In computer programming, a variadic function is a function of indefinite arity, i.e., one which accepts a variable number of arguments. Support for variadic functions differs widely among programming languages.
There are many mathematical and logical operations that come across naturally as variadic functions. For instance, the summing of numbers or the concatenation of strings or other sequences are operations that can logically apply to any number of operands.
Another operation that has been implemented as a variadic function in many languages is output formatting. The C function printf and the Common Lisp function format are two such examples. Both take one argument that specifies the formatting of the output, and any number of arguments that provide the values to be formatted.
Variadic functions can expose type-safety problems in some languages. For instance, C's printf, if used incautiously, can give rise to a class of security holes known as format string attacks. The attack is possible because the language support for variadic functions is not type-safe; it permits the function to attempt to pop more arguments off the stack than were placed there—corrupting the stack and leading to unexpected behavior.
Variadic functionality can be considered complementary to the apply function, which takes a function and a list/sequence/array as arguments and then calls the function once, with the arguments being the elements of the list.
The following provides an overview of specific implementations in different programming languages and environments.
Common Lisp uses the &rest argument specifier to implement variadic functions. At execution time, the &rest variable collects all following arguments into one list. For example:
(defun average (&rest args) (when args (/ (apply #'+ args) (length args))))
Note that this uses the standard lisp add function which itself is a variadic function.
Common Lisp also provides the &body specifier for situations where multiple statements of code are being passed in as parameters (generally when implementing syntactic macros). While such situations can equivalently be handled by &rest, &body is preferred because it conveys the author's intent more specifically. For example, here is a re-implementation of prog1, a macro that returns the result of the first of a block of statements:
(defmacro prog1-2 (&body statements) (let ((varname (gensym))) ,(let ((,varname ,(car statements))) ,@(cdr statements) ,varname)))
Like Python, Common Lisp provides a facility for optional parameters and named parameters (also called dictionary parameters). For instance, consider this function whose second parameter takes a default value if unspecified:
(defun plus (n &optional (number-to-add 1)) (+ n number-to-add))
When invoked with a single parameter N, the result will be N + 1. When invoked with two parameters X and Y, the result will be X + Y.
Parameters may also be given out-of-order. Consider, for instance, a function that constructs a tuple representing a data item. Such a function could be implemented like this:
(defun construct-data-record (&key (name "") (phone "") (fax "") (address "")) (list name phone fax address)) (construct-data-record :name "Name") ; => ("Name" "" "" "") (construct-data-record :address "10 Some Street" :phone "555-555-5555") ; => ("" "555-555-5555" "" "10 Some Street") (construct-data-record) ; => ("" "" "" "")
To portably implement variadic functions in the C programming language, the standard stdarg.h header file should be used. The older varargs.h header has been deprecated in favor of stdarg.h. In C++, the header file cstdarg should be used.[1]
To create a variadic function, an ellipsis (...) must be placed at the end of a parameter list. Inside the body of the function, a variable of type va_list must be defined. Then the macros va_start(va_list, last fixed param), va_arg(va_list, cast type), va_end(va_list) can be used. For example:
#include <stdarg.h> double average(int count, ...) { va_list ap; int j; double tot = 0; va_start(ap, count); //Requires the last fixed parameter (to get the address) for(j=0; j<count; j++) tot+=va_arg(ap, double); //Requires the type to cast to. Increments ap to the next argument. va_end(ap); return tot/count; }
This will compute the average of an arbitrary number of arguments. Note that the function does not know the number of arguments or their types. The above function requires that the types be double, and the number of arguments is passed in the first argument. In some other cases, for example printf, the number and types of arguments are figured out from a format string. In both cases, this depends on the programmer to supply the correct information. If fewer arguments are passed in than the function believes, or the types of arguments are incorrect, this could cause it to read into invalid areas of memory and can lead to vulnerabilities like the format string attack.
Objective-C uses the same varargs functionality as C. Like C, it has no way of knowing the number or types of the arguments. When the arguments are all objects, the convention is that, if the number of arguments is undetermined, then the list must be "terminated" with nil. Functions that follow this convention include the constructors of data structures that take an undetermined number of elements, like [NSArray arrayWithObjects:...].
