Function prototype

Some languages provide the facility to declare functions and subroutines through the use of function prototyping.

Task

Demonstrate the methods available for declaring prototypes within the language. The provided solutions should include:

An explanation of any placement restrictions for prototype declarations
A prototype declaration for a function that does not require arguments
A prototype declaration for a function that requires two arguments
A prototype declaration for a function that utilizes varargs
A prototype declaration for a function that utilizes optional arguments
A prototype declaration for a function that utilizes named parameters
Example of prototype declarations for subroutines or procedures (if these differ from functions)
An explanation and example of any special forms of prototyping not covered by the above

Languages that do not provide function prototyping facilities should be omitted from this task.

Ada

In Ada, prototypes are called specifications.

Specifications must be an exact copy of everything prior to the "is" statement for a function or procedure.
All specifications must appear in a declarative section, ie: before a "begin" statement.
For a main program, specifications are only necessary if a function call appears in the source before the function definition.
For a package, specifications must appear as part of the specification(.ads) file, and do not appear in the body file(.adb) (The file extensions apply to Gnat Ada and may not apply to all compilers).

function noargs return Integer; function twoargs (a, b : Integer) return Integer; -- varargs do not exist function optionalargs (a, b : Integer := 0) return Integer; -- all parameters are always named, only calling by name differs procedure dostuff (a : Integer);

Other Prototyping: Since pointers are not generic in Ada, a type must be defined before one can have a pointer to that type, thus for making linked-list type semantics another trivial prototyping exists:

type Box; -- tell Ada a box exists (undefined yet) type accBox is access Box; -- define a pointer to a box type Box is record -- later define what a box is next : accBox; -- including that a box holds access to other boxes end record;

Example of a package specification (i.e. prototype):

package Stack is procedure Push(Object:Integer); function Pull return Integer; end Stack;

Example of a package body:

package body Stack is procedure Push(Object:Integer) is begin -- implementation goes here end; function Pull return Integer; begin -- implementation goes here end; end Stack;

To use the package and function:

with Stack; procedure Main is N:integer:=5; begin Push(N); ... N := Pull; end Main;

Aime

integer f0(void); # No arguments void f1(integer, real); # Two arguments real f2(...); # Varargs void f3(integer, ...); # Varargs void f4(integer &, text &); # Two arguments (integer and string), pass by reference integer f5(integer, integer (*)(integer)); # Two arguments: integer and function returning integer and taking one integer argument integer f6(integer a, real b); # Parameters names are allowed record f7(void); # Function returning an associative array

ALGOL 68

Works with: ALGOL 68 version Revision 1.

Works with: ALGOL 68G version Any - tested with release algol68g-2.8.

Works with: ELLA ALGOL 68 version Any (with appropriate job cards) - tested with release 1.8-8d

File: Function_prototype.a68

#!/usr/bin/a68g --script # # -*- coding: utf-8 -*- # # An explanation of any placement restrictions for prototype declarations # PROC VOID no args; # Declare a function with no argument that returns an INTeger # PROC (INT #a#,INT #b#)VOID two args; # Declare a function with two arguments that returns an INTeger # MODE VARARGS = UNION(INT,REAL,COMPL); PROC ([]VARARGS)VOID var args; # An empty parameter list can be used to declare a function that accepts varargs # PROC (INT, []VARARGS)VOID at least one args; # One mandatory INTeger argument followed by varargs # MODE OPTINT = UNION(VOID,INT), OPTSTRING=UNION(VOID,STRING); # a function that utilizes optional arguments # PROC (OPTINT, OPTSTRING)VOID optional arguments; # A prototype declaration for a function that utilizes named parameters # MODE KWNAME = STRUCT(STRING name), KWSPECIES = STRUCT(STRING species), KWBREED = STRUCT(STRING breed), OWNER=STRUCT(STRING first name, middle name, last name); # due to the "Yoneda ambiguity" simple arguments must have an unique operator defined # OP NAME = (STRING name)KWNAME: (KWNAME opt; name OF opt := name; opt), SPECIES = (STRING species)KWSPECIES: (KWSPECIES opt; species OF opt := species; opt), BREED = (STRING breed)KWBREED: (KWBREED opt; breed OF opt := breed; opt); PROC ([]UNION(KWNAME,KWSPECIES,KWBREED,OWNER) #options#)VOID print pet; # subroutines, and fuctions are procedures, so have the same prototype declarations # # An explanation and example of any special forms of prototyping not covered by the above: # COMMENT If a function has no arguments, eg f, then it is not requied to pass it a "vacuum()" to call it, eg "f()" not correct! Rather is can be called without the () vacuum. eg "f" A GOTO "label" is equivalent to "PROC VOID label". END COMMENT SKIP

Amazing Hopper

Function prototypes are included in a #PROTO declaration in the header, at the beginning of a HOPPER source file, before the functional code (MAIN :). This is enforced by the HOPPER compiler, to declare pseudo functions.

