2 @c This is part of the GNU Emacs Lisp Reference Manual.
3 @c Copyright (C) 1990-1995, 1998, 2001-2015 Free Software Foundation,
5 @c See the file elisp.texi for copying conditions.
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.
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
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}.
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 * Problems with Macros:: Don't evaluate the macro arguments too many times.
31 Don't hide the user's variables.
32 * Indenting Macros:: Specifying how to indent macro calls.
36 @section A Simple Example of a Macro
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:
47 (list 'setq var (list '1+ var)))
51 When this is called with @code{(inc x)}, the argument @var{var} is the
52 symbol @code{x}---@emph{not} the @emph{value} of @code{x}, as it would
53 be in a function. The body of the macro uses this to construct the
54 expansion, which is @code{(setq x (1+ x))}. Once the macro definition
55 returns this expansion, Lisp proceeds to evaluate it, thus incrementing
59 This predicate tests whether its argument is a macro, and returns
60 @code{t} if so, @code{nil} otherwise.
64 @section Expansion of a Macro Call
65 @cindex expansion of macros
68 A macro call looks just like a function call in that it is a list which
69 starts with the name of the macro. The rest of the elements of the list
70 are the arguments of the macro.
72 Evaluation of the macro call begins like evaluation of a function call
73 except for one crucial difference: the macro arguments are the actual
74 expressions appearing in the macro call. They are not evaluated before
75 they are given to the macro definition. By contrast, the arguments of a
76 function are results of evaluating the elements of the function call
79 Having obtained the arguments, Lisp invokes the macro definition just
80 as a function is invoked. The argument variables of the macro are bound
81 to the argument values from the macro call, or to a list of them in the
82 case of a @code{&rest} argument. And the macro body executes and
83 returns its value just as a function body does.
85 The second crucial difference between macros and functions is that
86 the value returned by the macro body is an alternate Lisp expression,
87 also known as the @dfn{expansion} of the macro. The Lisp interpreter
88 proceeds to evaluate the expansion as soon as it comes back from the
91 Since the expansion is evaluated in the normal manner, it may contain
92 calls to other macros. It may even be a call to the same macro, though
95 Note that Emacs tries to expand macros when loading an uncompiled
96 Lisp file. This is not always possible, but if it is, it speeds up
97 subsequent execution. @xref{How Programs Do Loading}.
99 You can see the expansion of a given macro call by calling
102 @defun macroexpand form &optional environment
103 @cindex macro expansion
104 This function expands @var{form}, if it is a macro call. If the result
105 is another macro call, it is expanded in turn, until something which is
106 not a macro call results. That is the value returned by
107 @code{macroexpand}. If @var{form} is not a macro call to begin with, it
108 is returned as given.
110 Note that @code{macroexpand} does not look at the subexpressions of
111 @var{form} (although some macro definitions may do so). Even if they
112 are macro calls themselves, @code{macroexpand} does not expand them.
114 The function @code{macroexpand} does not expand calls to inline functions.
115 Normally there is no need for that, since a call to an inline function is
116 no harder to understand than a call to an ordinary function.
118 If @var{environment} is provided, it specifies an alist of macro
119 definitions that shadow the currently defined macros. Byte compilation
125 (list 'setq var (list '1+ var)))
129 (macroexpand '(inc r))
130 @result{} (setq r (1+ r))
134 (defmacro inc2 (var1 var2)
135 (list 'progn (list 'inc var1) (list 'inc var2)))
139 (macroexpand '(inc2 r s))
140 @result{} (progn (inc r) (inc s)) ; @r{@code{inc} not expanded here.}
146 @defun macroexpand-all form &optional environment
147 @code{macroexpand-all} expands macros like @code{macroexpand}, but
148 will look for and expand all macros in @var{form}, not just at the
149 top-level. If no macros are expanded, the return value is @code{eq}
152 Repeating the example used for @code{macroexpand} above with
153 @code{macroexpand-all}, we see that @code{macroexpand-all} @emph{does}
154 expand the embedded calls to @code{inc}:
157 (macroexpand-all '(inc2 r s))
158 @result{} (progn (setq r (1+ r)) (setq s (1+ s)))
163 @node Compiling Macros
164 @section Macros and Byte Compilation
165 @cindex byte-compiling macros
167 You might ask why we take the trouble to compute an expansion for a
168 macro and then evaluate the expansion. Why not have the macro body
169 produce the desired results directly? The reason has to do with
172 When a macro call appears in a Lisp program being compiled, the Lisp
173 compiler calls the macro definition just as the interpreter would, and
174 receives an expansion. But instead of evaluating this expansion, it
175 compiles the expansion as if it had appeared directly in the program.