#include <stdarg.h> void logObjects(id firstObject, ...) // <-- there is always at least one arg, "nil", so this is valid, even for "empty" list { va_list args; va_start(args, firstObject); id obj; for (obj = firstObject; obj != nil; obj = va_arg(args, id)) NSLog(@"%@", obj); va_end(args); } // This function can be called with any number or type of objects, as long as you terminate it with "nil": logObjects(@"foo", [NSNumber numberWithInt:4], @"bar", nil);
In the current standard of C++, C++11, templates may also take variadic argument lists; this feature is called variadic templates. This allows the creation of variadic template classes and variadic template functions. Variadic templates will finally allow the creation of true tuple classes in C++.
template<typename T> void print( const T& t ) { std::cout << t; } template< typename T, typename... Args> void print( const T& t, const Args&... args ) { print( t ); print( args... ); } template<typename... T> class tuple; template<typename T, typename... Ts> class tuple<T,Ts...>; template<> class tuple<>;
Variadic templates similar to this exist in D:
real average(T...)(T args) { real sum = 0; foreach(arg; args) { sum += arg; } return sum / args.length; } class tuple(T...) { T x; } tuple!(int, char[], Object) my_tuple; // my_tuple.x is now a tuple of three variables, // my_tuple.x[0] is of type int, my_tuple.x[1] char[] and my_tuple.x[2] Object.
Other languages, such as C#, VB.net, and Java use a different approach—they just allow a variable number of arguments of the same (super)type to be passed to a variadic function. Inside the method they are simply collected in an array.
This feature, called "varargs", has been introduced to Java in J2SE 5.0. Parameters are passed as an array:
public static void printSpaced(Object... objects) { for (Object o : objects) System.out.print(o + " "); } // Can be used to print: // printSpaced(1, 2, "three");
C# Example:
public static void PrintSpaced(params Object[] objects) { foreach (Object o in objects) Console.Write(o + " "); } // Can be used to print: // PrintSpaced(1, 2, "three");
VB.Net example:
Public Shared Sub PrintSpaced(ParamArray objects As Object()) For Each o As Object In objects Console.Write(o & " ") Next End Sub ' Can be used to print: ' PrintSpaced(1, 2, "three")
C++/CLI Example:
public static void PrintSpaced(...array<Object^> ^objects) { for each (Object^ o in objects) Console::Write(o + " "); } // Can be used to print: // PrintSpaced(1, 2, "three");
Haskell's type system can be used to make functions that seem to be able to take an unlimited number of arguments, of different types.[2]
In JavaScript, the arguments to a function may be accessed individually as local variables within the function. Additionally, the arguments may be accessed together as members of a local object called arguments. The arguments object is not a true array, but rather an array-like object that possesses a length property.[3]
function printSpaced() { for (var i = 0; i < arguments.length; i++) { document.write(arguments[i] + " "); } }
CoffeeScript provides a standard syntax for parsing all or part of JavaScript's arguments object into an array. The splat operator ... may be appended after a variable in the argument list to have that variable "soak up" any remaining arguments. For instance, the above JavaScript could be rewritten as
printSpaced = (all...) -> for x in all document.write "#{x} "
If there are any other variables in the argument list, those take precedence. So
middle = (first, middle..., last) ->
defines a function that, given two or fewer arguments, will return an empty array; but given three or more arguments, will return an array containing all but the first and last.
In Perl 5, functions don't have an explicit list of parameters (except optional prototypes that are used to imitate certain syntax, i.e., when reimplementing some built-in functions). All arguments are stored in an array called @_, which can be accessed in the function. Therefore no special effort is needed to accept variable number of arguments.
sub print_spaced { my @objects = @_; print "@objects\n"; } sub print_names { my ($title, @names) = @_; print "This is the list of $title names:\n" foreach (@names) { print "Name: $_\n"; } } print_spaced(1, 2, "three");
In PHP, variable-length argument lists are natively supported (without security risks) since version 4; dedicated functions (func_num_args, func_get_arg, func_get_args) allow the programmer to determine the number and values of unspecified arguments.
function sum() { $args = func_get_args(); return array_sum($args); // func_get_args() call cannot appear in function call prior to PHP 5.3 }
Python supports very flexible variadic functions. By marking variables with one asterisk (e.g. *var) the given variable is defined to be a tuple of all the extra arguments. By marking variables with two asterisks (e.g. **var) the given variable is a dictionary of all extra keyword arguments; the keys are strings, which are the names that were used to identify the arguments. If they exist, these arguments must be the last one in the list.
