1 @c -*- mode: texinfo; coding: utf-8 -*-
2 @c This is part of the GNU Emacs Lisp Reference Manual.
3 @c Copyright (C) 1990-1995, 1998-1999, 2001-2017 Free Software
5 @c See the file elisp.texi for copying conditions.
9 A Lisp program is composed mainly of Lisp functions. This chapter
10 explains what functions are, how they accept arguments, and how to
14 * What Is a Function:: Lisp functions vs. primitives; terminology.
15 * Lambda Expressions:: How functions are expressed as Lisp objects.
16 * Function Names:: A symbol can serve as the name of a function.
17 * Defining Functions:: Lisp expressions for defining functions.
18 * Calling Functions:: How to use an existing function.
19 * Mapping Functions:: Applying a function to each element of a list, etc.
20 * Anonymous Functions:: Lambda expressions are functions with no names.
21 * Generic Functions:: Polymorphism, Emacs-style.
22 * Function Cells:: Accessing or setting the function definition
24 * Closures:: Functions that enclose a lexical environment.
25 * Advising Functions:: Adding to the definition of a function.
26 * Obsolete Functions:: Declaring functions obsolete.
27 * Inline Functions:: Functions that the compiler will expand inline.
28 * Declare Form:: Adding additional information about a function.
29 * Declaring Functions:: Telling the compiler that a function is defined.
30 * Function Safety:: Determining whether a function is safe to call.
31 * Related Topics:: Cross-references to specific Lisp primitives
32 that have a special bearing on how functions work.
35 @node What Is a Function
36 @section What Is a Function?
39 @cindex value of function
41 In a general sense, a function is a rule for carrying out a
42 computation given input values called @dfn{arguments}. The result of
43 the computation is called the @dfn{value} or @dfn{return value} of the
44 function. The computation can also have side effects, such as lasting
45 changes in the values of variables or the contents of data structures.
47 In most computer languages, every function has a name. But in Lisp,
48 a function in the strictest sense has no name: it is an object which
49 can @emph{optionally} be associated with a symbol (e.g., @code{car})
50 that serves as the function name. @xref{Function Names}. When a
51 function has been given a name, we usually also refer to that symbol
52 as a ``function'' (e.g., we refer to ``the function @code{car}'').
53 In this manual, the distinction between a function name and the
54 function object itself is usually unimportant, but we will take note
55 wherever it is relevant.
57 Certain function-like objects, called @dfn{special forms} and
58 @dfn{macros}, also accept arguments to carry out computations.
59 However, as explained below, these are not considered functions in
62 Here are important terms for functions and function-like objects:
65 @item lambda expression
66 A function (in the strict sense, i.e., a function object) which is
67 written in Lisp. These are described in the following section.
69 @xref{Lambda Expressions}.
75 @cindex built-in function
76 A function which is callable from Lisp but is actually written in C@.
77 Primitives are also called @dfn{built-in functions}, or @dfn{subrs}.
78 Examples include functions like @code{car} and @code{append}. In
79 addition, all special forms (see below) are also considered
82 Usually, a function is implemented as a primitive because it is a
83 fundamental part of Lisp (e.g., @code{car}), or because it provides a
84 low-level interface to operating system services, or because it needs
85 to run fast. Unlike functions defined in Lisp, primitives can be
86 modified or added only by changing the C sources and recompiling
87 Emacs. See @ref{Writing Emacs Primitives}.
90 A primitive that is like a function but does not evaluate all of its
91 arguments in the usual way. It may evaluate only some of the
92 arguments, or may evaluate them in an unusual order, or several times.
93 Examples include @code{if}, @code{and}, and @code{while}.
98 A construct defined in Lisp, which differs from a function in that it
99 translates a Lisp expression into another expression which is to be
100 evaluated instead of the original expression. Macros enable Lisp
101 programmers to do the sorts of things that special forms can do.
106 An object which can be invoked via the @code{command-execute}
107 primitive, usually due to the user typing in a key sequence
108 @dfn{bound} to that command. @xref{Interactive Call}. A command is
109 usually a function; if the function is written in Lisp, it is made
110 into a command by an @code{interactive} form in the function
111 definition (@pxref{Defining Commands}). Commands that are functions
112 can also be called from Lisp expressions, just like other functions.
114 Keyboard macros (strings and vectors) are commands also, even though
115 they are not functions. @xref{Keyboard Macros}. We say that a symbol
116 is a command if its function cell contains a command (@pxref{Symbol
117 Components}); such a @dfn{named command} can be invoked with
121 A function object that is much like a lambda expression, except that
122 it also encloses an environment of lexical variable bindings.
125 @item byte-code function
126 A function that has been compiled by the byte compiler.
127 @xref{Byte-Code Type}.
129 @item autoload object
130 @cindex autoload object
131 A place-holder for a real function. If the autoload object is called,
132 Emacs loads the file containing the definition of the real function,
133 and then calls the real function. @xref{Autoload}.
136 You can use the function @code{functionp} to test if an object is a
139 @defun functionp object
140 This function returns @code{t} if @var{object} is any kind of
141 function, i.e., can be passed to @code{funcall}. Note that
142 @code{functionp} returns @code{t} for symbols that are function names,
143 and returns @code{nil} for special forms.
146 It is also possible to find out how many arguments an arbitrary
149 @defun func-arity function
150 This function provides information about the argument list of the
151 specified @var{function}. The returned value is a cons cell of the
152 form @w{@code{(@var{min} . @var{max})}}, where @var{min} is the
153 minimum number of arguments, and @var{max} is either the maximum
154 number of arguments, or the symbol @code{many} for functions with
155 @code{&rest} arguments, or the symbol @code{unevalled} if
156 @var{function} is a special form.
158 Note that this function might return inaccurate results in some
159 situations, such as the following:
163 Functions defined using @code{apply-partially} (@pxref{Calling
164 Functions, apply-partially}).
167 Functions that are advised using @code{advice-add} (@pxref{Advising
171 Functions that determine the argument list dynamically, as part of
178 Unlike @code{functionp}, the next three functions do @emph{not} treat
179 a symbol as its function definition.
182 This function returns @code{t} if @var{object} is a built-in function
183 (i.e., a Lisp primitive).
187 (subrp 'message) ; @r{@code{message} is a symbol,}
188 @result{} nil ; @r{not a subr object.}
191 (subrp (symbol-function 'message))
197 @defun byte-code-function-p object
198 This function returns @code{t} if @var{object} is a byte-code
199 function. For example:
203 (byte-code-function-p (symbol-function 'next-line))
209 @defun subr-arity subr
210 This works like @code{func-arity}, but only for built-in functions and
211 without symbol indirection. It signals an error for non-built-in
212 functions. We recommend to use @code{func-arity} instead.
215 @node Lambda Expressions
216 @section Lambda Expressions
217 @cindex lambda expression
219 A lambda expression is a function object written in Lisp. Here is
224 "Return the hyperbolic cosine of X."
225 (* 0.5 (+ (exp x) (exp (- x)))))
229 In Emacs Lisp, such a list is a valid expression which evaluates to
232 A lambda expression, by itself, has no name; it is an @dfn{anonymous
233 function}. Although lambda expressions can be used this way
234 (@pxref{Anonymous Functions}), they are more commonly associated with
235 symbols to make @dfn{named functions} (@pxref{Function Names}).
236 Before going into these details, the following subsections describe
237 the components of a lambda expression and what they do.
240 * Lambda Components:: The parts of a lambda expression.
241 * Simple Lambda:: A simple example.
242 * Argument List:: Details and special features of argument lists.
243 * Function Documentation:: How to put documentation in a function.
246 @node Lambda Components
247 @subsection Components of a Lambda Expression
249 A lambda expression is a list that looks like this:
252 (lambda (@var{arg-variables}@dots{})
253 [@var{documentation-string}]
254 [@var{interactive-declaration}]
255 @var{body-forms}@dots{})
259 The first element of a lambda expression is always the symbol
260 @code{lambda}. This indicates that the list represents a function. The
261 reason functions are defined to start with @code{lambda} is so that
262 other lists, intended for other uses, will not accidentally be valid as
265 The second element is a list of symbols---the argument variable names.
266 This is called the @dfn{lambda list}. When a Lisp function is called,
267 the argument values are matched up against the variables in the lambda
268 list, which are given local bindings with the values provided.
269 @xref{Local Variables}.
271 The documentation string is a Lisp string object placed within the
272 function definition to describe the function for the Emacs help
273 facilities. @xref{Function Documentation}.
275 The interactive declaration is a list of the form @code{(interactive
276 @var{code-string})}. This declares how to provide arguments if the
277 function is used interactively. Functions with this declaration are called
278 @dfn{commands}; they can be called using @kbd{M-x} or bound to a key.
279 Functions not intended to be called in this way should not have interactive
280 declarations. @xref{Defining Commands}, for how to write an interactive
283 @cindex body of function
284 The rest of the elements are the @dfn{body} of the function: the Lisp
285 code to do the work of the function (or, as a Lisp programmer would say,
286 ``a list of Lisp forms to evaluate''). The value returned by the
287 function is the value returned by the last element of the body.
290 @subsection A Simple Lambda Expression Example
292 Consider the following example:
295 (lambda (a b c) (+ a b c))
299 We can call this function by passing it to @code{funcall}, like this:
303 (funcall (lambda (a b c) (+ a b c))
309 This call evaluates the body of the lambda expression with the variable
310 @code{a} bound to 1, @code{b} bound to 2, and @code{c} bound to 3.
311 Evaluation of the body adds these three numbers, producing the result 6;
312 therefore, this call to the function returns the value 6.
314 Note that the arguments can be the results of other function calls, as in
319 (funcall (lambda (a b c) (+ a b c))
325 This evaluates the arguments @code{1}, @code{(* 2 3)}, and @code{(- 5
326 4)} from left to right. Then it applies the lambda expression to the
327 argument values 1, 6 and 1 to produce the value 8.
329 As these examples show, you can use a form with a lambda expression
330 as its @sc{car} to make local variables and give them values. In the
331 old days of Lisp, this technique was the only way to bind and
332 initialize local variables. But nowadays, it is clearer to use the
333 special form @code{let} for this purpose (@pxref{Local Variables}).
334 Lambda expressions are mainly used as anonymous functions for passing
335 as arguments to other functions (@pxref{Anonymous Functions}), or
336 stored as symbol function definitions to produce named functions
337 (@pxref{Function Names}).