176 As a result, the compiled code produces the value and side effects
177 intended for the macro, but executes at full compiled speed. This would
178 not work if the macro body computed the value and side effects
179 itself---they would be computed at compile time, which is not useful.
181 In order for compilation of macro calls to work, the macros must
182 already be defined in Lisp when the calls to them are compiled. The
183 compiler has a special feature to help you do this: if a file being
184 compiled contains a @code{defmacro} form, the macro is defined
185 temporarily for the rest of the compilation of that file.
187 Byte-compiling a file also executes any @code{require} calls at
188 top-level in the file, so you can ensure that necessary macro
189 definitions are available during compilation by requiring the files
190 that define them (@pxref{Named Features}). To avoid loading the macro
191 definition files when someone @emph{runs} the compiled program, write
192 @code{eval-when-compile} around the @code{require} calls (@pxref{Eval
195 @node Defining Macros
196 @section Defining Macros
197 @cindex defining macros
198 @cindex macro, how to define
200 A Lisp macro object is a list whose @sc{car} is @code{macro}, and
201 whose @sc{cdr} is a function. Expansion of the macro works
202 by applying the function (with @code{apply}) to the list of
203 @emph{unevaluated} arguments from the macro call.
205 It is possible to use an anonymous Lisp macro just like an anonymous
206 function, but this is never done, because it does not make sense to
207 pass an anonymous macro to functionals such as @code{mapcar}. In
208 practice, all Lisp macros have names, and they are almost always
209 defined with the @code{defmacro} macro.
211 @defmac defmacro name args [doc] [declare] body@dots{}
212 @code{defmacro} defines the symbol @var{name} (which should not be
213 quoted) as a macro that looks like this:
216 (macro lambda @var{args} . @var{body})
219 (Note that the @sc{cdr} of this list is a lambda expression.) This
220 macro object is stored in the function cell of @var{name}. The
221 meaning of @var{args} is the same as in a function, and the keywords
222 @code{&rest} and @code{&optional} may be used (@pxref{Argument List}).
223 Neither @var{name} nor @var{args} should be quoted. The return value
224 of @code{defmacro} is undefined.
226 @var{doc}, if present, should be a string specifying the macro's
227 documentation string. @var{declare}, if present, should be a
228 @code{declare} form specifying metadata for the macro (@pxref{Declare
229 Form}). Note that macros cannot have interactive declarations, since
230 they cannot be called interactively.
233 Macros often need to construct large list structures from a mixture
234 of constants and nonconstant parts. To make this easier, use the
235 @samp{`} syntax (@pxref{Backquote}). For example:
240 (defmacro t-becomes-nil (variable)
241 `(if (eq ,variable t)
242 (setq ,variable nil)))
247 @equiv{} (if (eq foo t) (setq foo nil))
252 The body of a macro definition can include a @code{declare} form,
253 which specifies additional properties about the macro. @xref{Declare
256 @node Problems with Macros
257 @section Common Problems Using Macros
258 @cindex macro caveats
260 Macro expansion can have counterintuitive consequences. This
261 section describes some important consequences that can lead to
262 trouble, and rules to follow to avoid trouble.
265 * Wrong Time:: Do the work in the expansion, not in the macro.
266 * Argument Evaluation:: The expansion should evaluate each macro arg once.
267 * Surprising Local Vars:: Local variable bindings in the expansion
268 require special care.
269 * Eval During Expansion:: Don't evaluate them; put them in the expansion.
270 * Repeated Expansion:: Avoid depending on how many times expansion is done.
274 @subsection Wrong Time
276 The most common problem in writing macros is doing some of the
277 real work prematurely---while expanding the macro, rather than in the
278 expansion itself. For instance, one real package had this macro
282 (defmacro my-set-buffer-multibyte (arg)
283 (if (fboundp 'set-buffer-multibyte)
284 (set-buffer-multibyte arg)))
287 With this erroneous macro definition, the program worked fine when
288 interpreted but failed when compiled. This macro definition called
289 @code{set-buffer-multibyte} during compilation, which was wrong, and
290 then did nothing when the compiled package was run. The definition
291 that the programmer really wanted was this:
294 (defmacro my-set-buffer-multibyte (arg)
295 (if (fboundp 'set-buffer-multibyte)
296 `(set-buffer-multibyte ,arg)))
300 This macro expands, if appropriate, into a call to
301 @code{set-buffer-multibyte} that will be executed when the compiled
302 program is actually run.
