacs can no
251 longer access the documentation strings for the functions and variables
252 in the file.
253
254 @item
255 If you alter the compiled file (such as by compiling a new version),
256 then further access to documentation strings in this file will give
257 nonsense results.
258 @end itemize
259
260 If your site installs Emacs following the usual procedures, these
261 problems will never normally occur. Installing a new version uses a new
262 directory with a different name; as long as the old version remains
263 installed, its files will remain unmodified in the places where they are
264 expected to be.
265
266 However, if you have built Emacs yourself and use it from the
267 directory where you built it, you will experience this problem
268 occasionally if you edit and recompile Lisp files. When it happens, you
269 can cure the problem by reloading the file after recompiling it.
270
271 Byte-compiled files made with Emacs 19.29 will not load into older
272 versions because the older versions don't support this feature. You can
273 turn off this feature by setting @code{byte-compile-dynamic-docstrings}
274 to @code{nil}. Once this is done, you can compile files that will load
275 into older Emacs versions. You can do this globally, or for one source
276 file by specifying a file-local binding for the variable. Here's one
277 way to do that:
278
279 @example
280 -*-byte-compile-dynamic-docstrings: nil;-*-
281 @end example
282
283 @defvar byte-compile-dynamic-docstrings
284 If this is non-@code{nil}, the byte compiler generates compiled files
285 that are set up for dynamic loading of documentation strings.
286 @end defvar
287
288 @cindex @samp{#@@@var{count}}
289 @cindex @samp{#$}
290 The dynamic documentation string feature writes compiled files that
291 use a special Lisp reader construct, @samp{#@@@var{count}}. This
292 construct skips the next @var{count} characters. It also uses the
293 @samp{#$} construct, which stands for ``the name of this file, as a
294 string.'' It is best not to use these constructs in Lisp source files.
295
296 @node Dynamic Loading
297 @section Dynamic Loading of Individual Functions
298
299 @cindex dynamic loading of functions
300 @cindex lazy loading
301 When you compile a file, you can optionally enable the @dfn{dynamic
302 function loading} feature (also known as @dfn{lazy loading}). With
303 dynamic function loading, loading the file doesn't fully read the
304 function definitions in the file. Instead, each function definition
305 contains a place-holder which refers to the file. The first time each
306 function is called, it reads the full definition from the file, to
307 replace the place-holder.
308
309 The advantage of dynamic function loading is that loading the file
310 becomes much faster. This is a good thing for a file which contains
311 many separate commands, provided that using one of them does not imply
312 you will soon (or ever) use the rest. A specialized mode which provides
313 many keyboard commands often has that usage pattern: a user may invoke
314 the mode, but use only a few of the commands it provides.
315
316 The dynamic loading feature has certain disadvantages:
317
318 @itemize @bullet
319 @item
320 If you delete or move the compiled file after loading it, Emacs can no
321 longer load the remaining function definitions not already loaded.
322
323 @item
324 If you alter the compiled file (such as by compiling a new version),
325 then trying to load any function not already loaded will get nonsense
326 results.
327 @end itemize
328
329 If you compile a new version of the file, the best thing to do is
330 immediately load the new compiled file. That will prevent any future
331 problems.
332
333 The byte compiler uses the dynamic function loading feature if the
334 variable @code{byte-compile-dynamic} is non-@code{nil} at compilation
335 time. Do not set this variable globally, since dynamic loading is
336 desirable only for certain files. Instead, enable the feature for
337 specific source files with file-local variable bindings, like this:
338
339 @example
340 -*-byte-compile-dynamic: t;-*-
341 @end example
342
343 @defvar byte-compile-dynamic
344 If this is non-@code{nil}, the byte compiler generates compiled files
345 that are set up for dynamic function loading.
346 @end defvar
347
348 @defun fetch-bytecode function
349 This immediately finishes loading the definition of @var{function} from
350 its byte-compiled file, if it is not fully loaded already. The argument
351 @var{function} may be a byte-code function object or a function name.