340 @subsection Other Features of Argument Lists
341 @kindex wrong-number-of-arguments
342 @cindex argument binding
343 @cindex binding arguments
344 @cindex argument lists, features
346 Our simple sample function, @code{(lambda (a b c) (+ a b c))},
347 specifies three argument variables, so it must be called with three
348 arguments: if you try to call it with only two arguments or four
349 arguments, you get a @code{wrong-number-of-arguments} error
352 It is often convenient to write a function that allows certain
353 arguments to be omitted. For example, the function @code{substring}
354 accepts three arguments---a string, the start index and the end
355 index---but the third argument defaults to the @var{length} of the
356 string if you omit it. It is also convenient for certain functions to
357 accept an indefinite number of arguments, as the functions @code{list}
360 @cindex optional arguments
361 @cindex rest arguments
364 To specify optional arguments that may be omitted when a function
365 is called, simply include the keyword @code{&optional} before the optional
366 arguments. To specify a list of zero or more extra arguments, include the
367 keyword @code{&rest} before one final argument.
369 Thus, the complete syntax for an argument list is as follows:
373 (@var{required-vars}@dots{}
374 @r{[}&optional @var{optional-vars}@dots{}@r{]}
375 @r{[}&rest @var{rest-var}@r{]})
380 The square brackets indicate that the @code{&optional} and @code{&rest}
381 clauses, and the variables that follow them, are optional.
383 A call to the function requires one actual argument for each of the
384 @var{required-vars}. There may be actual arguments for zero or more of
385 the @var{optional-vars}, and there cannot be any actual arguments beyond
386 that unless the lambda list uses @code{&rest}. In that case, there may
387 be any number of extra actual arguments.
389 If actual arguments for the optional and rest variables are omitted,
390 then they always default to @code{nil}. There is no way for the
391 function to distinguish between an explicit argument of @code{nil} and
392 an omitted argument. However, the body of the function is free to
393 consider @code{nil} an abbreviation for some other meaningful value.
394 This is what @code{substring} does; @code{nil} as the third argument to
395 @code{substring} means to use the length of the string supplied.
397 @cindex CL note---default optional arg
399 @b{Common Lisp note:} Common Lisp allows the function to specify what
400 default value to use when an optional argument is omitted; Emacs Lisp
401 always uses @code{nil}. Emacs Lisp does not support @code{supplied-p}
402 variables that tell you whether an argument was explicitly passed.
405 For example, an argument list that looks like this:
408 (a b &optional c d &rest e)
412 binds @code{a} and @code{b} to the first two actual arguments, which are
413 required. If one or two more arguments are provided, @code{c} and
414 @code{d} are bound to them respectively; any arguments after the first
415 four are collected into a list and @code{e} is bound to that list. If
416 there are only two arguments, @code{c} is @code{nil}; if two or three
417 arguments, @code{d} is @code{nil}; if four arguments or fewer, @code{e}
420 There is no way to have required arguments following optional
421 ones---it would not make sense. To see why this must be so, suppose
422 that @code{c} in the example were optional and @code{d} were required.
423 Suppose three actual arguments are given; which variable would the
424 third argument be for? Would it be used for the @var{c}, or for
425 @var{d}? One can argue for both possibilities. Similarly, it makes
426 no sense to have any more arguments (either required or optional)
427 after a @code{&rest} argument.
429 Here are some examples of argument lists and proper calls:
432 (funcall (lambda (n) (1+ n)) ; @r{One required:}
433 1) ; @r{requires exactly one argument.}
435 (funcall (lambda (n &optional n1) ; @r{One required and one optional:}
436 (if n1 (+ n n1) (1+ n))) ; @r{1 or 2 arguments.}
439 (funcall (lambda (n &rest ns) ; @r{One required and one rest:}
440 (+ n (apply '+ ns))) ; @r{1 or more arguments.}
445 @node Function Documentation
446 @subsection Documentation Strings of Functions
447 @cindex documentation of function
449 A lambda expression may optionally have a @dfn{documentation string}
450 just after the lambda list. This string does not affect execution of
451 the function; it is a kind of comment, but a systematized comment
452 which actually appears inside the Lisp world and can be used by the
453 Emacs help facilities. @xref{Documentation}, for how the
454 documentation string is accessed.
456 It is a good idea to provide documentation strings for all the
457 functions in your program, even those that are called only from within
458 your program. Documentation strings are like comments, except that they
459 are easier to access.
461 The first line of the documentation string should stand on its own,
462 because @code{apropos} displays just this first line. It should consist
463 of one or two complete sentences that summarize the function's purpose.
465 The start of the documentation string is usually indented in the
466 source file, but since these spaces come before the starting
467 double-quote, they are not part of the string. Some people make a
468 practice of indenting any additional lines of the string so that the
469 text lines up in the program source. @emph{That is a mistake.} The
470 indentation of the following lines is inside the string; what looks
471 nice in the source code will look ugly when displayed by the help
474 You may wonder how the documentation string could be optional, since
475 there are required components of the function that follow it (the body).
476 Since evaluation of a string returns that string, without any side effects,
477 it has no effect if it is not the last form in the body. Thus, in
478 practice, there is no confusion between the first form of the body and the
479 documentation string; if the only body form is a string then it serves both
480 as the return value and as the documentation.
482 The last line of the documentation string can specify calling
483 conventions different from the actual function arguments. Write
491 following a blank line, at the beginning of the line, with no newline
492 following it inside the documentation string. (The @samp{\} is used
493 to avoid confusing the Emacs motion commands.) The calling convention
494 specified in this way appears in help messages in place of the one
495 derived from the actual arguments of the function.
497 This feature is particularly useful for macro definitions, since the
498 arguments written in a macro definition often do not correspond to the
499 way users think of the parts of the macro call.
502 @section Naming a Function
503 @cindex function definition
504 @cindex named function
505 @cindex function name
507 A symbol can serve as the name of a function. This happens when the
508 symbol's @dfn{function cell} (@pxref{Symbol Components}) contains a
509 function object (e.g., a lambda expression). Then the symbol itself
510 becomes a valid, callable function, equivalent to the function object
511 in its function cell.
513 The contents of the function cell are also called the symbol's
514 @dfn{function definition}. The procedure of using a symbol's function
515 definition in place of the symbol is called @dfn{symbol function
516 indirection}; see @ref{Function Indirection}. If you have not given a
517 symbol a function definition, its function cell is said to be
518 @dfn{void}, and it cannot be used as a function.
520 In practice, nearly all functions have names, and are referred to by
521 their names. You can create a named Lisp function by defining a
522 lambda expression and putting it in a function cell (@pxref{Function
523 Cells}). However, it is more common to use the @code{defun} special
524 form, described in the next section.
526 @xref{Defining Functions}.
529 We give functions names because it is convenient to refer to them by
530 their names in Lisp expressions. Also, a named Lisp function can
531 easily refer to itself---it can be recursive. Furthermore, primitives
532 can only be referred to textually by their names, since primitive
533 function objects (@pxref{Primitive Function Type}) have no read
536 A function need not have a unique name. A given function object
537 @emph{usually} appears in the function cell of only one symbol, but
538 this is just a convention. It is easy to store it in several symbols
539 using @code{fset}; then each of the symbols is a valid name for the
542 Note that a symbol used as a function name may also be used as a
543 variable; these two uses of a symbol are independent and do not
544 conflict. (This is not the case in some dialects of Lisp, like
547 By convention, if a function's symbol consists of two names
548 separated by @samp{--}, the function is intended for internal use and
549 the first part names the file defining the function. For example, a
550 function named @code{vc-git--rev-parse} is an internal function
551 defined in @file{vc-git.el}. Internal-use functions written in C have
552 names ending in @samp{-internal}, e.g., @code{bury-buffer-internal}.
553 Emacs code contributed before 2018 may follow other internal-use
554 naming conventions, which are being phased out.
556 @node Defining Functions
557 @section Defining Functions
558 @cindex defining a function
560 We usually give a name to a function when it is first created. This
561 is called @dfn{defining a function}, and it is done with the
564 @defmac defun name args [doc] [declare] [interactive] body@dots{}
565 @code{defun} is the usual way to define new Lisp functions. It
566 defines the symbol @var{name} as a function with argument list
567 @var{args} and body forms given by @var{body}. Neither @var{name} nor
568 @var{args} should be quoted.
570 @var{doc}, if present, should be a string specifying the function's
571 documentation string (@pxref{Function Documentation}). @var{declare},
572 if present, should be a @code{declare} form specifying function
573 metadata (@pxref{Declare Form}). @var{interactive}, if present,
574 should be an @code{interactive} form specifying how the function is to
575 be called interactively (@pxref{Interactive Call}).
577 The return value of @code{defun} is undefined.
579 Here are some examples:
589 (defun bar (a &optional b &rest c)
592 @result{} (1 2 (3 4 5))
596 @result{} (1 nil nil)
600 @error{} Wrong number of arguments.
604 (defun capitalize-backwards ()
605 "Upcase the last letter of the word at point."
614 Be careful not to redefine existing functions unintentionally.
615 @code{defun} redefines even primitive functions such as @code{car}
616 without any hesitation or notification. Emacs does not prevent you
617 from doing this, because redefining a function is sometimes done
618 deliberately, and there is no way to distinguish deliberate
619 redefinition from unintentional redefinition.
622 @cindex function aliases
623 @cindex alias, for functions
624 @defun defalias name definition &optional doc
625 @anchor{Definition of defalias}
626 This function defines the symbol @var{name} as a function, with
627 definition @var{definition} (which can be any valid Lisp function).
628 Its return value is @emph{undefined}.
630 If @var{doc} is non-@code{nil}, it becomes the function documentation
631 of @var{name}. Otherwise, any documentation provided by
632 @var{definition} is used.
634 @cindex defalias-fset-function property
635 Internally, @code{defalias} normally uses @code{fset} to set the definition.
636 If @var{name} has a @code{defalias-fset-function} property, however,
637 the associated value is used as a function to call in place of @code{fset}.
639 The proper place to use @code{defalias} is where a specific function
640 name is being defined---especially where that name appears explicitly in
641 the source file being loaded. This is because @code{defalias} records
642 which file defined the function, just like @code{defun}
645 By contrast, in programs that manipulate function definitions for other
646 purposes, it is better to use @code{fset}, which does not keep such
647 records. @xref{Function Cells}.