304 @node Argument Evaluation
305 @subsection Evaluating Macro Arguments Repeatedly
307 When defining a macro you must pay attention to the number of times
308 the arguments will be evaluated when the expansion is executed. The
309 following macro (used to facilitate iteration) illustrates the
310 problem. This macro allows us to write a ``for'' loop construct.
315 (defmacro for (var from init to final do &rest body)
316 "Execute a simple \"for\" loop.
317 For example, (for i from 1 to 10 do (print i))."
318 (list 'let (list (list var init))
320 (cons (list '<= var final)
321 (append body (list (list 'inc var)))))))
325 (for i from 1 to 3 do
326 (setq square (* i i))
327 (princ (format "\n%d %d" i square)))
333 (setq square (* i i))
334 (princ (format "\n%d %d" i square))
347 The arguments @code{from}, @code{to}, and @code{do} in this macro are
348 ``syntactic sugar''; they are entirely ignored. The idea is that you
349 will write noise words (such as @code{from}, @code{to}, and @code{do})
350 in those positions in the macro call.
352 Here's an equivalent definition simplified through use of backquote:
356 (defmacro for (var from init to final do &rest body)
357 "Execute a simple \"for\" loop.
358 For example, (for i from 1 to 10 do (print i))."
360 (while (<= ,var ,final)
366 Both forms of this definition (with backquote and without) suffer from
367 the defect that @var{final} is evaluated on every iteration. If
368 @var{final} is a constant, this is not a problem. If it is a more
369 complex form, say @code{(long-complex-calculation x)}, this can slow
370 down the execution significantly. If @var{final} has side effects,
371 executing it more than once is probably incorrect.
373 @cindex macro argument evaluation
374 A well-designed macro definition takes steps to avoid this problem by
375 producing an expansion that evaluates the argument expressions exactly
376 once unless repeated evaluation is part of the intended purpose of the
377 macro. Here is a correct expansion for the @code{for} macro:
384 (setq square (* i i))
385 (princ (format "%d %d" i square))
390 Here is a macro definition that creates this expansion:
394 (defmacro for (var from init to final do &rest body)
395 "Execute a simple for loop: (for i from 1 to 10 do (print i))."
404 Unfortunately, this fix introduces another problem,
405 described in the following section.
407 @node Surprising Local Vars
408 @subsection Local Variables in Macro Expansions
411 In the previous section, the definition of @code{for} was fixed as
412 follows to make the expansion evaluate the macro arguments the proper
417 (defmacro for (var from init to final do &rest body)
418 "Execute a simple for loop: (for i from 1 to 10 do (print i))."
430 The new definition of @code{for} has a new problem: it introduces a
431 local variable named @code{max} which the user does not expect. This
432 causes trouble in examples such as the following:
437 (for x from 0 to 10 do
438 (let ((this (frob x)))
445 The references to @code{max} inside the body of the @code{for}, which
446 are supposed to refer to the user's binding of @code{max}, really access
447 the binding made by @code{for}.
449 The way to correct this is to use an uninterned symbol instead of
450 @code{max} (@pxref{Creating Symbols}). The uninterned symbol can be
451 bound and referred to just like any other symbol, but since it is
452 created by @code{for}, we know that it cannot already appear in the
453 user's program. Since it is not interned, there is no way the user can
454 put it into the program later. It will never appear anywhere except
455 where put by @code{for}. Here is a definition of @code{for} that works
460 (defmacro for (var from init to final do &rest body)
461 "Execute a simple for loop: (for i from 1 to 10 do (print i))."
462 (let ((tempvar (make-symbol "max")))
465 (while (<= ,var ,tempvar)
472 This creates an uninterned symbol named @code{max} and puts it in the
473 expansion instead of the usual interned symbol @code{max} that appears
474 in expressions ordinarily.
476 @node Eval During Expansion
477 @subsection Evaluating Macro Arguments in Expansion
479 Another problem can happen if the macro definition itself
480 evaluates any of the macro argument expressions, such as by calling
481 @code{eval} (@pxref{Eval}). If the argument is supposed to refer to the
482 user's variables, you may have trouble if the user happens to use a
483 variable with the same name as one of the macro arguments. Inside the
484 macro body, the macro argument binding is the most local binding of this
485 variable, so any references inside the form being evaluated do refer to
486 it. Here is an example:
491 (list 'setq (eval a) t))
495 (foo x) @expansion{} (setq b t)
496 @result{} t ; @r{and @code{b} has been set.}
499 (foo a) @expansion{} (setq a t)
500 @result{} t ; @r{but this set @code{a}, not @code{c}.}
505 It makes a difference whether the user's variable is named @code{a} or
506 @code{x}, because @code{a} conflicts with the macro argument variable
509 Another problem with calling @code{eval} in a macro definition is that
510 it probably won't do what you intend in a compiled program. The
511 byte compiler runs macro definitions while compiling the program, when
512 the program's own computations (which you might have wished to access
513 with @code{eval}) don't occur and its local variable bindings don't
516 To avoid these problems, @strong{don't evaluate an argument expression
517 while computing the macro expansion}. Instead, substitute the
518 expression into the macro expansion, so that its value will be computed
519 as part of executing the expansion. This is how the other examples in
522 @node Repeated Expansion
523 @subsection How Many Times is the Macro Expanded?