352 @end defun
353
354 @node Eval During Compile
355 @section Evaluation During Compilation
356
357 These features permit you to write code to be evaluated during
358 compilation of a program.
359
360 @defspec eval-and-compile body
361 This form marks @var{body} to be evaluated both when you compile the
362 containing code and when you run it (whether compiled or not).
363
364 You can get a similar result by putting @var{body} in a separate file
365 and referring to that file with @code{require}. Using @code{require} is
366 preferable if there is a substantial amount of code to be executed in
367 this way.
368 @end defspec
369
370 @defspec eval-when-compile body
371 This form marks @var{body} to be evaluated at compile time and not when
372 the compiled program is loaded. The result of evaluation by the
373 compiler becomes a constant which appears in the compiled program. When
374 the program is interpreted, not compiled at all, @var{body} is evaluated
375 normally.
376
377 At top level, this is analogous to the Common Lisp idiom
378 @code{(eval-when (compile eval) @dots{})}. Elsewhere, the Common Lisp
379 @samp{#.} reader macro (but not when interpreting) is closer to what
380 @code{eval-when-compile} does.
381 @end defspec
382
383 @node Byte-Code Objects
384 @section Byte-Code Function Objects
385 @cindex compiled function
386 @cindex byte-code function
387
388 Byte-compiled functions have a special data type: they are
389 @dfn{byte-code function objects}.
390
391 Internally, a byte-code function object is much like a vector;
392 however, the evaluator handles this data type specially when it appears
393 as a function to be called. The printed representation for a byte-code
394 function object is like that for a vector, with an additional @samp{#}
395 before the opening @samp{[}.
396
397 In Emacs version 18, there was no byte-code function object data type;
398 compiled functions used the function @code{byte-code} to run the byte
399 code.
400
401 A byte-code function object must have at least four elements; there is
402 no maximum number, but only the first six elements are actually used.
403 They are:
404
405 @table @var
406 @item arglist
407 The list of argument symbols.
408
409 @item byte-code
410 The string containing the byte-code instructions.
411
412 @item constants
413 The vector of Lisp objects referenced by the byte code. These include
414 symbols used as function names and variable names.
415
416 @item stacksize
417 The maximum stack size this function needs.
418
419 @item docstring
420 The documentation string (if any); otherwise, @code{nil}. The value may
421 be a number or a list, in case the documentation string is stored in a
422 file. Use the function @code{documentation} to get the real
423 documentation string (@pxref{Accessing Documentation}).
424
425 @item interactive
426 The interactive spec (if any). This can be a string or a Lisp
427 expression. It is @code{nil} for a function that isn't interactive.
428 @end table
429
430 Here's an example of a byte-code function object, in printed
431 representation. It is the definition of the command
432 @code{backward-sexp}.
433
434 @example
435 #[(&optional arg)
436 "^H\204^F^@@\301^P\302^H[!\207"
437 [arg 1 forward-sexp]
438 2
439 254435
440 "p"]
441 @end example
442
443 The primitive way to create a byte-code object is with
444 @code{make-byte-code}:
445
446 @defun make-byte-code &rest elements
447 This function constructs and returns a byte-code function object
448 with @var{elements} as its elements.
449 @end defun
450
451 You should not try to come up with the elements for a byte-code
452 function yourself, because if they are inconsistent, Emacs may crash
453 when you call the function. Always leave it to the byte compiler to
454 create these objects; it makes the elements consistent (we hope).
455
456 You can access the elements of a byte-code object using @code{aref};
457 you can also use @code{vconcat} to create a vector with the same
458 elements.
459
460 @node Disassembly
461 @section Disassembled Byte-Code
462 @cindex disassembled byte-code
463
464 People do not write byte-code; that job is left to the byte compiler.
465 But we provide a disassembler to satisfy a cat-like curiosity. The
466 disassembler converts the byte-compiled code into humanly readable
467 form.
468
469 The byte-code interpreter is implemented as a simple stack machine.
470 It pushes values onto a stack of its own, then pops them off to use them
471 in calculations whose results are themselves pushed back on the stack.