650 You cannot create a new primitive function with @code{defun} or
651 @code{defalias}, but you can use them to change the function definition of
652 any symbol, even one such as @code{car} or @code{x-popup-menu} whose
653 normal definition is a primitive. However, this is risky: for
654 instance, it is next to impossible to redefine @code{car} without
655 breaking Lisp completely. Redefining an obscure function such as
656 @code{x-popup-menu} is less dangerous, but it still may not work as
657 you expect. If there are calls to the primitive from C code, they
658 call the primitive's C definition directly, so changing the symbol's
659 definition will have no effect on them.
661 See also @code{defsubst}, which defines a function like @code{defun}
662 and tells the Lisp compiler to perform inline expansion on it.
663 @xref{Inline Functions}.
665 Alternatively, you can define a function by providing the code which
666 will inline it as a compiler macro. The following macros make this
669 @c FIXME: Can define-inline use the interactive spec?
670 @defmac define-inline name args [doc] [declare] body@dots{}
671 Define a function @var{name} by providing code that does its inlining,
672 as a compiler macro. The function will accept the argument list
673 @var{args} and will have the specified @var{body}.
675 If present, @var{doc} should be the function's documentation string
676 (@pxref{Function Documentation}); @var{declare}, if present, should be
677 a @code{declare} form (@pxref{Declare Form}) specifying the function's
681 Functions defined via @code{define-inline} have several advantages
682 with respect to macros defined by @code{defsubst} or @code{defmacro}:
686 They can be passed to @code{mapcar} (@pxref{Mapping Functions}).
689 They are more efficient.
692 They can be used as @dfn{place forms} to store values
693 (@pxref{Generalized Variables}).
696 They behave in a more predictable way than @code{cl-defsubst}
697 (@pxref{Argument Lists,,, cl, Common Lisp Extensions for GNU Emacs
701 Like @code{defmacro}, a function inlined with @code{define-inline}
702 inherits the scoping rules, either dynamic or lexical, from the call
703 site. @xref{Variable Scoping}.
705 The following macros should be used in the body of a function defined
706 by @code{define-inline}.
708 @defmac inline-quote expression
709 Quote @var{expression} for @code{define-inline}. This is similar to
710 the backquote (@pxref{Backquote}), but quotes code and accepts only
711 @code{,}, not @code{,@@}.
714 @defmac inline-letevals (bindings@dots{}) body@dots{}
715 This is similar to @code{let} (@pxref{Local Variables}): it sets up
716 local variables as specified by @var{bindings}, and then evaluates
717 @var{body} with those bindings in effect. Each element of
718 @var{bindings} should be either a symbol or a list of the form
719 @w{@code{(@var{var} @var{expr})}}; the result is to evaluate
720 @var{expr} and bind @var{var} to the result. The tail of
721 @var{bindings} can be either @code{nil} or a symbol which should hold
722 a list of arguments, in which case each argument is evaluated, and the
723 symbol is bound to the resulting list.
726 @defmac inline-const-p expression
727 Return non-@code{nil} if the value of @var{expression} is already
731 @defmac inline-const-val expression
732 Return the value of @var{expression}.
735 @defmac inline-error format &rest args
736 Signal an error, formatting @var{args} according to @var{format}.
739 Here's an example of using @code{define-inline}:
742 (define-inline myaccessor (obj)
743 (inline-letevals (obj)
744 (inline-quote (if (foo-p ,obj) (aref (cdr ,obj) 3) (aref ,obj 2)))))
748 This is equivalent to
751 (defsubst myaccessor (obj)
752 (if (foo-p obj) (aref (cdr obj) 3) (aref obj 2)))
755 @node Calling Functions
756 @section Calling Functions
757 @cindex function invocation
758 @cindex calling a function
760 Defining functions is only half the battle. Functions don't do
761 anything until you @dfn{call} them, i.e., tell them to run. Calling a
762 function is also known as @dfn{invocation}.
764 The most common way of invoking a function is by evaluating a list.
765 For example, evaluating the list @code{(concat "a" "b")} calls the
766 function @code{concat} with arguments @code{"a"} and @code{"b"}.
767 @xref{Evaluation}, for a description of evaluation.
769 When you write a list as an expression in your program, you specify
770 which function to call, and how many arguments to give it, in the text
771 of the program. Usually that's just what you want. Occasionally you
772 need to compute at run time which function to call. To do that, use
773 the function @code{funcall}. When you also need to determine at run
774 time how many arguments to pass, use @code{apply}.
776 @defun funcall function &rest arguments
777 @code{funcall} calls @var{function} with @var{arguments}, and returns
778 whatever @var{function} returns.
780 Since @code{funcall} is a function, all of its arguments, including
781 @var{function}, are evaluated before @code{funcall} is called. This
782 means that you can use any expression to obtain the function to be
783 called. It also means that @code{funcall} does not see the
784 expressions you write for the @var{arguments}, only their values.
785 These values are @emph{not} evaluated a second time in the act of
786 calling @var{function}; the operation of @code{funcall} is like the
787 normal procedure for calling a function, once its arguments have
788 already been evaluated.
790 The argument @var{function} must be either a Lisp function or a
791 primitive function. Special forms and macros are not allowed, because
792 they make sense only when given the unevaluated argument
793 expressions. @code{funcall} cannot provide these because, as we saw
794 above, it never knows them in the first place.
796 If you need to use @code{funcall} to call a command and make it behave
797 as if invoked interactively, use @code{funcall-interactively}
798 (@pxref{Interactive Call}).
810 (funcall f 'x 'y '(z))
815 @error{} Invalid function: #<subr and>
819 Compare these examples with the examples of @code{apply}.
822 @defun apply function &rest arguments
823 @code{apply} calls @var{function} with @var{arguments}, just like
824 @code{funcall} but with one difference: the last of @var{arguments} is a
825 list of objects, which are passed to @var{function} as separate
826 arguments, rather than a single list. We say that @code{apply}
827 @dfn{spreads} this list so that each individual element becomes an
830 @code{apply} returns the result of calling @var{function}. As with
831 @code{funcall}, @var{function} must either be a Lisp function or a
832 primitive function; special forms and macros do not make sense in
842 @error{} Wrong type argument: listp, z
845 (apply '+ 1 2 '(3 4))
849 (apply '+ '(1 2 3 4))
854 (apply 'append '((a b c) nil (x y z) nil))
855 @result{} (a b c x y z)
859 For an interesting example of using @code{apply}, see @ref{Definition
863 @cindex partial application of functions
865 Sometimes it is useful to fix some of the function's arguments at
866 certain values, and leave the rest of arguments for when the function
867 is actually called. The act of fixing some of the function's
868 arguments is called @dfn{partial application} of the function@footnote{
869 This is related to, but different from @dfn{currying}, which
870 transforms a function that takes multiple arguments in such a way that
871 it can be called as a chain of functions, each one with a single
873 The result is a new function that accepts the rest of
874 arguments and calls the original function with all the arguments
877 Here's how to do partial application in Emacs Lisp:
879 @defun apply-partially func &rest args
880 This function returns a new function which, when called, will call
881 @var{func} with the list of arguments composed from @var{args} and
882 additional arguments specified at the time of the call. If @var{func}
883 accepts @var{n} arguments, then a call to @code{apply-partially} with
884 @w{@code{@var{m} < @var{n}}} arguments will produce a new function of
885 @w{@code{@var{n} - @var{m}}} arguments.
887 Here's how we could define the built-in function @code{1+}, if it
888 didn't exist, using @code{apply-partially} and @code{+}, another
893 (defalias '1+ (apply-partially '+ 1)
894 "Increment argument by one.")
904 It is common for Lisp functions to accept functions as arguments or
905 find them in data structures (especially in hook variables and property
906 lists) and call them using @code{funcall} or @code{apply}. Functions
907 that accept function arguments are often called @dfn{functionals}.
909 Sometimes, when you call a functional, it is useful to supply a no-op
910 function as the argument. Here are two different kinds of no-op
914 This function returns @var{arg} and has no side effects.
917 @defun ignore &rest args
918 This function ignores any arguments and returns @code{nil}.
921 Some functions are user-visible @dfn{commands}, which can be called
922 interactively (usually by a key sequence). It is possible to invoke
923 such a command exactly as though it was called interactively, by using
924 the @code{call-interactively} function. @xref{Interactive Call}.
926 @node Mapping Functions
927 @section Mapping Functions
928 @cindex mapping functions
930 A @dfn{mapping function} applies a given function (@emph{not} a
931 special form or macro) to each element of a list or other collection.
932 Emacs Lisp has several such functions; this section describes
933 @code{mapcar}, @code{mapc}, @code{mapconcat}, and @code{mapcan}, which
934 map over a list. @xref{Definition of mapatoms}, for the function
935 @code{mapatoms} which maps over the symbols in an obarray.
936 @xref{Definition of maphash}, for the function @code{maphash} which
937 maps over key/value associations in a hash table.
939 These mapping functions do not allow char-tables because a char-table
940 is a sparse array whose nominal range of indices is very large. To map
941 over a char-table in a way that deals properly with its sparse nature,
942 use the function @code{map-char-table} (@pxref{Char-Tables}).
944 @defun mapcar function sequence
945 @anchor{Definition of mapcar}
946 @code{mapcar} applies @var{function} to each element of @var{sequence}
947 in turn, and returns a list of the results.
949 The argument @var{sequence} can be any kind of sequence except a
950 char-table; that is, a list, a vector, a bool-vector, or a string. The
951 result is always a list. The length of the result is the same as the
952 length of @var{sequence}. For example:
956 (mapcar 'car '((a b) (c d) (e f)))
960 (mapcar 'string "abc")
961 @result{} ("a" "b" "c")
965 ;; @r{Call each function in @code{my-hooks}.}
966 (mapcar 'funcall my-hooks)
970 (defun mapcar* (function &rest args)
971 "Apply FUNCTION to successive cars of all ARGS.
972 Return the list of results."
973 ;; @r{If no list is exhausted,}
974 (if (not (memq nil args))
975 ;; @r{apply function to @sc{car}s.}
976 (cons (apply function (mapcar 'car args))
977 (apply 'mapcar* function
978 ;; @r{Recurse for rest of elements.}
979 (mapcar 'cdr args)))))
983 (mapcar* 'cons '(a b c) '(1 2 3 4))
984 @result{} ((a . 1) (b . 2) (c . 3))
989 @defun mapcan function sequence
990 This function applies @var{function} to each element of
991 @var{sequence}, like @code{mapcar}, but instead of collecting the
992 results into a list, it returns a single list with all the elements of
993 the results (which must be lists), by altering the results (using
994 @code{nconc}; @pxref{Rearrangement}). Like with @code{mapcar},
995 @var{sequence} can be of any type except a char-table.
