| title |
|---|
Functions |
Creating named functions is done with the defun keyword. It follows this model:
(defun function-name (zero or some arguments)
"docstring"
(code of function body))The return value is the value returned by the last expression of the body (see below for more). There is no "return xx" statement.
So, for example:
(defun hello-world ()
;; ^^ no arguments
(print "hello world!"))Call it:
(hello-world)
;; "hello world!" <-- output
;; "hello world!" <-- a string is returned.The print function prints its one argument to standard output and
returns it. "hello world" is thus the returned value of our function.
Add in arguments like this:
(defun hello (name)
"Say hello to `name'."
(format t "hello ~a !~&" name))
;; HELLO(where ~a is the most used format directive to print a variable
aesthetically and ~& prints a newline)
Call the function:
(hello "me")
;; hello me ! <-- this is printed by `format`
;; NIL <-- return value: `format t` prints a string
;; to standard output and returns nil.If you don't specify the right amount of arguments, you'll be trapped into the interactive debugger with an explicit error message:
(hello)
invalid number of arguments: 0
Optional arguments are declared after the &optional keyword in the
lambda list. They are ordered, they must appear one after another.
This function:
(defun hello (name &optional age gender) …)must be called like this:
(hello "me") ;; a value for the required argument,
;; zero optional arguments
(hello "me" "7") ;; a value for age
(hello "me" 7 :h) ;; a value for age and genderIt is not always convenient to remember the order of the arguments. It
is thus possible to supply arguments by name: we declare them using
&key argname, we set them with :argname "value" in the function call,
and we use argname as a regular variable in the function body.
Key arguments are nil by default.
(defun hello (name &key happy)
"If `happy' is `t', print a smiley"
(format t "hello ~a " name)
(when happy
(format t ":)~&")))The following calls are possible:
(hello "me")
(hello "me" :happy t)
(hello "me" :happy nil) ;; useless, equivalent to (hello "me")
and this is not valid: (hello "me" :happy):
odd number of &KEY arguments
A similar example of a function declaration, with several key parameters:
(defun hello (name &key happy lisper cookbook-contributor-p) …)it can be called with zero or more key parameters, in any order:
(hello "me" :lisper t)
(hello "me" :lisper t :happy t)
(hello "me" :cookbook-contributor-p t :happy t)Last but not least, you would quickly realize it, but we can choose the keys programmatically (they can be variables):
(let ((key :happy)
(val t))
(hello "me" key val))
;; hello me :)
;; NILIt is generally a style warning, but it is possible.
(defun hello (&optional name &key happy)
(format t "hello ~a " name)
(when happy
(format t ":)~&")))In SBCL, this yields:
; in: DEFUN HELLO
; (SB-INT:NAMED-LAMBDA HELLO
; (&OPTIONAL NAME &KEY HAPPY)
; (BLOCK HELLO (FORMAT T "hello ~a " NAME) (WHEN HAPPY (FORMAT T ":)~&"))))
;
; caught STYLE-WARNING:
; &OPTIONAL and &KEY found in the same lambda list: (&OPTIONAL (NAME "John") &KEY
; HAPPY)
;
; compilation unit finished
; caught 1 STYLE-WARNING condition
We can call it:
(hello "me" :happy t)
;; hello me :)
;; NILIn the lambda list, use pairs to give a default value to an optional or a key argument, like (happy t) below:
(defun hello (name &key (happy t))Now happy is true by default.
You can skip this tip for now if you want, but come back later to it as it can turn handy.
We saw that a default key parameter is nil by default ((defun hello (name &key happy) …)). But how can be distinguish between "the value
is NIL by default" and "the user wants it to be NIL"?
We saw how to use a tuple to set its default value:
&key (happy t)
To answer our question, use a triple like this:
&key (happy t happy-p)
where happy-p serves as a predicate variable (using -p is only a
convention, give it the name you want) to know if the key was
supplied. If it was, then it will be T.
So now, we will print a sad face if :happy was explicitely set to
NIL. We don't print it by default.
(defun hello (name &key (happy nil happy-p))
(format t "Key supplied? ~a~&" happy-p)
(format t "hello ~a " name)
(when happy-p
(if happy
(format t ":)")
(format t ":("))))Sometimes you want a function to accept a variable number of
arguments. Use &rest <variable>, where <variable> will be a list.
