Mercurial > emacs
annotate lispref/macros.texi @ 27292:d0aba668c5af
(set_internal): Enter the new arg.
author | Richard M. Stallman <rms@gnu.org> |
---|---|
date | Tue, 11 Jan 2000 22:16:55 +0000 |
parents | 89afca54a135 |
children | 79e758f8571a |
rev | line source |
---|---|
6558 | 1 @c -*-texinfo-*- |
2 @c This is part of the GNU Emacs Lisp Reference Manual. | |
21007
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
3 @c Copyright (C) 1990, 1991, 1992, 1993, 1994, 1995, 1998 Free Software Foundation, Inc. |
6558 | 4 @c See the file elisp.texi for copying conditions. |
5 @setfilename ../info/macros | |
21007
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
6 @node Macros, Customization, Functions, Top |
6558 | 7 @chapter Macros |
8 @cindex macros | |
9 | |
10 @dfn{Macros} enable you to define new control constructs and other | |
11 language features. A macro is defined much like a function, but instead | |
12 of telling how to compute a value, it tells how to compute another Lisp | |
13 expression which will in turn compute the value. We call this | |
14 expression the @dfn{expansion} of the macro. | |
15 | |
16 Macros can do this because they operate on the unevaluated expressions | |
17 for the arguments, not on the argument values as functions do. They can | |
18 therefore construct an expansion containing these argument expressions | |
19 or parts of them. | |
20 | |
21 If you are using a macro to do something an ordinary function could | |
22 do, just for the sake of speed, consider using an inline function | |
23 instead. @xref{Inline Functions}. | |
24 | |
25 @menu | |
26 * Simple Macro:: A basic example. | |
27 * Expansion:: How, when and why macros are expanded. | |
28 * Compiling Macros:: How macros are expanded by the compiler. | |
29 * Defining Macros:: How to write a macro definition. | |
30 * Backquote:: Easier construction of list structure. | |
31 * Problems with Macros:: Don't evaluate the macro arguments too many times. | |
32 Don't hide the user's variables. | |
33 @end menu | |
34 | |
35 @node Simple Macro | |
36 @section A Simple Example of a Macro | |
37 | |
38 Suppose we would like to define a Lisp construct to increment a | |
39 variable value, much like the @code{++} operator in C. We would like to | |
40 write @code{(inc x)} and have the effect of @code{(setq x (1+ x))}. | |
41 Here's a macro definition that does the job: | |
42 | |
43 @findex inc | |
44 @example | |
45 @group | |
46 (defmacro inc (var) | |
47 (list 'setq var (list '1+ var))) | |
48 @end group | |
49 @end example | |
50 | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
51 When this is called with @code{(inc x)}, the argument @var{var} is the |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
52 symbol @code{x}---@emph{not} the @emph{value} of @code{x}, as it would |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
53 be in a function. The body of the macro uses this to construct the |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
54 expansion, which is @code{(setq x (1+ x))}. Once the macro definition |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
55 returns this expansion, Lisp proceeds to evaluate it, thus incrementing |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
56 @code{x}. |
6558 | 57 |
58 @node Expansion | |
59 @section Expansion of a Macro Call | |
60 @cindex expansion of macros | |
61 @cindex macro call | |
62 | |
63 A macro call looks just like a function call in that it is a list which | |
64 starts with the name of the macro. The rest of the elements of the list | |
65 are the arguments of the macro. | |
66 | |
67 Evaluation of the macro call begins like evaluation of a function call | |
68 except for one crucial difference: the macro arguments are the actual | |
69 expressions appearing in the macro call. They are not evaluated before | |
70 they are given to the macro definition. By contrast, the arguments of a | |
71 function are results of evaluating the elements of the function call | |
72 list. | |
73 | |
74 Having obtained the arguments, Lisp invokes the macro definition just | |
75 as a function is invoked. The argument variables of the macro are bound | |
76 to the argument values from the macro call, or to a list of them in the | |
77 case of a @code{&rest} argument. And the macro body executes and | |
78 returns its value just as a function body does. | |
79 | |
80 The second crucial difference between macros and functions is that the | |
81 value returned by the macro body is not the value of the macro call. | |
82 Instead, it is an alternate expression for computing that value, also | |
83 known as the @dfn{expansion} of the macro. The Lisp interpreter | |
84 proceeds to evaluate the expansion as soon as it comes back from the | |
85 macro. | |
86 | |
87 Since the expansion is evaluated in the normal manner, it may contain | |
88 calls to other macros. It may even be a call to the same macro, though | |
89 this is unusual. | |
90 | |
91 You can see the expansion of a given macro call by calling | |
92 @code{macroexpand}. | |
93 | |
94 @defun macroexpand form &optional environment | |
95 @cindex macro expansion | |