#!/usr/bin/hopper // Archivo Hopper #include <hopper.h> #context-free noargs /* Declare a pseudo-function with no argument */ #synon noargs no arguments #context multiargs /* Declare a pseudo-function with multi arguments */ #proto twoargs(_X_,_Y_) /* Declare a pseudo-function with two arguments. #PROTO need arguments */ main: no arguments _two args(2,2) // pseudo-function #proto need "_" sufix println {1,2,3,"hola mundo!","\n"}, multiargs exit(0) .locals multiargs: _PARAMS_={},pushall(_PARAMS_) [1:3]get(_PARAMS_),stats(SUMMATORY),println {"Mensaje: "}[4:5]get(_PARAMS_),println clear(_PARAMS_) back twoargs(a,b) {a}mulby(b) back // This function is as useful a s an ashtray on a motorcycle: no args: {0}minus(0),kill back {0}return

C

Function prototypes are typically included in a header file at the beginning of a source file prior to functional code. However, this is not enforced by a compiler.

int noargs(void); /* Declare a function with no argument that returns an integer */ int twoargs(int a,int b); /* Declare a function with two arguments that returns an integer */ int twoargs(int ,int); /* Parameter names are optional in a prototype definition */ int anyargs(); /* An empty parameter list can be used to declare a function that accepts varargs */ int atleastoneargs(int, ...); /* One mandatory integer argument followed by varargs */

C#

Abstract methods
Interfaces and abstract classes can define abstract methods that must be implemented by subclasses.

using System; abstract class Printer {  public abstract void Print(); } class PrinterImpl : Printer {  public override void Print() {  Console.WriteLine("Hello world!");  } }

Delegates
A delegate is similar to a function pointer. They are multicast: multiple methods can be attached to them.

using System; public delegate int IntFunction(int a, int b); public class Program {  public static int Add(int x, int y) {  return x + y;  }  public static int Multiply(int x, int y) {  return x * y;  }  public static void Main() {  IntFunction func = Add;  Console.WriteLine(func(2, 3)); //prints 5  func = Multiply;  Console.WriteLine(func(2, 3)); //prints 6  func += Add;  Console.WriteLine(func(2, 3)); //prints 5. Both functions are called, but only the last result is kept.  } }

Partial methods
A partial type is a type that is defined in multiple files.
A partial method has its signature defined in one part of a partial type, and its implementation defined in another part of the type. If it is not implemented, the compiler removes the signature at compile time.
The following conditions apply to partial methods:
- Signatures in both parts of the partial type must match.
- The method must return void.
- No access modifiers are allowed. Partial methods are implicitly private.

//file1.cs public partial class Program {  partial void Print(); } //file2.cs using System; public partial class Program {  partial void Print() {  Console.WriteLine("Hello world!");  }  static void Main() {  Program p = new Program();  p.Print(); //If the implementation above is not written, the compiler will remove this line.  } }

C++

Function declaration in C++ differs from that in C in some aspect.

int noargs(); // Declare a function with no arguments that returns an integer int twoargs(int a,int b); // Declare a function with two arguments that returns an integer int twoargs(int ,int); // Parameter names are optional in a prototype definition int anyargs(...); // An ellipsis is used to declare a function that accepts varargs int atleastoneargs(int, ...); // One mandatory integer argument followed by varargs template<typename T> T declval(T); //A function template template<typename ...T> tuple<T...> make_tuple(T...); //Function template using parameter pack (since c++11)

Clojure

If you want to make forward declarations, you can use declare.

(declare foo)

COBOL

Prototypes were introduced in COBOL 2002. In the following examples, PROGRAM-ID and PROGRAM can be replaced with the equivalents for functions and methods. However, in method prototypes the PROTOTYPE clause is not used.

  *> A subprogram taking no arguments and returning nothing.   PROGRAM-ID. no-args PROTOTYPE.   END PROGRAM no-args.   *> A subprogram taking two 8-digit numbers as arguments, and returning   *> an 8-digit number.   PROGRAM-ID. two-args PROTOTYPE.   DATA DIVISION.   LINKAGE SECTION.   01 arg-1 PIC 9(8).   01 arg-2 PIC 9(8).   01 ret PIC 9(8).   PROCEDURE DIVISION USING arg-1, arg-2 RETURNING ret.   END PROGRAM two-args.   *> A subprogram taking two optional arguments which are 8-digit   *> numbers (passed by reference (the default and compulsory for   *> optional arguments)).   PROGRAM-ID. optional-args PROTOTYPE.   DATA DIVISION.   LINKAGE SECTION.   01 arg-1 PIC 9(8).   01 arg-2 PIC 9(8).   PROCEDURE DIVISION USING OPTIONAL arg-1, OPTIONAL arg-2.   END PROGRAM optional-args.   *> Standard COBOL does not support varargs or named parameters.   *> A function from another language, taking a 32-bit integer by   *> value and returning a 32-bit integer (in Visual COBOL).   PROGRAM-ID. foreign-func PROTOTYPE.   OPTIONS.   ENTRY-CONVENTION some-langauge.   DATA DIVISION.   WORKING-STORAGE SECTION.   01 arg PIC S9(9) USAGE COMP-5.   01 ret PIC S9(9) USAGE COMP-5.   PROCEDURE DIVISION USING arg RETURNING ret.   END PROGRAM foreign-func.