(defun mean (x &rest numbers)
(/ (apply #'+ x numbers)
(1+ (length numbers))))(mean 1)
(mean 1 2) ;; => 3/2 (yes, it is printed as a ratio)
(mean 1 2 3 4 5) ;; => 3Observe:
(defun hello (name &key happy)
(format t "hello ~a~&" name))
(hello "me" :lisper t)
;; => Error: unknown keyword argumentwhereas
(defun hello (name &key happy &allow-other-keys)
(format t "hello ~a~&" name))
(hello "me" :lisper t)
;; hello meWe might need &allow-other-keys when passing around arguments or
with higher level manipulation of functions.
Here's a real example. We define a function to open a file that always
uses :if-exists :supersede, but still passes any other keys to the
open function.
(defun open-supersede (f &rest other-keys &key &allow-other-keys)
(apply #'open f :if-exists :supersede other-keys))In the case of a duplicated :if-exists argument, our first one takes precedence.
The return value of the function is the value returned by the last executed form of the body.
There are ways for non-local exits (return-from <function name> <value>), but they are usually not needed.
Common Lisp has also the concept of multiple return values.
Returning multiple values is not like returning a tuple or a list of results ;) This is a common misconception.
Multiple values are specially useful and powerful because a change in them needs little to no refactoring.
(defun foo (a b c)
a)This function returns a.
(defvar *res* (foo :a :b :c))
;; :AWe use values to return multiple values:
(defun foo (a b c)
(values a b c))(setf *res* (foo :a :b :c))
;; :AObserve here that *res* is still :A.
All functions that use the return value of foo need not to change, they
still work. If we had returned a list or an array, this would be
different.
multiple-value-bind
We destructure multiple values with multiple-value-bind (or
mvb+TAB in Slime for short) and we can get one given its position
with nth-value:
(multiple-value-bind (res1 res2 res3)
(foo :a :b :c)
(format t "res1 is ~a, res2 is ~a, res2 is ~a~&"
res1 res2 res3))
;; res1 is A, res2 is B, res2 is C
;; NILIts general form is
(multiple-value-bind (var-1 .. var-n) expr
body)The variables var-n are not available outside the scope of multiple-value-bind.
With nth-value:
(nth-value 0 (values :a :b :c)) ;; => :A
(nth-value 2 (values :a :b :c)) ;; => :C
(nth-value 9 (values :a :b :c)) ;; => NILLook here too that values is different from a list:
(nth-value 0 '(:a :b :c)) ;; => (:A :B :C)
(nth-value 1 '(:a :b :c)) ;; => NILNote that (values) with no values returns… no values at all.
multiple-value-list
While we are at it: multiple-value-list turns multiple values to a list:
(multiple-value-list (values 1 2 3))
;; (1 2 3)The reverse is values-list, it turns a list to multiple values:
(values-list '(1 2 3))
;; 1
;; 2
;; 3Anonymous functions are created with lambda:
(lambda (x) (print x))We can call a lambda with funcall or apply (see below).
If the first element of an unquoted list is a lambda expression, the lambda is called:
((lambda (x) (print x)) "hello")
;; hellofuncall is to be used with a known number of arguments, when apply
can be used on a list, for example from &rest:
(funcall #'+ 1 2)
(apply #'+ '(1 2))There is one thing to keep in mind with apply, it is that we can't
use it with super-large lists: the argument list of functions
have a length limit.
We can find this limit in the variable call-arguments-limit. It
depends on the implementation. While it is rather large on SBCL
(4611686018427387903), we have another option to apply a function with
arguments of arbitrary length: reduce.
reduce is used to apply functions on lists and vectors of arbitrary
length. It repeateadly calls the function with two arguments and walks
over the argument list.
For example, instead of using apply like above:
(apply #'min '(22 1 2 3)) ;; imagine a super large list
we can use reduce:
(reduce #'min '(22 1 2 3))
If our argument was 1000 elements long, apply would call the min
function with 1000 arguments, while reduce would call min (nearly)
a 1000 times with 2 arguments each time.
reduce walks over the list, which means the following:
minis first called with arguments 22 and 1, and it produces an intermediate result: 1.minis called again with this intermediate result as first argument, and the following argument of the argument list, 2. An intermediate result is produced, 1 again.minis called again with arguments 1 and 3, and returns the final result, 1.