96 This function expands @var{form}, if it is a macro call. If the result | |
97 is another macro call, it is expanded in turn, until something which is | |
98 not a macro call results. That is the value returned by | |
99 @code{macroexpand}. If @var{form} is not a macro call to begin with, it | |
100 is returned as given. | |
101 | |
102 Note that @code{macroexpand} does not look at the subexpressions of | |
103 @var{form} (although some macro definitions may do so). Even if they | |
104 are macro calls themselves, @code{macroexpand} does not expand them. | |
105 | |
106 The function @code{macroexpand} does not expand calls to inline functions. | |
107 Normally there is no need for that, since a call to an inline function is | |
108 no harder to understand than a call to an ordinary function. | |
109 | |
110 If @var{environment} is provided, it specifies an alist of macro | |
111 definitions that shadow the currently defined macros. Byte compilation | |
112 uses this feature. | |
113 | |
114 @smallexample | |
115 @group | |
116 (defmacro inc (var) | |
117 (list 'setq var (list '1+ var))) | |
118 @result{} inc | |
119 @end group | |
120 | |
121 @group | |
122 (macroexpand '(inc r)) | |
123 @result{} (setq r (1+ r)) | |
124 @end group | |
125 | |
126 @group | |
127 (defmacro inc2 (var1 var2) | |
128 (list 'progn (list 'inc var1) (list 'inc var2))) | |
129 @result{} inc2 | |
130 @end group | |
131 | |
132 @group | |
133 (macroexpand '(inc2 r s)) | |
134 @result{} (progn (inc r) (inc s)) ; @r{@code{inc} not expanded here.} | |
135 @end group | |
136 @end smallexample | |
137 @end defun | |
138 | |
139 @node Compiling Macros | |
140 @section Macros and Byte Compilation | |
141 @cindex byte-compiling macros | |
142 | |
143 You might ask why we take the trouble to compute an expansion for a | |
144 macro and then evaluate the expansion. Why not have the macro body | |
145 produce the desired results directly? The reason has to do with | |
146 compilation. | |
147 | |
148 When a macro call appears in a Lisp program being compiled, the Lisp | |
149 compiler calls the macro definition just as the interpreter would, and | |
150 receives an expansion. But instead of evaluating this expansion, it | |
151 compiles the expansion as if it had appeared directly in the program. | |
152 As a result, the compiled code produces the value and side effects | |
153 intended for the macro, but executes at full compiled speed. This would | |
154 not work if the macro body computed the value and side effects | |
155 itself---they would be computed at compile time, which is not useful. | |
156 | |
21007
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
157 In order for compilation of macro calls to work, the macros must |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
158 already be defined in Lisp when the calls to them are compiled. The |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
159 compiler has a special feature to help you do this: if a file being |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
160 compiled contains a @code{defmacro} form, the macro is defined |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
161 temporarily for the rest of the compilation of that file. To make this |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
162 feature work, you must put the @code{defmacro} in the same file where it |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
163 is used, and before its first use. |
6558 | 164 |
165 Byte-compiling a file executes any @code{require} calls at top-level | |
166 in the file. This is in case the file needs the required packages for | |
167 proper compilation. One way to ensure that necessary macro definitions | |
12098 | 168 are available during compilation is to require the files that define |
169 them (@pxref{Named Features}). To avoid loading the macro definition files | |
170 when someone @emph{runs} the compiled program, write | |
171 @code{eval-when-compile} around the @code{require} calls (@pxref{Eval | |
172 During Compile}). | |
6558 | 173 |
174 @node Defining Macros | |
175 @section Defining Macros | |
176 | |
177 A Lisp macro is a list whose @sc{car} is @code{macro}. Its @sc{cdr} should | |
178 be a function; expansion of the macro works by applying the function | |
179 (with @code{apply}) to the list of unevaluated argument-expressions | |
180 from the macro call. | |
181 | |
182 It is possible to use an anonymous Lisp macro just like an anonymous | |
183 function, but this is never done, because it does not make sense to pass | |
12098 | 184 an anonymous macro to functionals such as @code{mapcar}. In practice, |
185 all Lisp macros have names, and they are usually defined with the | |
186 special form @code{defmacro}. | |
6558 | 187 |
188 @defspec defmacro name argument-list body-forms@dots{} | |
189 @code{defmacro} defines the symbol @var{name} as a macro that looks | |
190 like this: | |
191 | |
192 @example | |
193 (macro lambda @var{argument-list} . @var{body-forms}) | |
194 @end example | |
195 | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
196 (Note that the @sc{cdr} of this list is a function---a lambda expression.) |
6558 | 197 This macro object is stored in the function cell of @var{name}. The |