999 ;; @r{Contrast this:}
1000 (mapcar 'list '(a b c d))
1001 @result{} ((a) (b) (c) (d))
1003 (mapcan 'list '(a b c d))
1009 @defun mapc function sequence
1010 @code{mapc} is like @code{mapcar} except that @var{function} is used for
1011 side-effects only---the values it returns are ignored, not collected
1012 into a list. @code{mapc} always returns @var{sequence}.
1015 @defun mapconcat function sequence separator
1016 @code{mapconcat} applies @var{function} to each element of
1017 @var{sequence}; the results, which must be sequences of characters
1018 (strings, vectors, or lists), are concatenated into a single string
1019 return value. Between each pair of result sequences, @code{mapconcat}
1020 inserts the characters from @var{separator}, which also must be a
1021 string, or a vector or list of characters. @xref{Sequences Arrays
1024 The argument @var{function} must be a function that can take one
1025 argument and returns a sequence of characters: a string, a vector, or
1026 a list. The argument @var{sequence} can be any kind of sequence
1027 except a char-table; that is, a list, a vector, a bool-vector, or a
1032 (mapconcat 'symbol-name
1033 '(The cat in the hat)
1035 @result{} "The cat in the hat"
1039 (mapconcat (function (lambda (x) (format "%c" (1+ x))))
1042 @result{} "IBM.9111"
1047 @node Anonymous Functions
1048 @section Anonymous Functions
1049 @cindex anonymous function
1051 Although functions are usually defined with @code{defun} and given
1052 names at the same time, it is sometimes convenient to use an explicit
1053 lambda expression---an @dfn{anonymous function}. Anonymous functions
1054 are valid wherever function names are. They are often assigned as
1055 variable values, or as arguments to functions; for instance, you might
1056 pass one as the @var{function} argument to @code{mapcar}, which
1057 applies that function to each element of a list (@pxref{Mapping
1058 Functions}). @xref{describe-symbols example}, for a realistic example
1061 When defining a lambda expression that is to be used as an anonymous
1062 function, you can in principle use any method to construct the list.
1063 But typically you should use the @code{lambda} macro, or the
1064 @code{function} special form, or the @code{#'} read syntax:
1066 @defmac lambda args [doc] [interactive] body@dots{}
1067 This macro returns an anonymous function with argument list
1068 @var{args}, documentation string @var{doc} (if any), interactive spec
1069 @var{interactive} (if any), and body forms given by @var{body}.
1071 In effect, this macro makes @code{lambda} forms self-quoting:
1072 evaluating a form whose @sc{car} is @code{lambda} yields the form
1076 (lambda (x) (* x x))
1077 @result{} (lambda (x) (* x x))
1080 The @code{lambda} form has one other effect: it tells the Emacs
1081 evaluator and byte-compiler that its argument is a function, by using
1082 @code{function} as a subroutine (see below).
1085 @defspec function function-object
1086 @cindex function quoting
1087 This special form returns @var{function-object} without evaluating it.
1088 In this, it is similar to @code{quote} (@pxref{Quoting}). But unlike
1089 @code{quote}, it also serves as a note to the Emacs evaluator and
1090 byte-compiler that @var{function-object} is intended to be used as a
1091 function. Assuming @var{function-object} is a valid lambda
1092 expression, this has two effects:
1096 When the code is byte-compiled, @var{function-object} is compiled into
1097 a byte-code function object (@pxref{Byte Compilation}).
1100 When lexical binding is enabled, @var{function-object} is converted
1101 into a closure. @xref{Closures}.
1105 @cindex @samp{#'} syntax
1106 The read syntax @code{#'} is a short-hand for using @code{function}.
1107 The following forms are all equivalent:
1110 (lambda (x) (* x x))
1111 (function (lambda (x) (* x x)))
1112 #'(lambda (x) (* x x))
1115 In the following example, we define a @code{change-property}
1116 function that takes a function as its third argument, followed by a
1117 @code{double-property} function that makes use of
1118 @code{change-property} by passing it an anonymous function:
1122 (defun change-property (symbol prop function)
1123 (let ((value (get symbol prop)))
1124 (put symbol prop (funcall function value))))
1128 (defun double-property (symbol prop)
1129 (change-property symbol prop (lambda (x) (* 2 x))))
1134 Note that we do not quote the @code{lambda} form.
1136 If you compile the above code, the anonymous function is also
1137 compiled. This would not happen if, say, you had constructed the
1138 anonymous function by quoting it as a list:
1140 @c Do not unquote this lambda!
1143 (defun double-property (symbol prop)
1144 (change-property symbol prop '(lambda (x) (* 2 x))))
1149 In that case, the anonymous function is kept as a lambda expression in
1150 the compiled code. The byte-compiler cannot assume this list is a
1151 function, even though it looks like one, since it does not know that
1152 @code{change-property} intends to use it as a function.
1154 @node Generic Functions
1155 @section Generic Functions
1156 @cindex generic functions
1157 @cindex polymorphism
1159 Functions defined using @code{defun} have a hard-coded set of
1160 assumptions about the types and expected values of their arguments.
1161 For example, a function that was designed to handle values of its
1162 argument that are either numbers or lists of numbers will fail or
1163 signal an error if called with a value of any other type, such as a
1164 vector or a string. This happens because the implementation of the
1165 function is not prepared to deal with types other than those assumed
1168 By contrast, object-oriented programs use @dfn{polymorphic
1169 functions}: a set of specialized functions having the same name, each
1170 one of which was written for a certain specific set of argument types.
1171 Which of the functions is actually called is decided at run time based
1172 on the types of the actual arguments.
1175 Emacs provides support for polymorphism. Like other Lisp
1176 environments, notably Common Lisp and its Common Lisp Object System
1177 (@acronym{CLOS}), this support is based on @dfn{generic functions}.
1178 The Emacs generic functions closely follow @acronym{CLOS}, including
1179 use of similar names, so if you have experience with @acronym{CLOS},
1180 the rest of this section will sound very familiar.
1182 A generic function specifies an abstract operation, by defining its
1183 name and list of arguments, but (usually) no implementation. The
1184 actual implementation for several specific classes of arguments is
1185 provided by @dfn{methods}, which should be defined separately. Each
1186 method that implements a generic function has the same name as the
1187 generic function, but the method's definition indicates what kinds of
1188 arguments it can handle by @dfn{specializing} the arguments defined by
1189 the generic function. These @dfn{argument specializers} can be more
1190 or less specific; for example, a @code{string} type is more specific
1191 than a more general type, such as @code{sequence}.
1193 Note that, unlike in message-based OO languages, such as C@t{++} and
1194 Simula, methods that implement generic functions don't belong to a
1195 class, they belong to the generic function they implement.
1197 When a generic function is invoked, it selects the applicable
1198 methods by comparing the actual arguments passed by the caller with
1199 the argument specializers of each method. A method is applicable if
1200 the actual arguments of the call are compatible with the method's
1201 specializers. If more than one method is applicable, they are
1202 combined using certain rules, described below, and the combination
1203 then handles the call.
1205 @defmac cl-defgeneric name arguments [documentation] [options-and-methods@dots{}] &rest body
1206 This macro defines a generic function with the specified @var{name}
1207 and @var{arguments}. If @var{body} is present, it provides the
1208 default implementation. If @var{documentation} is present (it should
1209 always be), it specifies the documentation string for the generic
1210 function, in the form @code{(:documentation @var{docstring})}. The
1211 optional @var{options-and-methods} can be one of the following forms:
1214 @item (declare @var{declarations})
1215 A declare form, as described in @ref{Declare Form}.
1216 @item (:argument-precedence-order &rest @var{args})
1217 This form affects the sorting order for combining applicable methods.
1218 Normally, when two methods are compared during combination, method
1219 arguments are examined left to right, and the first method whose
1220 argument specializer is more specific will come before the other one.
1221 The order defined by this form overrides that, and the arguments are
1222 examined according to their order in this form, and not left to right.
1223 @item (:method [@var{qualifiers}@dots{}] args &rest body)
1224 This form defines a method like @code{cl-defmethod} does.
1228 @defmac cl-defmethod name [qualifier] arguments [&context (expr spec)@dots{}] &rest [docstring] body
1229 This macro defines a particular implementation for the generic
1230 function called @var{name}. The implementation code is given by
1231 @var{body}. If present, @var{docstring} is the documentation string
1232 for the method. The @var{arguments} list, which must be identical in
1233 all the methods that implement a generic function, and must match the
1234 argument list of that function, provides argument specializers of the
1235 form @code{(@var{arg} @var{spec})}, where @var{arg} is the argument
1236 name as specified in the @code{cl-defgeneric} call, and @var{spec} is
1237 one of the following specializer forms:
1241 This specializer requires the argument to be of the given @var{type},
1242 one of the types from the type hierarchy described below.
1243 @item (eql @var{object})
1244 This specializer requires the argument be @code{eql} to the given
1246 @item (head @var{object})
1247 The argument must be a cons cell whose @code{car} is @code{eql} to
1249 @item @var{struct-tag}
1250 The argument must be an instance of a class named @var{struct-tag}
1251 defined with @code{cl-defstruct} (@pxref{Structures,,, cl, Common Lisp
1252 Extensions for GNU Emacs Lisp}), or of one of its parent classes.
1255 Method definitions can make use of a new argument-list keyword,
1256 @code{&context}, which introduces extra specializers that test the
1257 environment at the time the method is run. This keyword should appear
1258 after the list of required arguments, but before any @code{&rest} or
1259 @code{&optional} keywords. The @code{&context} specializers look much
1260 like regular argument specializers---(@var{expr} @var{spec})---except
1261 that @var{expr} is an expression to be evaluated in the current
1262 context, and the @var{spec} is a value to compare against. For
1263 example, @code{&context (overwrite-mode (eql t))} will make the method
1264 applicable only when @code{overwrite-mode} is turned on. The
1265 @code{&context} keyword can be followed by any number of context
1266 specializers. Because the context specializers are not part of the
1267 generic function's argument signature, they may be omitted in methods
1268 that don't require them.