Look, we can trace it:
CL-USER> (trace min)
CL-USER> (reduce #'min '(22 1 2 3))
0: (MIN 22 1)
0: MIN returned 1
0: (MIN 1 2)
0: MIN returned 1
0: (MIN 1 3)
0: MIN returned 1
1Its full signature is the following:
(reduce function sequence &key key from-end start end initial-value)where key, from-end, start and end are key arguments found in
other built-in functions (see our data-structures chapter). If given,
:initial-value is placed before the first subsequence.
Read more about reduce on the Community Spec:
In the example above, we used #', but a single quote also works, and
we can encounter it in the wild. Which one to use?
It is generally safer to use #', because it respects lexical scope. Observe:
(defun foo (x)
(* x 100))
(flet ((foo (x) (1+ x)))
(funcall #'foo 1))
;; => 2, as expected
;; But:
(flet ((foo (x) (1+ x)))
(funcall 'foo 1))
;; => 100#' is actually the shorthand for (function …):
(function +)
;; #<FUNCTION +>
(flet ((foo (x) (1+ x)))
(print (function foo))
(funcall (function foo) 1))
;; #<FUNCTION (FLET FOO) {1001C0ACFB}>
;; 2Using function or the #' shorthand allows us to refer to local
functions. If we pass instead a symbol to funcall, what is
called is always the function named by that symbol in the global environment.
In addition, #' catches the function by value. If the function is redefined, bindings that refered to this function by #' will still run its original behaviour.
Let's assign a function to a parameter:
(defparameter *foo-caller* #'foo)
(funcall *foo-caller* 1)
;; => 100Now, if we redefine foo, the behaviour of *foo-caller* will not change:
(defun foo (x) (1+ x))
;; WARNING: redefining CL-USER::FOO in DEFUN
;; FOO
(funcall *foo-caller* 1)
;; 100 ;; and not 2If we bind the caller with 'foo, a single quote, the function will be resolved at runtime:
(defun foo (x) (* x 100)) ;; back to original behavior.
(defparameter *foo-caller-2* 'foo)
;; *FOO-CALLER-2*
(funcall *foo-caller-2* 1)
;; 100
;; We change the definition:
(defun foo (x) (1+ x))
;; WARNING: redefining CL-USER::FOO in DEFUN
;; FOO
;; We try again:
(funcall *foo-caller-2* 1)
;; 2The behaviour you want depends on your use case. Generally, using sharpsign-quote is less surprising. But if you are running a tight loop and you want live-update mechanisms (think a game or live visualisations), you might want to use a single quote so that your loop picks up the user's new function definition.
Writing functions that return functions is simple enough:
(defun adder (n)
(lambda (x) (+ x n)))
;; ADDERHere we have defined the function adder which returns an object of type function.
To call the resulting function, we must use funcall or apply:
(adder 5)
;; #<CLOSURE (LAMBDA (X) :IN ADDER) {100994ACDB}>
(funcall (adder 5) 3)
;; 8Trying to call it right away leads to an illegal function call:
((adder 3) 5)
In: (ADDER 3) 5
((ADDER 3) 5)
Error: Illegal function call.Indeed, CL has different namespaces for functions and variables, i.e. the same name can refer to different things depending on its position in a form that's evaluated.
;; The symbol foo is bound to nothing:
CL-USER> (boundp 'foo)
NIL
CL-USER> (fboundp 'foo)
NIL
;; We create a variable:
CL-USER> (defparameter foo 42)
FOO
* foo
42
;; Now foo is "bound":
CL-USER> (boundp 'foo)
T
;; but still not as a function:
CL-USER> (fboundp 'foo)
NIL
;; So let's define a function:
CL-USER> (defun foo (x) (* x x))
FOO
;; Now the symbol foo is bound as a function too:
CL-USER> (fboundp 'foo)
T
;; Get the function:
CL-USER> (function foo)
#<FUNCTION FOO>
;; and the shorthand notation:
* #'foo
#<FUNCTION FOO>
;; We call it:
(funcall (function adder) 5)
#<CLOSURE (LAMBDA (X) :IN ADDER) {100991761B}>
;; and call the lambda:
(funcall (funcall (function adder) 5) 3)
8To simplify a bit, you can think of each symbol in CL having (at least) two "cells" in which information is stored. One cell - sometimes referred to as its value cell - can hold a value that is bound to this symbol, and you can use boundp to test whether the symbol is bound to a value (in the global environment). You can access the value cell of a symbol with symbol-value.