Common Lisp

In Common Lisp, function prototypes can be used with (declaim (inline func-name)) function arguments are taken when the function is defined. In addition, the argument types aren't needed.

Caveat -- This works with specific implementations of CL. This was tested in SBCL.

(declaim (inline no-args)) (declaim (inline one-arg)) (declaim (inline two-args)) (declaim (inline optional-args)) (defun no-args ()  (format nil "no arguments!")) (defun one-arg (x)   ; inserts the value of x into a string  (format nil "~a" x)) (defun two-args (x y)  ; same as function `one-arg', but with two arguments  (format nil "~a ~a" x y)) (defun optional-args (x &optional y) ; optional args are denoted with &optional beforehand  ; same as function `two-args', but if y is not given it just prints NIL  (format nil "~a ~a~%" x y)) (no-args) ;=> "no arguments!"  (one-arg 1) ;=> "1" (two-args 1 "example") ;=> "1 example"  (optional-args 1.0) ;=> "1.0 NIL"  (optional-args 1.0 "example") ;=> "1.0 example"

More about declaim here

D

Beside function prototypes similar to the ones available in C (plus templates, D-style varargs), you can define class method prototypes in abstract classes and interfaces. The exact rules for this are best explained by the documentation

/// Declare a function with no arguments that returns an integer. int noArgs(); /// Declare a function with two arguments that returns an integer. int twoArgs(int a, int b); /// Parameter names are optional in a prototype definition. int twoArgs2(int, int); /// An ellipsis can be used to declare a function that accepts /// C-style varargs. int anyArgs(...); /// One mandatory integer argument followed by C-style varargs. int atLeastOneArg(int, ...); /// Declare a function that accepts any number of integers. void anyInts(int[] a...); /// Declare a function that accepts D-style varargs. void anyArgs2(TArgs...)(TArgs args); /// Declare a function that accepts two or more D-style varargs. void anyArgs3(TArgs...)(TArgs args) if (TArgs.length > 2); /// Currently D doesn't support named arguments. /// One implementation. int twoArgs(int a, int b) {  return a + b; } interface SomeInterface {  void foo();  void foo(int, int);  // Varargs  void foo(...); // C-style.  void foo(int[] a...);  void bar(T...)(T args); // D-style.  // Optional arguments are only supported if a default is provided,  // the default arg(s) has/have to be at the end of the args list.  void foo(int a, int b = 10); } void main() {}

Delphi

Library: System.SysUtils

In Delphi, prototype function is named Class/Record Helper. For now, is not possible has more then one helper active in a same object, if two or more was declareted, just the present in the last Unit declareted will be active, the others will be ignored. In case two or more helpers was declareted in same Unit, just the last helper declareted will be active. Can not inherit record helpers, but class helper can be.
Patten: " identifierName = record helper for TypeIdentifierName"
See Documentation for more details.

program Function_prototype; {$APPTYPE CONSOLE} uses  System.SysUtils; type  TIntArray = TArray<Integer>;  TIntArrayHelper = record helper for TIntArray  const  DEFAULT_VALUE = -1;  // A prototype declaration for a function that does not require arguments  function ToString(): string;  // A prototype declaration for a function that requires two arguments  procedure Insert(Index: Integer; value: Integer);  // A prototype declaration for a function that utilizes varargs  // varargs is not available, but a equivalent is array of const  procedure From(Args: array of const);  //A prototype declaration for a function that utilizes optional arguments  procedure Delete(Index: Integer; Count: Integer = 1);  //A prototype declaration for a function that utilizes named parameters  // Named parameters is not supported in Delphi  //Example of prototype declarations for subroutines or procedures  //(if these differ from functions)  procedure Sqr; //Procedure return nothing  function Averange: double; //Function return a value  end; { TIntHelper } function TIntArrayHelper.Averange: double; begin  Result := 0;  for var e in self do  Result := Result + e;  Result := Result / Length(self); end; procedure TIntArrayHelper.Delete(Index, Count: Integer); begin  System.Delete(self, Index, Count); end; procedure TIntArrayHelper.From(Args: array of const); var  I, Count: Integer; begin  Count := Length(Args);  SetLength(self, Count);  if Count = 0 then  exit;  for I := 0 to High(Args) do  with Args[I] do  case VType of  vtInteger:  self[I] := VInteger;  vtBoolean:  self[I] := ord(VBoolean);  vtChar, vtWideChar:  self[I] := StrToIntDef(string(VChar), DEFAULT_VALUE);  vtExtended:  self[I] := Round(VExtended^);  vtString:  self[I] := StrToIntDef(VString^, DEFAULT_VALUE);  vtPChar:  self[I] := StrToIntDef(VPChar, DEFAULT_VALUE);  vtObject:  self[I] := cardinal(VObject);  vtClass:  self[I] := cardinal(VClass);  vtAnsiString:  self[I] := StrToIntDef(string(VAnsiString), DEFAULT_VALUE);  vtCurrency:  self[I] := Round(VCurrency^);  vtVariant:  self[I] := Integer(VVariant^);  vtInt64:  self[I] := Integer(VInt64^);  vtUnicodeString:  self[I] := StrToIntDef(string(VUnicodeString), DEFAULT_VALUE);  end; end; procedure TIntArrayHelper.Insert(Index, value: Integer); begin  system.Insert([value], self, Index); end; procedure TIntArrayHelper.Sqr; begin  for var I := 0 to High(self) do  Self[I] := Self[I] * Self[I]; end; function TIntArrayHelper.ToString: string; begin  Result := '[';  for var e in self do  Result := Result + e.ToString + ', ';  Result := Result + ']'; end; begin  var val: TArray<Integer>;  val.From([1, '2', PI]);  val.Insert(0, -1); // insert -1 at position 0  writeln(' Array: ', val.ToString, ' ');  writeln(' Averange: ', val.Averange: 3: 2);  val.Sqr;  writeln(' Sqr: ', val.ToString);  Readln; end.