198 value returned by evaluating the @code{defmacro} form is @var{name}, but | |
199 usually we ignore this value. | |
200 | |
201 The shape and meaning of @var{argument-list} is the same as in a | |
202 function, and the keywords @code{&rest} and @code{&optional} may be used | |
203 (@pxref{Argument List}). Macros may have a documentation string, but | |
204 any @code{interactive} declaration is ignored since macros cannot be | |
205 called interactively. | |
206 @end defspec | |
207 | |
208 @node Backquote | |
209 @section Backquote | |
210 @cindex backquote (list substitution) | |
211 @cindex ` (list substitution) | |
12067 | 212 @findex ` |
6558 | 213 |
214 Macros often need to construct large list structures from a mixture of | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
215 constants and nonconstant parts. To make this easier, use the @samp{`} |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
216 syntax (usually called @dfn{backquote}). |
6558 | 217 |
218 Backquote allows you to quote a list, but selectively evaluate | |
219 elements of that list. In the simplest case, it is identical to the | |
220 special form @code{quote} (@pxref{Quoting}). For example, these | |
221 two forms yield identical results: | |
222 | |
223 @example | |
224 @group | |
12067 | 225 `(a list of (+ 2 3) elements) |
6558 | 226 @result{} (a list of (+ 2 3) elements) |
227 @end group | |
228 @group | |
12067 | 229 '(a list of (+ 2 3) elements) |
6558 | 230 @result{} (a list of (+ 2 3) elements) |
231 @end group | |
232 @end example | |
233 | |
7194
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
234 @findex , @r{(with Backquote)} |
12098 | 235 The special marker @samp{,} inside of the argument to backquote |
6558 | 236 indicates a value that isn't constant. Backquote evaluates the |
12098 | 237 argument of @samp{,} and puts the value in the list structure: |
6558 | 238 |
239 @example | |
240 @group | |
241 (list 'a 'list 'of (+ 2 3) 'elements) | |
242 @result{} (a list of 5 elements) | |
243 @end group | |
244 @group | |
12067 | 245 `(a list of ,(+ 2 3) elements) |
6558 | 246 @result{} (a list of 5 elements) |
247 @end group | |
248 @end example | |
249 | |
21007
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
250 Substitution with @samp{,} is allowed at deeper levels of the list |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
251 structure also. For example: |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
252 |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
253 @example |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
254 @group |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
255 (defmacro t-becomes-nil (variable) |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
256 `(if (eq ,variable t) |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
257 (setq ,variable nil))) |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
258 @end group |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
259 |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
260 @group |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
261 (t-becomes-nil foo) |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
262 @equiv{} (if (eq foo t) (setq foo nil)) |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
263 @end group |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
264 @end example |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
265 |
7194
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
266 @findex ,@@ @r{(with Backquote)} |
6558 | 267 @cindex splicing (with backquote) |
21007
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
268 You can also @dfn{splice} an evaluated value into the resulting list, |
12098 | 269 using the special marker @samp{,@@}. The elements of the spliced list |
6558 | 270 become elements at the same level as the other elements of the resulting |
12098 | 271 list. The equivalent code without using @samp{`} is often unreadable. |
6558 | 272 Here are some examples: |
273 | |
274 @example | |
275 @group | |
276 (setq some-list '(2 3)) | |
277 @result{} (2 3) | |
278 @end group | |
279 @group | |
280 (cons 1 (append some-list '(4) some-list)) | |
281 @result{} (1 2 3 4 2 3) | |
282 @end group | |
283 @group | |
12067 | 284 `(1 ,@@some-list 4 ,@@some-list) |
6558 | 285 @result{} (1 2 3 4 2 3) |
286 @end group | |
287 | |
288 @group | |
289 (setq list '(hack foo bar)) | |
290 @result{} (hack foo bar) | |
291 @end group | |
292 @group | |
293 (cons 'use | |
294 (cons 'the | |
295 (cons 'words (append (cdr list) '(as elements))))) | |
296 @result{} (use the words foo bar as elements) | |
297 @end group | |
298 @group | |
12067 | 299 `(use the words ,@@(cdr list) as elements) |
6558 | 300 @result{} (use the words foo bar as elements) |
301 @end group | |
302 @end example | |
303 | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
304 In old Emacs versions, before version 19.29, @samp{`} used a different |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
305 syntax which required an extra level of parentheses around the entire |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