1270 The type specializer, @code{(@var{arg} @var{type})}, can specify one
1271 of the @dfn{system types} in the following list. When a parent type
1272 is specified, an argument whose type is any of its more specific child
1273 types, as well as grand-children, grand-grand-children, etc. will also
1278 Parent type: @code{number}.
1281 Parent type: @code{symbol}
1284 Parent type: @code{array}.
1286 Parent type: @code{sequence}.
1288 Parent type: @code{list}.
1290 Parent type: @code{sequence}.
1294 Parent type: @code{number}.
1295 @item window-configuration
1299 @item compiled-function
1302 Parent type: @code{array}.
1304 Parent type: @code{array}.
1306 Parent type: @code{array}.
1314 The optional @var{qualifier} allows combining several applicable
1315 methods. If it is not present, the defined method is a @dfn{primary}
1316 method, responsible for providing the primary implementation of the
1317 generic function for the specialized arguments. You can also define
1318 @dfn{auxiliary methods}, by using one of the following values as
1323 This auxiliary method will run before the primary method. More
1324 accurately, all the @code{:before} methods will run before the
1325 primary, in the most-specific-first order.
1327 This auxiliary method will run after the primary method. More
1328 accurately, all such methods will run after the primary, in the
1329 most-specific-last order.
1331 This auxiliary method will run @emph{instead} of the primary method.
1332 The most specific of such methods will be run before any other method.
1333 Such methods normally use @code{cl-call-next-method}, described below,
1334 to invoke the other auxiliary or primary methods.
1335 @item :extra @var{string}
1336 This allows you to add more methods, distinguished by @var{string},
1337 for the same specializers and qualifiers.
1341 @cindex dispatch of methods for generic function
1342 @cindex multiple-dispatch methods
1343 Each time a generic function is called, it builds the @dfn{effective
1344 method} which will handle this invocation by combining the applicable
1345 methods defined for the function. The process of finding the
1346 applicable methods and producing the effective method is called
1347 @dfn{dispatch}. The applicable methods are those all of whose
1348 specializers are compatible with the actual arguments of the call.
1349 Since all of the arguments must be compatible with the specializers,
1350 they all determine whether a method is applicable. Methods that
1351 explicitly specialize more than one argument are called
1352 @dfn{multiple-dispatch methods}.
1354 The applicable methods are sorted into the order in which they will be
1355 combined. The method whose left-most argument specializer is the most
1356 specific one will come first in the order. (Specifying
1357 @code{:argument-precedence-order} as part of @code{cl-defmethod}
1358 overrides that, as described above.) If the method body calls
1359 @code{cl-call-next-method}, the next most-specific method will run.
1360 If there are applicable @code{:around} methods, the most-specific of
1361 them will run first; it should call @code{cl-call-next-method} to run
1362 any of the less specific @code{:around} methods. Next, the
1363 @code{:before} methods run in the order of their specificity, followed
1364 by the primary method, and lastly the @code{:after} methods in the
1365 reverse order of their specificity.
1367 @defun cl-call-next-method &rest args
1368 When invoked from within the lexical body of a primary or an
1369 @code{:around} auxiliary method, call the next applicable method for
1370 the same generic function. Normally, it is called with no arguments,
1371 which means to call the next applicable method with the same arguments
1372 that the calling method was invoked. Otherwise, the specified
1373 arguments are used instead.
1376 @defun cl-next-method-p
1377 This function, when called from within the lexical body of a primary
1378 or an @code{:around} auxiliary method, returns non-@code{nil} if there
1379 is a next method to call.
1383 @node Function Cells
1384 @section Accessing Function Cell Contents
1386 The @dfn{function definition} of a symbol is the object stored in the
1387 function cell of the symbol. The functions described here access, test,
1388 and set the function cell of symbols.
1390 See also the function @code{indirect-function}. @xref{Definition of
1393 @defun symbol-function symbol
1394 @kindex void-function
1395 This returns the object in the function cell of @var{symbol}. It does
1396 not check that the returned object is a legitimate function.
1398 If the function cell is void, the return value is @code{nil}. To
1399 distinguish between a function cell that is void and one set to
1400 @code{nil}, use @code{fboundp} (see below).
1404 (defun bar (n) (+ n 2))
1405 (symbol-function 'bar)
1406 @result{} (lambda (n) (+ n 2))
1413 (symbol-function 'baz)
1419 @cindex void function cell
1420 If you have never given a symbol any function definition, we say
1421 that that symbol's function cell is @dfn{void}. In other words, the
1422 function cell does not have any Lisp object in it. If you try to call
1423 the symbol as a function, Emacs signals a @code{void-function} error.
1425 Note that void is not the same as @code{nil} or the symbol
1426 @code{void}. The symbols @code{nil} and @code{void} are Lisp objects,
1427 and can be stored into a function cell just as any other object can be
1428 (and they can be valid functions if you define them in turn with
1429 @code{defun}). A void function cell contains no object whatsoever.
1431 You can test the voidness of a symbol's function definition with
1432 @code{fboundp}. After you have given a symbol a function definition, you
1433 can make it void once more using @code{fmakunbound}.
1435 @defun fboundp symbol
1436 This function returns @code{t} if the symbol has an object in its
1437 function cell, @code{nil} otherwise. It does not check that the object
1438 is a legitimate function.
1441 @defun fmakunbound symbol
1442 This function makes @var{symbol}'s function cell void, so that a
1443 subsequent attempt to access this cell will cause a
1444 @code{void-function} error. It returns @var{symbol}. (See also
1445 @code{makunbound}, in @ref{Void Variables}.)
1459 @error{} Symbol's function definition is void: foo
1464 @defun fset symbol definition
1465 This function stores @var{definition} in the function cell of
1466 @var{symbol}. The result is @var{definition}. Normally
1467 @var{definition} should be a function or the name of a function, but
1468 this is not checked. The argument @var{symbol} is an ordinary evaluated
1471 The primary use of this function is as a subroutine by constructs that define
1472 or alter functions, like @code{defun} or @code{advice-add} (@pxref{Advising
1473 Functions}). You can also use it to give a symbol a function definition that
1474 is not a function, e.g., a keyboard macro (@pxref{Keyboard Macros}):
1477 ;; @r{Define a named keyboard macro.}
1478 (fset 'kill-two-lines "\^u2\^k")
1482 It you wish to use @code{fset} to make an alternate name for a
1483 function, consider using @code{defalias} instead. @xref{Definition of
1490 As explained in @ref{Variable Scoping}, Emacs can optionally enable
1491 lexical binding of variables. When lexical binding is enabled, any
1492 named function that you create (e.g., with @code{defun}), as well as
1493 any anonymous function that you create using the @code{lambda} macro
1494 or the @code{function} special form or the @code{#'} syntax
1495 (@pxref{Anonymous Functions}), is automatically converted into a
1499 A closure is a function that also carries a record of the lexical
1500 environment that existed when the function was defined. When it is
1501 invoked, any lexical variable references within its definition use the
1502 retained lexical environment. In all other respects, closures behave
1503 much like ordinary functions; in particular, they can be called in the
1504 same way as ordinary functions.
1506 @xref{Lexical Binding}, for an example of using a closure.
1508 Currently, an Emacs Lisp closure object is represented by a list
1509 with the symbol @code{closure} as the first element, a list
1510 representing the lexical environment as the second element, and the
1511 argument list and body forms as the remaining elements:
1514 ;; @r{lexical binding is enabled.}
1515 (lambda (x) (* x x))
1516 @result{} (closure (t) (x) (* x x))
1520 However, the fact that the internal structure of a closure is
1521 exposed to the rest of the Lisp world is considered an internal
1522 implementation detail. For this reason, we recommend against directly
1523 examining or altering the structure of closure objects.
1525 @node Advising Functions
1526 @section Advising Emacs Lisp Functions
1527 @cindex advising functions
1528 @cindex piece of advice
1530 When you need to modify a function defined in another library, or when you need
1531 to modify a hook like @code{@var{foo}-function}, a process filter, or basically
1532 any variable or object field which holds a function value, you can use the
1533 appropriate setter function, such as @code{fset} or @code{defun} for named
1534 functions, @code{setq} for hook variables, or @code{set-process-filter} for
1535 process filters, but those are often too blunt, completely throwing away the
1538 The @dfn{advice} feature lets you add to the existing definition of
1539 a function, by @dfn{advising the function}. This is a cleaner method
1540 than redefining the whole function.
1542 Emacs's advice system provides two sets of primitives for that: the core set,
1543 for function values held in variables and object fields (with the corresponding
1544 primitives being @code{add-function} and @code{remove-function}) and another
1545 set layered on top of it for named functions (with the main primitives being
1546 @code{advice-add} and @code{advice-remove}).
1548 For example, in order to trace the calls to the process filter of a process
1549 @var{proc}, you could use:
1552 (defun my-tracing-function (proc string)
1553 (message "Proc %S received %S" proc string))
1555 (add-function :before (process-filter @var{proc}) #'my-tracing-function)
1558 This will cause the process's output to be passed to @code{my-tracing-function}
1559 before being passed to the original process filter. @code{my-tracing-function}
1560 receives the same arguments as the original function. When you're done with
1561 it, you can revert to the untraced behavior with:
1564 (remove-function (process-filter @var{proc}) #'my-tracing-function)
1567 Similarly, if you want to trace the execution of the function named
1568 @code{display-buffer}, you could use:
1571 (defun his-tracing-function (orig-fun &rest args)
1572 (message "display-buffer called with args %S" args)
1573 (let ((res (apply orig-fun args)))
1574 (message "display-buffer returned %S" res)
1577 (advice-add 'display-buffer :around #'his-tracing-function)
1580 Here, @code{his-tracing-function} is called instead of the original function
1581 and receives the original function (additionally to that function's arguments)
1582 as argument, so it can call it if and when it needs to.
1583 When you're tired of seeing this output, you can revert to the untraced
1587 (advice-remove 'display-buffer #'his-tracing-function)
1590 The arguments @code{:before} and @code{:around} used in the above examples
1591 specify how the two functions are composed, since there are many different
1592 ways to do it. The added function is also called a piece of @emph{advice}.
1595 * Core Advising Primitives:: Primitives to manipulate advice.
1596 * Advising Named Functions:: Advising named functions.
1597 * Advice combinators:: Ways to compose advice.
1598 * Porting old advice:: Adapting code using the old defadvice.
1601 @node Core Advising Primitives
1602 @subsection Primitives to manipulate advices
1603 @cindex advice, add and remove
1605 @defmac add-function where place function &optional props
1606 This macro is the handy way to add the advice @var{function} to the function
1607 stored in @var{place} (@pxref{Generalized Variables}).
1609 @var{where} determines how @var{function} is composed with the
1610 existing function, e.g., whether @var{function} should be called before, or
1611 after the original function. @xref{Advice combinators}, for the list of
1612 available ways to compose the two functions.