Output:

 Array: [-1, 1, 2, 3, ] Averange: 1.25 Sqr: [1, 1, 4, 9, ]

Class helper example, with inherited helpers:
Patten: " identifierName = class helper (ancestor list*) for TypeIdentifierName"
Ancertor list is optional, and only will be used if this helper will inhret a parent helper.

Library: System.SysUtils

Library: System.Classes

program Function_prototype_class; {$APPTYPE CONSOLE} uses  System.SysUtils,  System.Classes; type  TStringListHelper1 = class helper for TStringList  constructor Create(FileName: TFileName); overload;  end;  TStringListHelper2 = class helper (TStringListHelper1) for TStringList  procedure SaveAndFree(FileName: TFileName);  end;  TStringListHelper3 = class helper (TStringListHelper2) for TStringList  procedure AddDateTime;  end; { TStringListHelper1 } constructor TStringListHelper1.Create(FileName: TFileName); begin  inherited Create;  if FileExists(FileName) then  LoadFromFile(FileName); end; { TStringListHelper2 } procedure TStringListHelper2.SaveAndFree(FileName: TFileName); begin  SaveToFile(FileName);  Free; end; { TStringListHelper3 } procedure TStringListHelper3.AddDateTime; begin  self.Add(DateToStr(now)); end; begin  with TStringList.Create('d:\Text.txt') do  begin  AddDateTime;  SaveAndFree('d:\Text_done.txt');  end;  readln; end.

Diego

// A prototype declaration for a function that does not require arguments begin_funct(foo); // function as a 'method' with no arguments, no return type end_funct(foo); // or with_funct(foo); // function as a 'method' with no arguments, no return type // A prototype declaration for a function that requires two arguments begin_funct(goo)_arg({string}, str1, str2); // two arguments, no return type end_funct[]; // derived name of function using [], like 'this' with_funct(goo)_arg({str}, str1, str2); // two arguments, no return type with_funct(hoo)_param({integer}, i, j); // 'param' posit can be used instead of 'arg' // A prototype declaration for a function that utilizes varargs begin_funct(voo)_arg({int}, [vararg], v); // variable number of arguments, no return type, 'int' can be used instead of 'integer' end_funct[]; begin_funct({int}, voo)_arg({int}, ..., v); // variable number of arguments, with return type add_var({int}, sum)_v(0); forall_var(v)_calc([sum]+=[v]); [voo]_ret([sum]); end_funct[]; // A prototype declaration for a function that utilizes optional arguments begin_funct({int}, ooo)_arg(o)_value(1); // optional argument with default value and return type integer with_funct(ooo)_return([o]); // Can be shortened to [ooo]_ret([0]); end_funct[]; begin_funct({int}, oxo)_arg(o,u,v)_opt(u)_value(1); // optional argument of second argument with default value and return type integer [ooo]_ret(1); // the execution has to name arguments or missing in comma-separated list of arguments end_funct[]; // A prototype declaration for a function that utilizes named parameters begin_funct({int}, poo)_param({int}, a, b, c); // to enforce named parameters '_param' posit can be used. [poo]_ret([a]+[b]+[c]); end_funct[]; exec_funct(poo)_param(a)_value(1)_param(b, c)_value(2, 3) ? me_msg()_funct(poo); ; begin_funct({int}, poo)_arg({int}, a, b, c); // named parameters can still be used with '_arg' posit. [poo]_ret([a]+[b]+[c]); end_funct[]; me_msg()_funct(poo)_arg(a)_value(1)_value(2, 3); // Callee has to figure out unnamed arguments by extraction // 'exec_' verb is implied before '_funct' action // Example of prototype declarations for subroutines or procedures (if these differ from functions) begin_instruct(foo); // instructions are 'methods', no arguments, no return type end_instruct[foo]; // explicit end of itself // or with_instruct(foo); // instructions are 'methods', no arguments, no return type begin_funct(yoo)_arg(robotMoniker)_param(b); // A '_funct' can be used as a subroutine when missing the '{}' return datatype // a mix of '_arg' and '_param' posits can be used with_robot[robotMoniker]_var(sq)_calc([b]^2); // create a variable called 'sq' on robot 'robotMoniker' end_funct(yoo); begin_instruct(woo)_arg(robotType)_param(b); // An '_instuct' is only used for subroutines and return datatypes are not accepted with_robot()_type[robotType]_var(sq)_calc([b]^2); // create a variable called 'sq' on all robots of type 'robotType' end_funct(woo); // An explanation and example of any special forms of prototyping not covered by the above begin_funct({double}, voo)_arg({int}, [numArgs], v); // variable-defined number of arguments, with return type add_var({int}, sum)_v(0); add_var({double}, average)_v(0); for_var(v)_until[numArgs]_calc([sum]+=[v]); // the number of arguments [numArgs] does not have to be number of arguments of v [voo]_ret([sum]/[numArgs]); end_funct[]; begin_funct({int}, [numArgsOut], voo)_arg({int}, [numArgsIn], v); // variable-defined number of arguments, with variable-defined number of return types add_var({int}, sum)_v(0); add_var({double}, average)_v(0); for_var(v)_until[numArgsOut]_calc([sum]+=[v]); [voo]_ret([sum]/[numArgsOut]); end_funct[];