def f(*args, **kwargs): print args print kwargs >>> f(1, 2, "cow", "kitty") (1, 2, "cow", "kitty") {} >>> f(arg1=1, sample=2, name="cow", hero="kitty") () {"arg1": 1, "sample": 2, "name": "cow", "hero": "kitty"} >>> f(1, 2, name="cow", hero="kitty") (1, 2) {"name": "cow", "hero": "kitty"} >>> f(arg1=1, sample=2, name="cow", "kitty") SyntaxError "Non-keyword arg after keyword arg"
Conversely you may also pass in a tuple or dictionary using the same asterisk-notation and have it automatically expand to fill.
def g(a, b, c): print a, b, c >>> mytuple = 1,2,3 >>> mydict = {"a": "first", "b": "second", "c": "third"} >>> g(*mytuple) 1 2 3 >>> g(**mydict) first second third >>> g(**{"a": "first"}) TypeError "g() takes exactly 3 non-keyword arguments (got 1)" >>> g(**{"a": "first", "b": "second", "c": "third", "d": "fourth"}) TypeError "g() got an unexpected keyword argument 'd'"
S-Lang supports two mechanisms for passing optional arguments to a function. The two mechanisms are illustrated by the readascii function that is distributed as part of the S-Lang library. The function is used to read one or more data columns from an ascii (as opposed to binary) file:
readascii ("xy.dat", &x, &y); readascii ("xydy.dat", &x, &y, &dy); readascii ("score.dat", &name, &score, &date, &flags; format="%s %lf %s %d", skip=5);
The optional keyword/value pairs that follow the semi-colon in the argument list are called qualifiers in S-Lang parlance.
When a function is called, the "_NARGS" variable contains the number of arguments passed to the function. This value does not include the number of qualifiers. Here is a simple example that performs the sum of the arguments (assumed here to be numeric) passed to a function:
define sum_values () { if (_NARGS == 0) usage ("result = sum_values (val1, val2, ...);"); variable args = __pop_args (_NARGS); variable total = 0, arg; foreach arg (args) total += arg.value; return total; } vmessage ("Sum = %g", sum_values (1,2,3,4,5));
Here is a variant of the above that uses a qualifier to alternate between the sum and difference of the successive arguments. This example makes use of the built-in sum function to add the elements of an array.
define sum_values () { if (_NARGS == 0) usage ("result = sum_values (val1, val2, ... ; alternate);"); variable args = __pop_args (_NARGS); % Create an array from the arguments variable array = [__push_args(args)]; % If the sum is to alternate, multiply the odd elements by -1 if (qualifier_exists ("alternate")) array[[1::2]] *= -1; return sum(array); } vmessage ("Sum = %g", sum_values (1,2,3,4,5 ; alternate));
Ruby supports variadic functions natively. An asterisk before a single parameter in the function declaration (not necessarily the last one) declares that parameter to be an array into which all remaining input arguments are placed.
def printNames(listTitle, *nameList, post_list_message) puts "This is the list of #{listTitle} names:" nameList.each do |name| puts "Name: #{name}" end puts post_list_message end printNames("employee", "John", "Bob", "Chris","And that was the list!") #This code produces the following output: # #This is the list of employee names: #Name: John #Name: Bob #Name: Chris #And that was the list!
In Scheme, there is a dotted notation for representing improper lists (chains of pairs whose last cdr is not null). You can use this notation in argument lists to specify an argument to "dump" the rest of the arguments into.
If a function is defined with the "define" syntax, then you put a dot before the varargs argument
" ""This is the list of "" names:""Name: ""three"
If a function is defined with the "lambda" syntax, then it is the same thing, except that if there are no arguments before the varargs argument (i.e. if you want to put all the arguments into one thing), then you remove the parentheses altogether; as if to say, let this be the entire argument list.
" ""This is the list of "" names:""Name: "
In Tcl a variadic argument can be defined as last argument and must be named args. The content of args is a list.
proc variadic { p1 p2 args } { puts [info level 0] puts $p1 puts $p2 puts "Number of variadic elements: [llength $args]" foreach arg $args { puts $arg } }
Variadic functions are used very often to process key-value lists.
proc process-key-value { args } { puts [info level 0] puts "Number of variadic elements: [llength $args]" foreach {key value} $args { switch $key { "-loglevel" { # Do something useful } default { set options($key) $value } } } }
Lua supports variadic functions, by use of a special named parameter "..."
function foo(bar, ...) -- ... must be passed as the last param print('number of variable params:', select('#', ...)) print('omit first param:', select(2, ...)) local baz = {...} -- pack variable number of params into a table print('all variadic params are:', ...) print('param no 2', baz[2]) end foo('bar', 1, 2, 3)