1614 When modifying a variable (whose name will usually end with @code{-function}),
1615 you can choose whether @var{function} is used globally or only in the current
1616 buffer: if @var{place} is just a symbol, then @var{function} is added to the
1617 global value of @var{place}. Whereas if @var{place} is of the form
1618 @code{(local @var{symbol})}, where @var{symbol} is an expression which returns
1619 the variable name, then @var{function} will only be added in the
1620 current buffer. Finally, if you want to modify a lexical variable, you will
1621 have to use @code{(var @var{variable})}.
1623 Every function added with @code{add-function} can be accompanied by an
1624 association list of properties @var{props}. Currently only two of those
1625 properties have a special meaning:
1629 This gives a name to the advice, which @code{remove-function} can use to
1630 identify which function to remove. Typically used when @var{function} is an
1634 This specifies how to order the advice, should several pieces of
1635 advice be present. By default, the depth is 0. A depth of 100
1636 indicates that this piece of advice should be kept as deep as
1637 possible, whereas a depth of -100 indicates that it should stay as the
1638 outermost piece. When two pieces of advice specify the same depth,
1639 the most recently added one will be outermost.
1641 For @code{:before} advice, being outermost means that this advice will
1642 be run first, before any other advice, whereas being innermost means
1643 that it will run right before the original function, with no other
1644 advice run between itself and the original function. Similarly, for
1645 @code{:after} advice innermost means that it will run right after the
1646 original function, with no other advice run in between, whereas
1647 outermost means that it will be run right at the end after all other
1648 advice. An innermost @code{:override} piece of advice will only
1649 override the original function and other pieces of advice will apply
1650 to it, whereas an outermost @code{:override} piece of advice will
1651 override not only the original function but all other advice applied
1655 If @var{function} is not interactive, then the combined function will inherit
1656 the interactive spec, if any, of the original function. Else, the combined
1657 function will be interactive and will use the interactive spec of
1658 @var{function}. One exception: if the interactive spec of @var{function}
1659 is a function (rather than an expression or a string), then the interactive
1660 spec of the combined function will be a call to that function with as sole
1661 argument the interactive spec of the original function. To interpret the spec
1662 received as argument, use @code{advice-eval-interactive-spec}.
1664 Note: The interactive spec of @var{function} will apply to the combined
1665 function and should hence obey the calling convention of the combined function
1666 rather than that of @var{function}. In many cases, it makes no difference
1667 since they are identical, but it does matter for @code{:around},
1668 @code{:filter-args}, and @code{filter-return}, where @var{function}.
1671 @defmac remove-function place function
1672 This macro removes @var{function} from the function stored in
1673 @var{place}. This only works if @var{function} was added to @var{place}
1674 using @code{add-function}.
1676 @var{function} is compared with functions added to @var{place} using
1677 @code{equal}, to try and make it work also with lambda expressions. It is
1678 additionally compared also with the @code{name} property of the functions added
1679 to @var{place}, which can be more reliable than comparing lambda expressions
1683 @defun advice-function-member-p advice function-def
1684 Return non-@code{nil} if @var{advice} is already in @var{function-def}.
1685 Like for @code{remove-function} above, instead of @var{advice} being the actual
1686 function, it can also be the @code{name} of the piece of advice.
1689 @defun advice-function-mapc f function-def
1690 Call the function @var{f} for every piece of advice that was added to
1691 @var{function-def}. @var{f} is called with two arguments: the advice function
1695 @defun advice-eval-interactive-spec spec
1696 Evaluate the interactive @var{spec} just like an interactive call to a function
1697 with such a spec would, and then return the corresponding list of arguments
1698 that was built. E.g., @code{(advice-eval-interactive-spec "r\nP")} will
1699 return a list of three elements, containing the boundaries of the region and
1700 the current prefix argument.
1703 @node Advising Named Functions
1704 @subsection Advising Named Functions
1705 @cindex advising named functions
1707 A common use of advice is for named functions and macros.
1708 You could just use @code{add-function} as in:
1711 (add-function :around (symbol-function '@var{fun}) #'his-tracing-function)
1714 But you should use @code{advice-add} and @code{advice-remove} for that
1715 instead. This separate set of functions to manipulate pieces of advice applied
1716 to named functions, offers the following extra features compared to
1717 @code{add-function}: they know how to deal with macros and autoloaded
1718 functions, they let @code{describe-function} preserve the original docstring as
1719 well as document the added advice, and they let you add and remove advice
1720 before a function is even defined.
1722 @code{advice-add} can be useful for altering the behavior of existing calls
1723 to an existing function without having to redefine the whole function.
1724 However, it can be a source of bugs, since existing callers to the function may
1725 assume the old behavior, and work incorrectly when the behavior is changed by
1726 advice. Advice can also cause confusion in debugging, if the person doing the
1727 debugging does not notice or remember that the function has been modified
1730 For these reasons, advice should be reserved for the cases where you
1731 cannot modify a function's behavior in any other way. If it is
1732 possible to do the same thing via a hook, that is preferable
1733 (@pxref{Hooks}). If you simply want to change what a particular key
1734 does, it may be better to write a new command, and remap the old
1735 command's key bindings to the new one (@pxref{Remapping Commands}).
1736 In particular, Emacs's own source files should not put advice on
1737 functions in Emacs. (There are currently a few exceptions to this
1738 convention, but we aim to correct them.)
1740 Special forms (@pxref{Special Forms}) cannot be advised, however macros can
1741 be advised, in much the same way as functions. Of course, this will not affect
1742 code that has already been macro-expanded, so you need to make sure the advice
1743 is installed before the macro is expanded.
1745 It is possible to advise a primitive (@pxref{What Is a Function}),
1746 but one should typically @emph{not} do so, for two reasons. Firstly,
1747 some primitives are used by the advice mechanism, and advising them
1748 could cause an infinite recursion. Secondly, many primitives are
1749 called directly from C, and such calls ignore advice; hence, one ends
1750 up in a confusing situation where some calls (occurring from Lisp
1751 code) obey the advice and other calls (from C code) do not.
1753 @defmac define-advice symbol (where lambda-list &optional name depth) &rest body
1754 This macro defines a piece of advice and adds it to the function named
1755 @var{symbol}. The advice is an anonymous function if @var{name} is
1756 nil or a function named @code{symbol@@name}. See @code{advice-add}
1757 for explanation of other arguments.
1760 @defun advice-add symbol where function &optional props
1761 Add the advice @var{function} to the named function @var{symbol}.
1762 @var{where} and @var{props} have the same meaning as for @code{add-function}
1763 (@pxref{Core Advising Primitives}).
1766 @defun advice-remove symbol function
1767 Remove the advice @var{function} from the named function @var{symbol}.
1768 @var{function} can also be the @code{name} of a piece of advice.
1771 @defun advice-member-p function symbol
1772 Return non-@code{nil} if the advice @var{function} is already in the named
1773 function @var{symbol}. @var{function} can also be the @code{name} of
1777 @defun advice-mapc function symbol
1778 Call @var{function} for every piece of advice that was added to the
1779 named function @var{symbol}. @var{function} is called with two
1780 arguments: the advice function and its properties.
1783 @node Advice combinators
1784 @subsection Ways to compose advice
1786 Here are the different possible values for the @var{where} argument of
1787 @code{add-function} and @code{advice-add}, specifying how the advice
1788 @var{function} and the original function should be composed.
1792 Call @var{function} before the old function. Both functions receive the
1793 same arguments, and the return value of the composition is the return value of
1794 the old function. More specifically, the composition of the two functions
1797 (lambda (&rest r) (apply @var{function} r) (apply @var{oldfun} r))
1799 @code{(add-function :before @var{funvar} @var{function})} is comparable for
1800 single-function hooks to @code{(add-hook '@var{hookvar} @var{function})} for
1804 Call @var{function} after the old function. Both functions receive the
1805 same arguments, and the return value of the composition is the return value of
1806 the old function. More specifically, the composition of the two functions
1809 (lambda (&rest r) (prog1 (apply @var{oldfun} r) (apply @var{function} r)))
1811 @code{(add-function :after @var{funvar} @var{function})} is comparable for
1812 single-function hooks to @code{(add-hook '@var{hookvar} @var{function}
1813 'append)} for normal hooks.
1816 This completely replaces the old function with the new one. The old function
1817 can of course be recovered if you later call @code{remove-function}.
1820 Call @var{function} instead of the old function, but provide the old function
1821 as an extra argument to @var{function}. This is the most flexible composition.
1822 For example, it lets you call the old function with different arguments, or
1823 many times, or within a let-binding, or you can sometimes delegate the work to
1824 the old function and sometimes override it completely. More specifically, the
1825 composition of the two functions behaves like:
1827 (lambda (&rest r) (apply @var{function} @var{oldfun} r))
1831 Call @var{function} before the old function and don't call the old
1832 function if @var{function} returns @code{nil}. Both functions receive the
1833 same arguments, and the return value of the composition is the return value of
1834 the old function. More specifically, the composition of the two functions
1837 (lambda (&rest r) (and (apply @var{function} r) (apply @var{oldfun} r)))
1839 @code{(add-function :before-while @var{funvar} @var{function})} is comparable
1840 for single-function hooks to @code{(add-hook '@var{hookvar} @var{function})}
1841 when @var{hookvar} is run via @code{run-hook-with-args-until-failure}.
1844 Call @var{function} before the old function and only call the old function if
1845 @var{function} returns @code{nil}. More specifically, the composition of the
1846 two functions behaves like:
1848 (lambda (&rest r) (or (apply @var{function} r) (apply @var{oldfun} r)))
1850 @code{(add-function :before-until @var{funvar} @var{function})} is comparable
1851 for single-function hooks to @code{(add-hook '@var{hookvar} @var{function})}
1852 when @var{hookvar} is run via @code{run-hook-with-args-until-success}.
1855 Call @var{function} after the old function and only if the old function
1856 returned non-@code{nil}. Both functions receive the same arguments, and the
1857 return value of the composition is the return value of @var{function}.
1858 More specifically, the composition of the two functions behaves like:
1860 (lambda (&rest r) (and (apply @var{oldfun} r) (apply @var{function} r)))
1862 @code{(add-function :after-while @var{funvar} @var{function})} is comparable
1863 for single-function hooks to @code{(add-hook '@var{hookvar} @var{function}
1864 'append)} when @var{hookvar} is run via
1865 @code{run-hook-with-args-until-failure}.