EasyLang

Easylang is a one-pass compiler. Functions must therefore be defined or declared before use. For mutually recursive functions, you therefore need function prototypes.

# No arguments funcdecl f0 . # # Two arguments procdecl f1 a s$ . # # Varargs are not available # # Two arguments (integer and string), pass by reference procdecl f4 &a &s$ . # # Function returning an array funcdecl[] f7 . # print f0 # # Each declared function must also be implemented # func f0 . return 3 . proc f1 a s$ . print s$ & a . proc f4 &a &s$ . a = 7 s$ = "hello" . func[] f7 . return [ 1 2 3 ] .

Output:

F#

In F#, prototypes are called signatures. Signature files are used to bulk annotate the accessibility of the things within them. If something is in an implementation file but not in the signature file, it is assumed to be private to that file. If it is in the signature file without the internal accessibility modifier, then it is assumed to public, otherwise it is internal to the assembly. For more details, see the documentation. Below is an example of the signatures produced for the functions specified in the task (without any accessibility modifiers):

// A function taking and returning nothing (unit). val noArgs : unit -> unit // A function taking two integers, and returning an integer. val twoArgs : int -> int -> int // A function taking a ParamPack array of ints, and returning an int. The ParamPack // attribute is not included in the signature. val varArgs : int [] -> int // A function taking an int and a ParamPack array of ints, and returning an // object of the same type. val atLeastOnArg : int -> int [] -> int // A function taking an int Option, and returning an int. val optionalArg : Option<int> -> int // Named arguments and the other form of optional arguments are only available on // methods. type methodClass =  class  // A method taking an int named x, and returning an int.  member NamedArg : x:int -> int  // A method taking two optional ints in a tuple, and returning an int. The  //optional arguments must be tupled.  member OptionalArgs : ?x:int * ?y:int -> int  end

FreeBASIC

' FB 1.05.0 Win64 ' The position regarding prototypes is broadly similar to that of the C language in that functions,  ' sub-routines or operators (unless they have already been fully defined) must be declared before they can be used.  ' This is usually done near the top of a file or in a separate header file which is then 'included'. ' Parameter names are optional in declarations. When calling functions, using parameter names  ' (as opposed to identifying arguments by position) is not supported. Type MyType ' needed for operator declaration  i As Integer End Type Declare Function noArgs() As Integer ' function with no argument that returns an integer  Declare Function twoArgs(As Integer, As Integer) As Integer ' function with two arguments that returns an integer  Declare Function atLeastOneArg CDecl(As Integer, ...) As Integer ' one mandatory integer argument followed by varargs Declare Function optionalArg(As Integer = 0) As Integer ' function with a (single) optional argument with default value  Declare Sub noArgs2() ' sub-routine with no argument Declare Operator + (As MyType, As MyType) As MyType ' operator declaration (no hidden 'This' parameter for MyType) ' FreeBASIC also supports object-oriented programming and here all constructors, destructors,  ' methods (function or sub), properties and operators (having a hidden 'This' parameter) must be  ' declared within a user defined type and then defined afterwards. Type MyType2  Public:  Declare Constructor(As Integer)  Declare Destructor()  Declare Sub MySub()   Declare Function MyFunction(As Integer) As Integer   Declare Property MyProperty As Integer  Declare Operator Cast() As String  Private:  i As Integer End Type

FutureBasic

A function prototype must appear above any calls to the function.

def fn NoArgs def fn TwoArgs( a as long, b as long ) def fn VarArgs( n as long, ... ) long def fn Multiply( a as long, b as long ) // function returns a value

Go

While the language specification does not use the word prototype it states, "A function declaration may omit the body. Such a declaration provides the signature for a function implemented outside Go, such as an assembly routine." This is the closest analogy to a C (e.g.) prototype.