The other cell - sometimes referred to as its function cell - can hold the definition of the symbol's (global) function binding. In this case, the symbol is said to be fbound to this definition. You can use fboundp to test whether a symbol is fbound. You can access the function cell of a symbol (in the global environment) with symbol-function.
Now, if a symbol is evaluated, it is treated as a variable in that its value cell is returned (just foo). If a compound form, i.e. a cons, is evaluated and its car is a symbol, then the function cell of this symbol is used (as in (foo 3)).
In Common Lisp, as opposed to Scheme, it is not possible that the car of the compound form to be evaluated is an arbitrary form. If it is not a symbol, it must be a lambda expression, which looks like (lambda lambda-list form*).
This explains the error message we got above - (adder 3) is neither a symbol nor a lambda expression.
If we want to be able to use the symbol *my-fun* in the car of a compound form, we have to explicitly store something in its function cell (which is normally done for us by the macro defun):
;;; continued from above
CL-USER> (fboundp '*my-fun*)
NIL
CL-USER> (setf (symbol-function '*my-fun*) (adder 3))
#<CLOSURE (LAMBDA (X) :IN ADDER) {10099A5EFB}>
CL-USER> (fboundp '*my-fun*)
T
CL-USER> (*my-fun* 5)
8Read the CLHS section about form evaluation for more.
Closures allow to capture lexical bindings:
(let ((limit 3)
(counter -1))
(defun my-counter ()
(if (< counter limit)
(incf counter)
(setf counter 0))))
(my-counter)
0
(my-counter)
1
(my-counter)
2
(my-counter)
3
(my-counter)
0Or similarly:
(defun repeater (n)
(let ((counter -1))
(lambda ()
(if (< counter n)
(incf counter)
(setf counter 0)))))
(defparameter *my-repeater* (repeater 3))
;; *MY-REPEATER*
(funcall *my-repeater*)
0
(funcall *my-repeater*)
1
(funcall *my-repeater*)
2
(funcall *my-repeater*)
3
(funcall *my-repeater*)
0See more on Practical Common Lisp.
A function name can also be a list of two symbols with setf as the
first one, and where the first argument is the new value:
(defun (setf function-name) (new-value other optional arguments)
body)This mechanism is often used for CLOS methods.
Let's work towards an example. Let's say we manipulate a hash-table that represents a square. We store the square width in this hash-table:
(defparameter *square* (make-hash-table))
(setf (gethash :width *square*) 21)During our program life cycle, we can change the square width, with setf as we did above.
We define a function to compute a square area. We don't store it in the hash-table as it is redundant with the dimension.
(defun area (square)
(expt (gethash :width square) 2))Now, our programming needs lead to the situation where it would be very handy to change the area of the square… and have this reflected on the square's dimensions. It can be ergonomic for your program's application interface to define a setf-function, like this:
(defun (setf area) (new-area square)
(let ((width (sqrt new-area)))
(setf (gethash :width square) width)))And now you can do:
(setf (area *SQUARE*) 100)
;; => 10.0and check your square (describe, inspect…), the new width was set.
A related concept is that of currying which you might be familiar with if you're coming from a functional language. After we've read the last section that's rather easy to implement:
CL-USER> (defun curry (function &rest args)
(lambda (&rest more-args)
(apply function (append args more-args))))
CURRY
CL-USER> (funcall (curry #'+ 3) 5)
8
CL-USER> (funcall (curry #'+ 3) 6)
9
CL-USER> (setf (symbol-function 'power-of-ten) (curry #'expt 10))
#<Interpreted Function "LAMBDA (FUNCTION &REST ARGS)" {482DB969}>
CL-USER> (power-of-ten 3)
1000Now that you know how to do it, you may appreciate using the implementation of the Alexandria library (in Quicklisp).
(ql:quickload "alexandria")
(defun adder (foo bar)
"Add the two arguments."
(+ foo bar))
(defvar add-one (alexandria:curry #'adder 1) "Add 1 to the argument.")
(funcall add-one 10) ;; => 11
(setf (symbol-function 'add-one) add-one)
(add-one 10) ;; => 11- functions: http://www.lispworks.com/documentation/HyperSpec/Body/t_fn.htm#function
- ordinary lambda lists: http://www.lispworks.com/documentation/HyperSpec/Body/03_da.htm
- multiple-value-bind: http://clhs.lisp.se/Body/m_multip.htm