1868 Call @var{function} after the old function and only if the old function
1869 returned @code{nil}. More specifically, the composition of the two functions
1872 (lambda (&rest r) (or (apply @var{oldfun} r) (apply @var{function} r)))
1874 @code{(add-function :after-until @var{funvar} @var{function})} is comparable
1875 for single-function hooks to @code{(add-hook '@var{hookvar} @var{function}
1876 'append)} when @var{hookvar} is run via
1877 @code{run-hook-with-args-until-success}.
1880 Call @var{function} first and use the result (which should be a list) as the
1881 new arguments to pass to the old function. More specifically, the composition
1882 of the two functions behaves like:
1884 (lambda (&rest r) (apply @var{oldfun} (funcall @var{function} r)))
1887 @item :filter-return
1888 Call the old function first and pass the result to @var{function}.
1889 More specifically, the composition of the two functions behaves like:
1891 (lambda (&rest r) (funcall @var{function} (apply @var{oldfun} r)))
1896 @node Porting old advice
1897 @subsection Adapting code using the old defadvice
1898 @cindex old advices, porting
1899 @c NB: The following index entries deliberately avoid ``old'',
1900 @c an adjective that does not come to mind for those who grew up
1901 @c on ‘defadvice’ et al. For those folks, that way is ``current''.
1902 @c They discover its oldness reading this node.
1903 @cindex advices, porting from @code{defadvice}
1907 A lot of code uses the old @code{defadvice} mechanism, which is largely made
1908 obsolete by the new @code{advice-add}, whose implementation and semantics is
1909 significantly simpler.
1911 An old piece of advice such as:
1914 (defadvice previous-line (before next-line-at-end
1915 (&optional arg try-vscroll))
1916 "Insert an empty line when moving up from the top line."
1917 (if (and next-line-add-newlines (= arg 1)
1918 (save-excursion (beginning-of-line) (bobp)))
1924 could be translated in the new advice mechanism into a plain function:
1927 (defun previous-line--next-line-at-end (&optional arg try-vscroll)
1928 "Insert an empty line when moving up from the top line."
1929 (if (and next-line-add-newlines (= arg 1)
1930 (save-excursion (beginning-of-line) (bobp)))
1936 Obviously, this does not actually modify @code{previous-line}. For that the
1939 (ad-activate 'previous-line)
1941 whereas the new advice mechanism needs:
1943 (advice-add 'previous-line :before #'previous-line--next-line-at-end)
1946 Note that @code{ad-activate} had a global effect: it activated all pieces of
1947 advice enabled for that specified function. If you wanted to only activate or
1948 deactivate a particular piece, you needed to @emph{enable} or @emph{disable}
1949 it with @code{ad-enable-advice} and @code{ad-disable-advice}.
1950 The new mechanism does away with this distinction.
1952 Around advice such as:
1955 (defadvice foo (around foo-around)
1956 "Ignore case in `foo'."
1957 (let ((case-fold-search t))
1962 could translate into:
1965 (defun foo--foo-around (orig-fun &rest args)
1966 "Ignore case in `foo'."
1967 (let ((case-fold-search t))
1968 (apply orig-fun args)))
1969 (advice-add 'foo :around #'foo--foo-around)
1972 Regarding the advice's @emph{class}, note that the new @code{:before} is not
1973 quite equivalent to the old @code{before}, because in the old advice you could
1974 modify the function's arguments (e.g., with @code{ad-set-arg}), and that would
1975 affect the argument values seen by the original function, whereas in the new
1976 @code{:before}, modifying an argument via @code{setq} in the advice has no
1977 effect on the arguments seen by the original function.
1978 When porting @code{before} advice which relied on this behavior, you'll need
1979 to turn it into new @code{:around} or @code{:filter-args} advice instead.
1981 Similarly old @code{after} advice could modify the returned value by
1982 changing @code{ad-return-value}, whereas new @code{:after} advice cannot, so
1983 when porting such old @code{after} advice, you'll need to turn it into new
1984 @code{:around} or @code{:filter-return} advice instead.
1986 @node Obsolete Functions
1987 @section Declaring Functions Obsolete
1988 @cindex obsolete functions
1990 You can mark a named function as @dfn{obsolete}, meaning that it may
1991 be removed at some point in the future. This causes Emacs to warn
1992 that the function is obsolete whenever it byte-compiles code
1993 containing that function, and whenever it displays the documentation
1994 for that function. In all other respects, an obsolete function
1995 behaves like any other function.
1997 The easiest way to mark a function as obsolete is to put a
1998 @code{(declare (obsolete @dots{}))} form in the function's
1999 @code{defun} definition. @xref{Declare Form}. Alternatively, you can
2000 use the @code{make-obsolete} function, described below.
2002 A macro (@pxref{Macros}) can also be marked obsolete with
2003 @code{make-obsolete}; this has the same effects as for a function. An
2004 alias for a function or macro can also be marked as obsolete; this
2005 makes the alias itself obsolete, not the function or macro which it
2008 @defun make-obsolete obsolete-name current-name &optional when
2009 This function marks @var{obsolete-name} as obsolete.
2010 @var{obsolete-name} should be a symbol naming a function or macro, or
2011 an alias for a function or macro.
2013 If @var{current-name} is a symbol, the warning message says to use
2014 @var{current-name} instead of @var{obsolete-name}. @var{current-name}
2015 does not need to be an alias for @var{obsolete-name}; it can be a
2016 different function with similar functionality. @var{current-name} can
2017 also be a string, which serves as the warning message. The message
2018 should begin in lower case, and end with a period. It can also be
2019 @code{nil}, in which case the warning message provides no additional
2022 If provided, @var{when} should be a string indicating when the function
2023 was first made obsolete---for example, a date or a release number.
2026 @defmac define-obsolete-function-alias obsolete-name current-name &optional when doc
2027 This convenience macro marks the function @var{obsolete-name} obsolete
2028 and also defines it as an alias for the function @var{current-name}.
2029 It is equivalent to the following:
2032 (defalias @var{obsolete-name} @var{current-name} @var{doc})
2033 (make-obsolete @var{obsolete-name} @var{current-name} @var{when})
2037 In addition, you can mark a particular calling convention for a
2038 function as obsolete:
2040 @defun set-advertised-calling-convention function signature when
2041 This function specifies the argument list @var{signature} as the
2042 correct way to call @var{function}. This causes the Emacs byte
2043 compiler to issue a warning whenever it comes across an Emacs Lisp
2044 program that calls @var{function} any other way (however, it will
2045 still allow the code to be byte compiled). @var{when} should be a
2046 string indicating when the variable was first made obsolete (usually a
2047 version number string).
2049 For instance, in old versions of Emacs the @code{sit-for} function
2050 accepted three arguments, like this
2053 (sit-for seconds milliseconds nodisp)
2056 However, calling @code{sit-for} this way is considered obsolete
2057 (@pxref{Waiting}). The old calling convention is deprecated like
2061 (set-advertised-calling-convention
2062 'sit-for '(seconds &optional nodisp) "22.1")
2066 @node Inline Functions
2067 @section Inline Functions
2068 @cindex inline functions
2070 An @dfn{inline function} is a function that works just like an
2071 ordinary function, except for one thing: when you byte-compile a call
2072 to the function (@pxref{Byte Compilation}), the function's definition
2073 is expanded into the caller. To define an inline function, use
2074 @code{defsubst} instead of @code{defun}.
2076 @defmac defsubst name args [doc] [declare] [interactive] body@dots{}
2077 This macro defines an inline function. Its syntax is exactly the same
2078 as @code{defun} (@pxref{Defining Functions}).
2081 Making a function inline often makes its function calls run faster.
2082 But it also has disadvantages. For one thing, it reduces flexibility;
2083 if you change the definition of the function, calls already inlined
2084 still use the old definition until you recompile them.
2086 Another disadvantage is that making a large function inline can
2087 increase the size of compiled code both in files and in memory. Since
2088 the speed advantage of inline functions is greatest for small
2089 functions, you generally should not make large functions inline.
2091 Also, inline functions do not behave well with respect to debugging,
2092 tracing, and advising (@pxref{Advising Functions}). Since ease of
2093 debugging and the flexibility of redefining functions are important
2094 features of Emacs, you should not make a function inline, even if it's
2095 small, unless its speed is really crucial, and you've timed the code
2096 to verify that using @code{defun} actually has performance problems.
2098 After an inline function is defined, its inline expansion can be
2099 performed later on in the same file, just like macros.
2101 It's possible to use @code{defsubst} to define a macro to expand
2102 into the same code that an inline function would execute
2103 (@pxref{Macros}). But the macro would be limited to direct use in
2104 expressions---a macro cannot be called with @code{apply},
2105 @code{mapcar} and so on. Also, it takes some work to convert an
2106 ordinary function into a macro. To convert it into an inline function
2107 is easy; just replace @code{defun} with @code{defsubst}. Since each
2108 argument of an inline function is evaluated exactly once, you needn't
2109 worry about how many times the body uses the arguments, as you do for
2112 As an alternative to @code{defsubst}, you can use
2113 @code{define-inline} to define functions via their exhaustive compiler
2114 macro. @xref{Defining Functions, define-inline}.
2117 @section The @code{declare} Form
2120 @code{declare} is a special macro which can be used to add meta
2121 properties to a function or macro: for example, marking it as
2122 obsolete, or giving its forms a special @key{TAB} indentation
2123 convention in Emacs Lisp mode.
2125 @anchor{Definition of declare}
2126 @defmac declare specs@dots{}
2127 This macro ignores its arguments and evaluates to @code{nil}; it has
2128 no run-time effect. However, when a @code{declare} form occurs in the
2129 @var{declare} argument of a @code{defun} or @code{defsubst} function
2130 definition (@pxref{Defining Functions}) or a @code{defmacro} macro
2131 definition (@pxref{Defining Macros}), it appends the properties
2132 specified by @var{specs} to the function or macro. This work is
2133 specially performed by @code{defun}, @code{defsubst}, and
2136 Each element in @var{specs} should have the form @code{(@var{property}
2137 @var{args}@dots{})}, which should not be quoted. These have the
2141 @item (advertised-calling-convention @var{signature} @var{when})
2142 This acts like a call to @code{set-advertised-calling-convention}
2143 (@pxref{Obsolete Functions}); @var{signature} specifies the correct
2144 argument list for calling the function or macro, and @var{when} should
2145 be a string indicating when the old argument list was first made obsolete.