Function declarations whether with a body or without must be "top level" declarations, that is, after the package clause and outside of any other function. Examples of function delarations without bodies are,

func a() // function with no arguments func b(x, y int) // function with two arguments func c(...int) // varargs are called "variadic parameters" in Go.

Go does not directly support optional or named parameters and does not have any concept of procedures or subroutines as distinct from functions.

Otherwise, Go does have the concept of a function signature which includes parameters and return values. Go is strongly typed and functions are first class objects so function signatures are used in a variety of ways. These might be considered distinct from the concept of function prototype.

Haskell

A function can be declared without giving it's prototype in Haskell. The haskell compiler has got type inference whereby it can infer the return type and type of variable given to function. You can still hardcode the prototype which specifies the datatype of variables and return type. For ex. Consider a function add which takes two integers and returns their sum. It can be prototyped and declared as :

add :: Int -> Int -> Int add x y = x+y

Actually all functions in haskell are functions with just one arguments. Haskell will treat above function as a function which takes an int and returns a function with type (:: (Int->Int)) . Then this function which is returned is such that it takes an int and returns an int. Similarly for any function add which takes 3 integers and adds them the actual prototype will be as follows:

add :: Int->(Int ->(Int->Int))

The one that does not require arguements could just be:

printThis = putStrLn("This is being printed.")

But haskell would rather consider the function to be of return type IO() in this case.

Two arguments:

add :: Int -> Int -> Int add x y = x+y

The same thing can be done using the lambda function as :

add :: Int -> Int -> Int add = \x->\y -> x+y

Two arguments with unnamed parameters:

doThis :: Int-> Int-> String doThis _ _ = "Function with unnamed parameters"

Function with var args requires creation of type class as per the requirement.

J

J assumes an unknown name is a verb of infinite rank. Rank determines the frames on which the verb executes. As the demonstration shows, changing the rank, assigning the rank before the verb is used in other definitions affects the result. We could, of course, play other games by changing the unknown name from verb to another part of speech.

 NB. j assumes an unknown name f is a verb of infinite rank  NB. f has infinite ranks  f b. 0 _ _ _    NB. The verb g makes a table.  g=: f/~  NB. * has rank 0  f=: *  NB. indeed, make a multiplication table  f/~ i.5 0 0 0 0 0 0 1 2 3 4 0 2 4 6 8 0 3 6 9 12 0 4 8 12 16    NB. g was defined as if f had infinite rank.  g i.5 0 1 4 9 16    NB. f is known to have rank 0.  g=: f/~  NB. Now we reproduce the table  g i.5 0 0 0 0 0 0 1 2 3 4 0 2 4 6 8 0 3 6 9 12 0 4 8 12 16      NB. change f to another rank 0 verb  f=: +   NB. and construct an addition table  g i.5 0 1 2 3 4 1 2 3 4 5 2 3 4 5 6 3 4 5 6 7 4 5 6 7 8    NB. f is multiplication at infinite rank  f=: *"_    NB. g, however, has rank 0  g i.5 0 0 0 0 0 0 1 2 3 4 0 2 4 6 8 0 3 6 9 12 0 4 8 12 16

Java

The order of declarations in Java is unimportant, so forward declaration of functions is neither needed nor supported. The only place where function prototypes are needed is in abstract classes or interfaces, where implementation will be provided by the derived or implementing class.

public final class FunctionPrototype {  public static void main(String[] aArgs) { Rectangle rectangle = new Rectangle(10.0, 20.0); System.out.println("Area = " + rectangle.area());  Calculator calculator = new Calculator(); System.out.println("Sum = " + calculator.sum(2, 2)); calculator.version(); } private static class Rectangle implements Shape {  public Rectangle(double aWidth, double aLength) { width = aWidth; length = aLength; }  @Override  public double area() {  return length * width;  }  private final double width, length;   }  private static class Calculator extends Arithmetic {  public Calculator() { // Statements to create the graphical  // representation of the calculator }  @Override public int sum(int aOne, int aTwo) { return aOne + aTwo; }  @Override public void version() { System.out.println("0.0.1"); }  } } interface Shape {  public double area(); }  abstract class Arithmetic { public abstract int sum(int aOne, int aTwo); public abstract void version(); }

Output:

Area = 200.0 Sum = 4 0.0.1

JavaScript

ES5

JavaScript functions may also be used to define prototype objects

// A prototype declaration for a function that does not require arguments function List() {} List.prototype.push = function() {  return [].push.apply(this, arguments); }; List.prototype.pop = function() {  return [].pop.call(this); }; var l = new List(); l.push(5); l.length; // 1 l[0]; 5 l.pop(); // 5 l.length; // 0 // A prototype declaration for a function that utilizes varargs function List() {  this.push.apply(this, arguments); } List.prototype.push = function() {  return [].push.apply(this, arguments); }; List.prototype.pop = function() {  return [].pop.call(this); }; var l = new List(5, 10, 15); l.length; // 3 l[0]; 5 l.pop(); // 15 l.length; // 2