306 backquote construct. Likewise, each @samp{,} or @samp{,@@} substitution |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
307 required an extra level of parentheses surrounding both the @samp{,} or |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
308 @samp{,@@} and the following expression. The old syntax required |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
309 whitespace between the @samp{`}, @samp{,} or @samp{,@@} and the |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
310 following expression. |
6558 | 311 |
21007
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
312 This syntax is still accepted, for compatibility with old Emacs |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
313 versions, but we recommend not using it in new programs. |
6558 | 314 |
315 @node Problems with Macros | |
316 @section Common Problems Using Macros | |
317 | |
318 The basic facts of macro expansion have counterintuitive consequences. | |
319 This section describes some important consequences that can lead to | |
320 trouble, and rules to follow to avoid trouble. | |
321 | |
322 @menu | |
323 * Argument Evaluation:: The expansion should evaluate each macro arg once. | |
324 * Surprising Local Vars:: Local variable bindings in the expansion | |
325 require special care. | |
326 * Eval During Expansion:: Don't evaluate them; put them in the expansion. | |
327 * Repeated Expansion:: Avoid depending on how many times expansion is done. | |
328 @end menu | |
329 | |
330 @node Argument Evaluation | |
331 @subsection Evaluating Macro Arguments Repeatedly | |
332 | |
333 When defining a macro you must pay attention to the number of times | |
334 the arguments will be evaluated when the expansion is executed. The | |
335 following macro (used to facilitate iteration) illustrates the problem. | |
336 This macro allows us to write a simple ``for'' loop such as one might | |
337 find in Pascal. | |
338 | |
339 @findex for | |
340 @smallexample | |
341 @group | |
342 (defmacro for (var from init to final do &rest body) | |
343 "Execute a simple \"for\" loop. | |
344 For example, (for i from 1 to 10 do (print i))." | |
345 (list 'let (list (list var init)) | |
346 (cons 'while (cons (list '<= var final) | |
347 (append body (list (list 'inc var))))))) | |
348 @end group | |
349 @result{} for | |
350 | |
351 @group | |
352 (for i from 1 to 3 do | |
353 (setq square (* i i)) | |
354 (princ (format "\n%d %d" i square))) | |
355 @expansion{} | |
356 @end group | |
357 @group | |
358 (let ((i 1)) | |
359 (while (<= i 3) | |
360 (setq square (* i i)) | |
361 (princ (format "%d %d" i square)) | |
362 (inc i))) | |
363 @end group | |
364 @group | |
365 | |
366 @print{}1 1 | |
367 @print{}2 4 | |
368 @print{}3 9 | |
369 @result{} nil | |
370 @end group | |
371 @end smallexample | |
372 | |
373 @noindent | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
374 The arguments @code{from}, @code{to}, and @code{do} in this macro are |
6558 | 375 ``syntactic sugar''; they are entirely ignored. The idea is that you |
376 will write noise words (such as @code{from}, @code{to}, and @code{do}) | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
377 in those positions in the macro call. |
6558 | 378 |
7194
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
379 Here's an equivalent definition simplified through use of backquote: |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
380 |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
381 @smallexample |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
382 @group |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
383 (defmacro for (var from init to final do &rest body) |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
384 "Execute a simple \"for\" loop. |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
385 For example, (for i from 1 to 10 do (print i))." |
12098 | 386 `(let ((,var ,init)) |
387 (while (<= ,var ,final) | |
388 ,@@body | |
389 (inc ,var)))) | |
7194
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
390 @end group |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
391 @end smallexample |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
392 |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
393 Both forms of this definition (with backquote and without) suffer from |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
394 the defect that @var{final} is evaluated on every iteration. If |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
395 @var{final} is a constant, this is not a problem. If it is a more |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
396 complex form, say @code{(long-complex-calculation x)}, this can slow |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
397 down the execution significantly. If @var{final} has side effects, |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
398 executing it more than once is probably incorrect. |
6558 | 399 |
400 @cindex macro argument evaluation | |
401 A well-designed macro definition takes steps to avoid this problem by | |
402 producing an expansion that evaluates the argument expressions exactly | |
403 once unless repeated evaluation is part of the intended purpose of the | |