472 When a byte-code function returns, it pops a value off the stack and
473 returns it as the value of the function.
474
475 In addition to the stack, byte-code functions can use, bind, and set
476 ordinary Lisp variables, by transferring values between variables and
477 the stack.
478
479 @deffn Command disassemble object &optional stream
480 This function prints the disassembled code for @var{object}. If
481 @var{stream} is supplied, then output goes there. Otherwise, the
482 disassembled code is printed to the stream @code{standard-output}. The
483 argument @var{object} can be a function name or a lambda expression.
484
485 As a special exception, if this function is used interactively,
486 it outputs to a buffer named @samp{*Disassemble*}.
487 @end deffn
488
489 Here are two examples of using the @code{disassemble} function. We
490 have added explanatory comments to help you relate the byte-code to the
491 Lisp source; these do not appear in the output of @code{disassemble}.
492 These examples show unoptimized byte-code. Nowadays byte-code is
493 usually optimized, but we did not want to rewrite these examples, since
494 they still serve their purpose.
495
496 @example
497 @group
498 (defun factorial (integer)
499 "Compute factorial of an integer."
500 (if (= 1 integer) 1
501 (* integer (factorial (1- integer)))))
502 @result{} factorial
503 @end group
504
505 @group
506 (factorial 4)
507 @result{} 24
508 @end group
509
510 @group
511 (disassemble 'factorial)
512 @print{} byte-code for factorial:
513 doc: Compute factorial of an integer.
514 args: (integer)
515 @end group
516
517 @group
518 0 constant 1 ; @r{Push 1 onto stack.}
519
520 1 varref integer ; @r{Get value of @code{integer}}
521 ; @r{from the environment}
522 ; @r{and push the value}
523 ; @r{onto the stack.}
524 @end group
525
526 @group
527 2 eqlsign ; @r{Pop top two values off stack,}
528 ; @r{compare them,}
529 ; @r{and push result onto stack.}
530 @end group
531
532 @group
533 3 goto-if-nil 10 ; @r{Pop and test top of stack;}
534 ; @r{if @code{nil}, go to 10,}
535 ; @r{else continue.}
536 @end group
537
538 @group
539 6 constant 1 ; @r{Push 1 onto top of stack.}
540
541 7 goto 17 ; @r{Go to 17 (in this case, 1 will be}
542 ; @r{returned by the function).}
543 @end group
544
545 @group
546 10 constant * ; @r{Push symbol @code{*} onto stack.}
547
548 11 varref integer ; @r{Push value of @code{integer} onto stack.}
549 @end group
550
551 @group
552 12 constant factorial ; @r{Push @code{factorial} onto stack.}
553
554 13 varref integer ; @r{Push value of @code{integer} onto stack.}
555
556 14 sub1 ; @r{Pop @code{integer}, decrement value,}
557 ; @r{push new value onto stack.}
558 @end group
559
560 @group
561 ; @r{Stack now contains:}
562 ; @minus{} @r{decremented value of @code{integer}}
563 ; @minus{} @r{@code{factorial}}
564 ; @minus{} @r{value of @code{integer}}
565 ; @minus{} @r{@code{*}}
566 @end group
567
568 @group
569 15 call 1 ; @r{Call function @code{factorial} using}
570 ; @r{the first (i.e., the top) element}
571 ; @r{of the stack as the argument;}
572 ; @r{push returned value onto stack.}
573 @end group
574
575 @group
576 ; @r{Stack now contains:}
577 ; @minus{} @r{result of recursive}
578 ; @r{call to @code{factorial}}
579 ; @minus{} @r{value of @code{integer}}
580 ; @minus{} @r{@code{*}}
581 @end group
582
583 @group
584 16 call 2 ; @r{Using the first two}
585 ; @r{(i.e., the top two)}
586 ; @r{elements of the stack}
587 ; @r{as arguments,}
588 ; @r{call the function @code{*},}
589 ; @r{pushing the result onto the stack.}
590 @end group
591
592 @group
593 17 return ; @r{Return the top element}
594 ; @r{of the stack.}
595 @result{} nil
596 @end group
597 @end example
598
599 The @code{silly-loop} function is somewhat more complex:
600
601 @example
602 @group
603 (defun silly-loop (n)
604 "Return time before and after N iterations of a loop."