ES6

Class Declarations are used to define prototype objects

// A prototype declaration for a function that does not require arguments class List {  push() {  return [].push.apply(this, arguments);  }  pop() {  return [].pop.call(this);   } } var l = new List(); l.push(5); l.length; // 1 l[0]; 5 l.pop(); // 5 l.length; // 0 // A prototype declaration for a function that utilizes varargs class List {  constructor(...args) {  this.push(...args);  }  push() {  return [].push.apply(this, arguments);  }  pop() {  return [].pop.call(this);   } } var l = new List(5, 10, 15); l.length; // 3 l[0]; 5 l.pop(); // 15 l.length; // 2

jq

jq does not have "function prototypes" in the strict sense, so this entry focuses on jq function signatures as specified in function definitions.

jq does not limit the number of formal parameters of a function and supports multi-arity functions, but each allowed "restriction" to a particular arity must be specified explicitly.

Although jq does not support varargs functions, their effect can be achieved by using an array-valued argument in conjunction with "destructuring", as illustrated below.

Note also that:

any restrictions on the allowed values of the parameters must be specified programmatically and are only checked at run-time;
function definitions may be included within function definitions;
recursive and mutually recursive functions are allowed, but calls to a function can only occur within the scope of its definition.
a function of a particular arity can be defined more than once, with lexical scoping rules determining how each invocation will be handled.

In the following examples, only `def` and `as` are jq keywords, and .... is used to indicate ellipsis.

def Func: # no arguments def Func(a;b): # two arguments def Vararg(v): v as [$a1, $a2] .... # if v is an array, then $a1 will be the first item specified by v, or else null, and so on def Vararg(a; v): v as [$a1, $a2] .... # if v is an array, then $a1 will be the first item specified by v, or else null, and so on

Julia

Julia does not need or use function prototypes in general. Generic functions are further specialized as to argument type and return type during just-in-time compilation if required. However, when interacting with other languages such a C which use function prototypes, Julia can prototype its functions for passing its functions to external languages with the @cfunction macro:

julia > function mycompare(a, b)::Cint  (a < b) ? -1 : ((a > b) ? +1 : 0)  end mycompare (generic function with 1 method)

Using @cfunction to create a prototype for passing this to C's quicksort:

julia> mycompare_c = @cfunction(mycompare, Cint, (Ref{Cdouble}, Ref{Cdouble}))

Kotlin

The order of declarations in Kotlin is unimportant and so forward declaration of 'top level' functions is neither needed nor supported.

The only place where function (or property) prototypes are needed is for abstract members of classes or interfaces whose implementation will be provided by overriding those members in a derived or implementing class or object.

Here's an example of this. Note that since Kotlin allows arguments to be passed either by name or position for all functions, there is no separate prototype for this situation. Moreover, since arguments may be passed by name, it is strongly recommended (but not obligatory) that the parameter names for overriding members should be the same as for the functions they override. The compiler will issue a warning if this recommendation is not followed.

// version 1.0.6 interface MyInterface {  fun foo() // no arguments, no return type  fun goo(i: Int, j: Int) // two arguments, no return type  fun voo(vararg v: Int) // variable number of arguments, no return type  fun ooo(o: Int = 1): Int // optional argument with default value and return type Int  fun roo(): Int // no arguments with return type Int  val poo: Int // read only property of type Int } abstract class MyAbstractClass {  abstract fun afoo() // abstract member function, no arguments or return type  abstract var apoo: Int // abstract read/write member property of type Int } class Derived : MyAbstractClass(), MyInterface {  override fun afoo() {}  override var apoo: Int = 0  override fun foo() {}  override fun goo(i: Int, j: Int) {}  override fun voo(vararg v: Int) {}  override fun ooo(o: Int): Int = o // can't specify default argument again here but same as in interface  override fun roo(): Int = 2  override val poo: Int = 3 } fun main(args: Array<String>) {  val d = Derived()  println(d.apoo)  println(d.ooo()) // default argument of 1 inferred  println(d.roo())  println(d.poo) }

Output:

0 1 2 3

Lua

function Func() -- Does not require arguments return 1 end function Func(a,b) -- Requires arguments return a + b end function Func(a,b) -- Arguments are optional return a or 4 + b or 2 end function Func(a,...) -- One argument followed by varargs return a,{...} -- Returns both arguments, varargs as table end

Luck

function noargs(): int = ? ;; function twoargs(x:int, y:int): int = ? ;; /* underscore means ignore and is not bound to lexical scope */ function twoargs(_:bool, _:bool): int = ? ;; function anyargs(xs: ...): int = ? ;; function plusargs(x:int, xs: ...): int = ? ;;

M2000 Interpreter

Functions/modules are declared before used. So the flow matter, position in code not matter (perhaps we can put functions in a simple routine, using a label, execute a gosub to label, then make the functions, and then return. Functions amd modules added to a specific list of functions/modules, so every time interpreter check that list (a hash table). They can change definition including signature. Any module/function before executed has no declared local modules/functions. Declarations performed as they executed. For modules, we can prepare a non changed module before begin execute module's code, and when declaration for same name module comes to execute, it just skipped.