404 macro. Here is a correct expansion for the @code{for} macro: | |
405 | |
406 @smallexample | |
407 @group | |
408 (let ((i 1) | |
409 (max 3)) | |
410 (while (<= i max) | |
411 (setq square (* i i)) | |
412 (princ (format "%d %d" i square)) | |
413 (inc i))) | |
414 @end group | |
415 @end smallexample | |
416 | |
417 Here is a macro definition that creates this expansion: | |
418 | |
419 @smallexample | |
420 @group | |
421 (defmacro for (var from init to final do &rest body) | |
422 "Execute a simple for loop: (for i from 1 to 10 do (print i))." | |
12098 | 423 `(let ((,var ,init) |
424 (max ,final)) | |
425 (while (<= ,var max) | |
426 ,@@body | |
427 (inc ,var)))) | |
6558 | 428 @end group |
429 @end smallexample | |
430 | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
431 Unfortunately, this fix introduces another problem, |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
432 described in the following section. |
6558 | 433 |
434 @node Surprising Local Vars | |
435 @subsection Local Variables in Macro Expansions | |
436 | |
27193 | 437 @ifnottex |
6558 | 438 In the previous section, the definition of @code{for} was fixed as |
439 follows to make the expansion evaluate the macro arguments the proper | |
440 number of times: | |
441 | |
442 @smallexample | |
443 @group | |
444 (defmacro for (var from init to final do &rest body) | |
445 "Execute a simple for loop: (for i from 1 to 10 do (print i))." | |
446 @end group | |
447 @group | |
12098 | 448 `(let ((,var ,init) |
449 (max ,final)) | |
450 (while (<= ,var max) | |
451 ,@@body | |
452 (inc ,var)))) | |
6558 | 453 @end group |
454 @end smallexample | |
27193 | 455 @end ifnottex |
6558 | 456 |
457 The new definition of @code{for} has a new problem: it introduces a | |
458 local variable named @code{max} which the user does not expect. This | |
459 causes trouble in examples such as the following: | |
460 | |
7734 | 461 @smallexample |
6558 | 462 @group |
463 (let ((max 0)) | |
464 (for x from 0 to 10 do | |
465 (let ((this (frob x))) | |
466 (if (< max this) | |
467 (setq max this))))) | |
468 @end group | |
7734 | 469 @end smallexample |
6558 | 470 |
471 @noindent | |
472 The references to @code{max} inside the body of the @code{for}, which | |
473 are supposed to refer to the user's binding of @code{max}, really access | |
474 the binding made by @code{for}. | |
475 | |
476 The way to correct this is to use an uninterned symbol instead of | |
477 @code{max} (@pxref{Creating Symbols}). The uninterned symbol can be | |
7194
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
478 bound and referred to just like any other symbol, but since it is |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
479 created by @code{for}, we know that it cannot already appear in the |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
480 user's program. Since it is not interned, there is no way the user can |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
481 put it into the program later. It will never appear anywhere except |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
482 where put by @code{for}. Here is a definition of @code{for} that works |
3112fb627aa0
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
6558
diff
changeset
|
483 this way: |
6558 | 484 |
485 @smallexample | |
486 @group | |
487 (defmacro for (var from init to final do &rest body) | |
488 "Execute a simple for loop: (for i from 1 to 10 do (print i))." | |
489 (let ((tempvar (make-symbol "max"))) | |
12098 | 490 `(let ((,var ,init) |
491 (,tempvar ,final)) | |
492 (while (<= ,var ,tempvar) | |
493 ,@@body | |
494 (inc ,var))))) | |
6558 | 495 @end group |
496 @end smallexample | |
497 | |
498 @noindent | |
499 This creates an uninterned symbol named @code{max} and puts it in the | |
500 expansion instead of the usual interned symbol @code{max} that appears | |
501 in expressions ordinarily. | |
502 | |
503 @node Eval During Expansion | |
504 @subsection Evaluating Macro Arguments in Expansion | |
505 | |
22138
d4ac295a98b3
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21682
diff
changeset
|
506 Another problem can happen if the macro definition itself |
21007
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
507 evaluates any of the macro argument expressions, such as by calling |
6558 | 508 @code{eval} (@pxref{Eval}). If the argument is supposed to refer to the |
509 user's variables, you may have trouble if the user happens to use a | |
510 variable with the same name as one of the macro arguments. Inside the | |
511 macro body, the macro argument binding is the most local binding of this | |
21007
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
512 variable, so any references inside the form being evaluated do refer to |
66d807bdc5b4
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
12098
diff
changeset
|
513 it. Here is an example: |
6558 | 514 |
515 @example | |
516 @group | |
517 (defmacro foo (a) | |
518 (list 'setq (eval a) t)) | |
519 @result{} foo | |
520 @end group | |
521 @group | |
522 (setq x 'b) | |
523 (foo x) @expansion{} (setq b t) | |
524 @result{} t ; @r{and @code{b} has been set.} | |
525 ;; @r{but} | |
526 (setq a 'c) | |