605 (let ((t1 (current-time-string)))
606 (while (> (setq n (1- n))
607 0))
608 (list t1 (current-time-string))))
609 @result{} silly-loop
610 @end group
611
612 @group
613 (disassemble 'silly-loop)
614 @print{} byte-code for silly-loop:
615 doc: Return time before and after N iterations of a loop.
616 args: (n)
617
618 0 constant current-time-string ; @r{Push}
619 ; @r{@code{current-time-string}}
620 ; @r{onto top of stack.}
621 @end group
622
623 @group
624 1 call 0 ; @r{Call @code{current-time-string}}
625 ; @r{ with no argument,}
626 ; @r{ pushing result onto stack.}
627 @end group
628
629 @group
630 2 varbind t1 ; @r{Pop stack and bind @code{t1}}
631 ; @r{to popped value.}
632 @end group
633
634 @group
635 3 varref n ; @r{Get value of @code{n} from}
636 ; @r{the environment and push}
637 ; @r{the value onto the stack.}
638 @end group
639
640 @group
641 4 sub1 ; @r{Subtract 1 from top of stack.}
642 @end group
643
644 @group
645 5 dup ; @r{Duplicate the top of the stack;}
646 ; @r{i.e., copy the top of}
647 ; @r{the stack and push the}
648 ; @r{copy onto the stack.}
649 @end group
650
651 @group
652 6 varset n ; @r{Pop the top of the stack,}
653 ; @r{and bind @code{n} to the value.}
654
655 ; @r{In effect, the sequence @code{dup varset}}
656 ; @r{copies the top of the stack}
657 ; @r{into the value of @code{n}}
658 ; @r{without popping it.}
659 @end group
660
661 @group
662 7 constant 0 ; @r{Push 0 onto stack.}
663 @end group
664
665 @group
666 8 gtr ; @r{Pop top two values off stack,}
667 ; @r{test if @var{n} is greater than 0}
668 ; @r{and push result onto stack.}
669 @end group
670
671 @group
672 9 goto-if-nil-else-pop 17 ; @r{Goto 17 if @code{n} <= 0}
673 ; @r{(this exits the while loop).}
674 ; @r{else pop top of stack}
675 ; @r{and continue}
676 @end group
677
678 @group
679 12 constant nil ; @r{Push @code{nil} onto stack}
680 ; @r{(this is the body of the loop).}
681 @end group
682
683 @group
684 13 discard ; @r{Discard result of the body}
685 ; @r{of the loop (a while loop}
686 ; @r{is always evaluated for}
687 ; @r{its side effects).}
688 @end group
689
690 @group
691 14 goto 3 ; @r{Jump back to beginning}
692 ; @r{of while loop.}
693 @end group
694
695 @group
696 17 discard ; @r{Discard result of while loop}
697 ; @r{by popping top of stack.}
698 ; @r{This result is the value @code{nil} that}
699 ; @r{was not popped by the goto at 9.}
700 @end group
701
702 @group
703 18 varref t1 ; @r{Push value of @code{t1} onto stack.}
704 @end group
705
706 @group
707 19 constant current-time-string ; @r{Push}
708 ; @r{@code{current-time-string}}
709 ; @r{onto top of stack.}
710 @end group
711
712 @group
713 20 call 0 ; @r{Call @code{current-time-string} again.}
714 @end group
715
716 @group
717 21 list2 ; @r{Pop top two elements off stack,}
718 ; @r{create a list of them,}
719 ; @r{and push list onto stack.}
720 @end group
721
722 @group
723 22 unbind 1 ; @r{Unbind @code{t1} in local environment.}
724
725 23 return ; @r{Return value of the top of stack.}
726
727 @result{} nil
728 @end group
729 @end example
730
731