525 Occasionally problems result from the fact that a macro call is
526 expanded each time it is evaluated in an interpreted function, but is
527 expanded only once (during compilation) for a compiled function. If the
528 macro definition has side effects, they will work differently depending
529 on how many times the macro is expanded.
531 Therefore, you should avoid side effects in computation of the
532 macro expansion, unless you really know what you are doing.
534 One special kind of side effect can't be avoided: constructing Lisp
535 objects. Almost all macro expansions include constructed lists; that is
536 the whole point of most macros. This is usually safe; there is just one
537 case where you must be careful: when the object you construct is part of a
538 quoted constant in the macro expansion.
540 If the macro is expanded just once, in compilation, then the object is
541 constructed just once, during compilation. But in interpreted
542 execution, the macro is expanded each time the macro call runs, and this
543 means a new object is constructed each time.
545 In most clean Lisp code, this difference won't matter. It can matter
546 only if you perform side-effects on the objects constructed by the macro
547 definition. Thus, to avoid trouble, @strong{avoid side effects on
548 objects constructed by macro definitions}. Here is an example of how
549 such side effects can get you into trouble:
553 (defmacro empty-object ()
554 (list 'quote (cons nil nil)))
558 (defun initialize (condition)
559 (let ((object (empty-object)))
561 (setcar object condition))
567 If @code{initialize} is interpreted, a new list @code{(nil)} is
568 constructed each time @code{initialize} is called. Thus, no side effect
569 survives between calls. If @code{initialize} is compiled, then the
570 macro @code{empty-object} is expanded during compilation, producing a
571 single ``constant'' @code{(nil)} that is reused and altered each time
572 @code{initialize} is called.
574 One way to avoid pathological cases like this is to think of
575 @code{empty-object} as a funny kind of constant, not as a memory
576 allocation construct. You wouldn't use @code{setcar} on a constant such
577 as @code{'(nil)}, so naturally you won't use it on @code{(empty-object)}
580 @node Indenting Macros
581 @section Indenting Macros
583 Within a macro definition, you can use the @code{declare} form
584 (@pxref{Defining Macros}) to specify how @key{TAB} should indent
585 calls to the macro. An indentation specification is written like this:
588 (declare (indent @var{indent-spec}))
592 Here are the possibilities for @var{indent-spec}:
596 This is the same as no property---use the standard indentation pattern.
598 Handle this function like a @samp{def} construct: treat the second
599 line as the start of a @dfn{body}.
600 @item an integer, @var{number}
601 The first @var{number} arguments of the function are
602 @dfn{distinguished} arguments; the rest are considered the body
603 of the expression. A line in the expression is indented according to
604 whether the first argument on it is distinguished or not. If the
605 argument is part of the body, the line is indented @code{lisp-body-indent}
606 more columns than the open-parenthesis starting the containing
607 expression. If the argument is distinguished and is either the first
608 or second argument, it is indented @emph{twice} that many extra columns.
609 If the argument is distinguished and not the first or second argument,
610 the line uses the standard pattern.
611 @item a symbol, @var{symbol}
612 @var{symbol} should be a function name; that function is called to
613 calculate the indentation of a line within this expression. The
614 function receives two arguments:
618 The position at which the line being indented begins.
620 The value returned by @code{parse-partial-sexp} (a Lisp primitive for
621 indentation and nesting computation) when it parses up to the
622 beginning of this line.
626 It should return either a number, which is the number of columns of
627 indentation for that line, or a list whose car is such a number. The
628 difference between returning a number and returning a list is that a
629 number says that all following lines at the same nesting level should
630 be indented just like this one; a list says that following lines might
631 call for different indentations. This makes a difference when the
632 indentation is being computed by @kbd{C-M-q}; if the value is a
633 number, @kbd{C-M-q} need not recalculate indentation for the following
634 lines until the end of the list.