527 (foo a) @expansion{} (setq a t) | |
528 @result{} t ; @r{but this set @code{a}, not @code{c}.} | |
529 | |
530 @end group | |
531 @end example | |
532 | |
533 It makes a difference whether the user's variable is named @code{a} or | |
534 @code{x}, because @code{a} conflicts with the macro argument variable | |
535 @code{a}. | |
536 | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
537 Another problem with calling @code{eval} in a macro definition is that |
6558 | 538 it probably won't do what you intend in a compiled program. The |
539 byte-compiler runs macro definitions while compiling the program, when | |
540 the program's own computations (which you might have wished to access | |
541 with @code{eval}) don't occur and its local variable bindings don't | |
542 exist. | |
543 | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
544 To avoid these problems, @strong{don't evaluate an argument expression |
22138
d4ac295a98b3
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21682
diff
changeset
|
545 while computing the macro expansion}. Instead, substitute the |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
546 expression into the macro expansion, so that its value will be computed |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
547 as part of executing the expansion. This is how the other examples in |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
548 this chapter work. |
6558 | 549 |
550 @node Repeated Expansion | |
551 @subsection How Many Times is the Macro Expanded? | |
552 | |
553 Occasionally problems result from the fact that a macro call is | |
554 expanded each time it is evaluated in an interpreted function, but is | |
555 expanded only once (during compilation) for a compiled function. If the | |
556 macro definition has side effects, they will work differently depending | |
557 on how many times the macro is expanded. | |
558 | |
21682
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
559 Therefore, you should avoid side effects in computation of the |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
560 macro expansion, unless you really know what you are doing. |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
561 |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
562 One special kind of side effect can't be avoided: constructing Lisp |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
563 objects. Almost all macro expansions include constructed lists; that is |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
564 the whole point of most macros. This is usually safe; there is just one |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
565 case where you must be careful: when the object you construct is part of a |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
566 quoted constant in the macro expansion. |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
567 |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
568 If the macro is expanded just once, in compilation, then the object is |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
569 constructed just once, during compilation. But in interpreted |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
570 execution, the macro is expanded each time the macro call runs, and this |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
571 means a new object is constructed each time. |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
572 |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
573 In most clean Lisp code, this difference won't matter. It can matter |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
574 only if you perform side-effects on the objects constructed by the macro |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
575 definition. Thus, to avoid trouble, @strong{avoid side effects on |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
576 objects constructed by macro definitions}. Here is an example of how |
90da2489c498
*** empty log message ***
Richard M. Stallman <rms@gnu.org>
parents:
21007
diff
changeset
|
577 such side effects can get you into trouble: |
6558 | 578 |
579 @lisp | |
580 @group | |
581 (defmacro empty-object () | |
582 (list 'quote (cons nil nil))) | |
583 @end group | |
584 | |
585 @group | |
586 (defun initialize (condition) | |
587 (let ((object (empty-object))) | |
588 (if condition | |
589 (setcar object condition)) | |
590 object)) | |
591 @end group | |
592 @end lisp | |
593 | |
594 @noindent | |
595 If @code{initialize} is interpreted, a new list @code{(nil)} is | |
596 constructed each time @code{initialize} is called. Thus, no side effect | |
597 survives between calls. If @code{initialize} is compiled, then the | |
598 macro @code{empty-object} is expanded during compilation, producing a | |
599 single ``constant'' @code{(nil)} that is reused and altered each time | |
600 @code{initialize} is called. | |
601 | |
602 One way to avoid pathological cases like this is to think of | |
603 @code{empty-object} as a funny kind of constant, not as a memory | |
604 allocation construct. You wouldn't use @code{setcar} on a constant such | |
605 as @code{'(nil)}, so naturally you won't use it on @code{(empty-object)} | |
606 either. |