2147 @item (debug @var{edebug-form-spec})
2148 This is valid for macros only. When stepping through the macro with
2149 Edebug, use @var{edebug-form-spec}. @xref{Instrumenting Macro Calls}.
2151 @item (doc-string @var{n})
2152 This is used when defining a function or macro which itself will be used to
2153 define entities like functions, macros, or variables. It indicates that
2154 the @var{n}th argument, if any, should be considered
2155 as a documentation string.
2157 @item (indent @var{indent-spec})
2158 Indent calls to this function or macro according to @var{indent-spec}.
2159 This is typically used for macros, though it works for functions too.
2160 @xref{Indenting Macros}.
2162 @item (interactive-only @var{value})
2163 Set the function's @code{interactive-only} property to @var{value}.
2164 @xref{The interactive-only property}.
2166 @item (obsolete @var{current-name} @var{when})
2167 Mark the function or macro as obsolete, similar to a call to
2168 @code{make-obsolete} (@pxref{Obsolete Functions}). @var{current-name}
2169 should be a symbol (in which case the warning message says to use that
2170 instead), a string (specifying the warning message), or @code{nil} (in
2171 which case the warning message gives no extra details). @var{when}
2172 should be a string indicating when the function or macro was first
2175 @item (compiler-macro @var{expander})
2176 This can only be used for functions, and tells the compiler to use
2177 @var{expander} as an optimization function. When encountering a call to the
2178 function, of the form @code{(@var{function} @var{args}@dots{})}, the macro
2179 expander will call @var{expander} with that form as well as with
2180 @var{args}@dots{}, and @var{expander} can either return a new expression to use
2181 instead of the function call, or it can return just the form unchanged,
2182 to indicate that the function call should be left alone. @var{expander} can
2183 be a symbol, or it can be a form @code{(lambda (@var{arg}) @var{body})} in
2184 which case @var{arg} will hold the original function call expression, and the
2185 (unevaluated) arguments to the function can be accessed using the function's
2188 @item (gv-expander @var{expander})
2189 Declare @var{expander} to be the function to handle calls to the macro (or
2190 function) as a generalized variable, similarly to @code{gv-define-expander}.
2191 @var{expander} can be a symbol or it can be of the form @code{(lambda
2192 (@var{arg}) @var{body})} in which case that function will additionally have
2193 access to the macro (or function)'s arguments.
2195 @item (gv-setter @var{setter})
2196 Declare @var{setter} to be the function to handle calls to the macro (or
2197 function) as a generalized variable. @var{setter} can be a symbol in which
2198 case it will be passed to @code{gv-define-simple-setter}, or it can be of the
2199 form @code{(lambda (@var{arg}) @var{body})} in which case that function will
2200 additionally have access to the macro (or function)'s arguments and it will
2201 passed to @code{gv-define-setter}.
2207 @node Declaring Functions
2208 @section Telling the Compiler that a Function is Defined
2209 @cindex function declaration
2210 @cindex declaring functions
2211 @findex declare-function
2213 Byte-compiling a file often produces warnings about functions that the
2214 compiler doesn't know about (@pxref{Compiler Errors}). Sometimes this
2215 indicates a real problem, but usually the functions in question are
2216 defined in other files which would be loaded if that code is run. For
2217 example, byte-compiling @file{simple.el} used to warn:
2220 simple.el:8727:1:Warning: the function ‘shell-mode’ is not known to be
2224 In fact, @code{shell-mode} is used only in a function that executes
2225 @code{(require 'shell)} before calling @code{shell-mode}, so
2226 @code{shell-mode} will be defined properly at run-time. When you know
2227 that such a warning does not indicate a real problem, it is good to
2228 suppress the warning. That makes new warnings which might mean real
2229 problems more visible. You do that with @code{declare-function}.
2231 All you need to do is add a @code{declare-function} statement before the
2232 first use of the function in question:
2235 (declare-function shell-mode "shell" ())
2238 This says that @code{shell-mode} is defined in @file{shell.el} (the
2239 @samp{.el} can be omitted). The compiler takes for granted that that file
2240 really defines the function, and does not check.
2242 The optional third argument specifies the argument list of
2243 @code{shell-mode}. In this case, it takes no arguments
2244 (@code{nil} is different from not specifying a value). In other
2245 cases, this might be something like @code{(file &optional overwrite)}.
2246 You don't have to specify the argument list, but if you do the
2247 byte compiler can check that the calls match the declaration.
2249 @defmac declare-function function file &optional arglist fileonly
2250 Tell the byte compiler to assume that @var{function} is defined in the
2251 file @var{file}. The optional third argument @var{arglist} is either
2252 @code{t}, meaning the argument list is unspecified, or a list of
2253 formal parameters in the same style as @code{defun}. An omitted
2254 @var{arglist} defaults to @code{t}, not @code{nil}; this is atypical
2255 behavior for omitted arguments, and it means that to supply a fourth
2256 but not third argument one must specify @code{t} for the third-argument
2257 placeholder instead of the usual @code{nil}. The optional fourth
2258 argument @var{fileonly} non-@code{nil} means check only that
2259 @var{file} exists, not that it actually defines @var{function}.
2262 To verify that these functions really are declared where
2263 @code{declare-function} says they are, use @code{check-declare-file}
2264 to check all @code{declare-function} calls in one source file, or use
2265 @code{check-declare-directory} check all the files in and under a
2268 These commands find the file that ought to contain a function's
2269 definition using @code{locate-library}; if that finds no file, they
2270 expand the definition file name relative to the directory of the file
2271 that contains the @code{declare-function} call.
2273 You can also say that a function is a primitive by specifying a file
2274 name ending in @samp{.c} or @samp{.m}. This is useful only when you
2275 call a primitive that is defined only on certain systems. Most
2276 primitives are always defined, so they will never give you a warning.
2278 Sometimes a file will optionally use functions from an external package.
2279 If you prefix the filename in the @code{declare-function} statement with
2280 @samp{ext:}, then it will be checked if it is found, otherwise skipped
2283 There are some function definitions that @samp{check-declare} does not
2284 understand (e.g., @code{defstruct} and some other macros). In such cases,
2285 you can pass a non-@code{nil} @var{fileonly} argument to
2286 @code{declare-function}, meaning to only check that the file exists, not
2287 that it actually defines the function. Note that to do this without
2288 having to specify an argument list, you should set the @var{arglist}
2289 argument to @code{t} (because @code{nil} means an empty argument list, as
2290 opposed to an unspecified one).
2292 @node Function Safety
2293 @section Determining whether a Function is Safe to Call
2294 @cindex function safety
2295 @cindex safety of functions
2297 Some major modes, such as SES, call functions that are stored in user
2298 files. (@inforef{Top, ,ses}, for more information on SES@.) User
2299 files sometimes have poor pedigrees---you can get a spreadsheet from
2300 someone you've just met, or you can get one through email from someone
2301 you've never met. So it is risky to call a function whose source code
2302 is stored in a user file until you have determined that it is safe.
2304 @defun unsafep form &optional unsafep-vars
2305 Returns @code{nil} if @var{form} is a @dfn{safe} Lisp expression, or
2306 returns a list that describes why it might be unsafe. The argument
2307 @var{unsafep-vars} is a list of symbols known to have temporary
2308 bindings at this point; it is mainly used for internal recursive
2309 calls. The current buffer is an implicit argument, which provides a
2310 list of buffer-local bindings.
2313 Being quick and simple, @code{unsafep} does a very light analysis and
2314 rejects many Lisp expressions that are actually safe. There are no
2315 known cases where @code{unsafep} returns @code{nil} for an unsafe
2316 expression. However, a safe Lisp expression can return a string
2317 with a @code{display} property, containing an associated Lisp
2318 expression to be executed after the string is inserted into a buffer.
2319 This associated expression can be a virus. In order to be safe, you
2320 must delete properties from all strings calculated by user code before
2321 inserting them into buffers.
2324 What is a safe Lisp expression? Basically, it's an expression that
2325 calls only built-in functions with no side effects (or only innocuous
2326 ones). Innocuous side effects include displaying messages and
2327 altering non-risky buffer-local variables (but not global variables).
2330 @item Safe expression
2333 An atom or quoted thing.
2335 A call to a safe function (see below), if all its arguments are
2338 One of the special forms @code{and}, @code{catch}, @code{cond},
2339 @code{if}, @code{or}, @code{prog1}, @code{prog2}, @code{progn},
2340 @code{while}, and @code{unwind-protect}], if all its arguments are
2343 A form that creates temporary bindings (@code{condition-case},
2344 @code{dolist}, @code{dotimes}, @code{lambda}, @code{let}, or
2345 @code{let*}), if all args are safe and the symbols to be bound are not
2346 explicitly risky (see @pxref{File Local Variables}).
2348 An assignment using @code{add-to-list}, @code{setq}, @code{push}, or
2349 @code{pop}, if all args are safe and the symbols to be assigned are
2350 not explicitly risky and they already have temporary or buffer-local
2353 One of [apply, mapc, mapcar, mapconcat] if the first argument is a
2354 safe explicit lambda and the other args are safe expressions.
2360 A lambda containing safe expressions.
2362 A symbol on the list @code{safe-functions}, so the user says it's safe.
2364 A symbol with a non-@code{nil} @code{side-effect-free} property.
2366 A symbol with a non-@code{nil} @code{safe-function} property. The
2367 value @code{t} indicates a function that is safe but has innocuous
2368 side effects. Other values will someday indicate functions with
2369 classes of side effects that are not always safe.
2372 The @code{side-effect-free} and @code{safe-function} properties are
2373 provided for built-in functions and for low-level functions and macros
2374 defined in @file{subr.el}. You can assign these properties for the
2375 functions you write.
2379 @node Related Topics
2380 @section Other Topics Related to Functions
2382 Here is a table of several functions that do things related to
2383 function calling and function definitions. They are documented
2384 elsewhere, but we provide cross references here.
2388 See @ref{Calling Functions}.
2393 @item call-interactively
2394 See @ref{Interactive Call}.
2396 @item called-interactively-p
2397 See @ref{Distinguish Interactive}.
2400 See @ref{Interactive Call}.
2403 See @ref{Accessing Documentation}.
2409 See @ref{Calling Functions}.
2412 See @ref{Anonymous Functions}.
2415 See @ref{Calling Functions}.
2417 @item indirect-function
2418 See @ref{Function Indirection}.
2421 See @ref{Using Interactive}.
2424 See @ref{Distinguish Interactive}.
2427 See @ref{Creating Symbols}.
2430 See @ref{Mapping Functions}.
2432 @item map-char-table
2433 See @ref{Char-Tables}.
2436 See @ref{Mapping Functions}.
2439 See @ref{Functions for Key Lookup}.