Subroutines are parts of functions/modules and first searched from bottom, added to a list of subs positions,therefore they can't changed. Example of change an inner module using another module with same signature. Module MyBeta {Read x : ... } or Module MyBeta (x) { ... } or Module MyBeta(x) { } is the same.

Module Check { Module MyBeta (a) { Print "MyBeta", a/2 } Module TestMe { Module Beta (x) { Print "TestMeBeta", x } Beta 100 } TestMe ; Beta as MyBeta } Check

Signatures needed for Event object. An event object get a list of functions, called as modules, and call every function with same signature. We can provide arguments by reference too. We can define simple functions (without visibility except local and global), or group functions (static groups) with visibility local, global and group level, or we can define local scope functions.

Module Check {, \\ make an event object \\ with a prototype signature \\ first parameter is numeric/object by value, and second is by reference Event Alfa { Read x, &z } \\ make a function with same signature Function ServiceAlfa { read a, &b b+=a } \\ add function to event Event Alfa new &ServiceAlfa() \\ call event in this module var=30 Call Event Alfa, 10, &var Print var=40 \\ make a local module, and pass event by value Module checkinside (ev) { \\ ev is a copy of Alfa m=10 call event ev, 4, &m Print m=14 \\ clear functions from ev Event ev Clear \\ we can call it again, but nothing happen call event ev, 4, &m Print m=14 } checkinside Alfa \\ so now we call Alfa Call Event Alfa, 10, &var Print var=50 Event Alfa Hold \\ calling do nothing, because of Hold state Call Event Alfa, 10, &var Event Alfa Release Call Event Alfa, 10, &var Print var=60 } Check

Using a function for local call (module visibility)

Module Check {, \\ make an event object \\ with a prototype signature \\ first parameter is numeric/object by value, and second is by reference Event Alfa { Read x, &z } \\ make a function with same signature \\ but here prepared to used with current module visibility m=0 Function ServiceAlfa { \ this code "see" m variable \\ we have to use new, to make new a, b for sure read new a, &b b+=a m++ } \\ add function to event, making reference as local to module Event Alfa new Lazy$(&ServiceAlfa()) \\ call event in this module var=30 Call Event Alfa, 10, &var Print var=40 \\ make a local module, and pass event by value Module checkinside (ev) { \\ ev is a copy of Alfa m=10 call event ev, 4, &m Print m=14 \\ clear functions from ev Event ev Clear \\ we can call it again, but nothing happen call event ev, 4, &m Print m=14 } checkinside Alfa \\ so now we call Alfa Call Event Alfa, 10, &var Print var=50 Event Alfa Hold \\ calling do nothing, because of Hold state Call Event Alfa, 10, &var Event Alfa Release Call Event Alfa, 10, &var Print var=60 Print m=4 ' 4 times called ServiceAlfa } Check

Using a Function in a Group (Groups are the User objects in M2000)

Module Check {, \\ make an event object \\ with a prototype signature \\ first parameter is numeric/object by value, and second is by reference Event Alfa { Read x, &z } \\ make a group function with same signature Group IamStatic { m=0 Function ServiceAlfa(a, &b) { b+=a .m++ } } \\ add function to event, making reference as local to module Event Alfa new &IamStatic.ServiceAlfa() \\ call event in this module var=30 Call Event Alfa, 10, &var Print var=40 \\ make a local module, and pass event by value Module checkinside (ev) { \\ ev is a copy of Alfa m=10 call event ev, 4, &m Print m=14 \\ clear functions from ev Event ev Clear \\ we can call it again, but nothing happen call event ev, 4, &m Print m=14 } checkinside Alfa \\ so now we call Alfa Call Event Alfa, 10, &var Print var=50 Event Alfa Hold \\ calling do nothing, because of Hold state Call Event Alfa, 10, &var Event Alfa Release Call Event Alfa, 10, &var Print var=60 Print IamStatic.m=4 ' 4 times called IamStatic.ServiceAlfa } Check

Nim

Procedure declarations can be used if a proc is to be used before its definition.

# Procedure declarations. All are named proc noargs(): int proc twoargs(a, b: int): int proc anyargs(x: varargs[int]): int proc optargs(a, b: int = 10): int # Usage discard noargs() discard twoargs(1,2) discard anyargs(1,2,3,4,5,6,7,8) discard optargs(5) # Procedure definitions proc noargs(): int = echo "noargs" proc twoargs(a, b: int): int = echo "twoargs" proc anyargs(x: varargs[int]): int = echo "anyargs" proc optargs(a: int, b = 10): int = echo "optargs"

OCaml

(* Usually prototype declarations are put in an interface file,  a file with .mli filename extension *) (* A prototype declaration for a function that does not require arguments *) val no_arg : unit -> unit (* A prototype declaration for a function that requires two arguments *) val two_args : int -> int -> unit (* A prototype declaration for a function that utilizes optional arguments *) val opt_arg : ?param:int -> unit -> unit (* in this case we add a unit parameter in order to omit the argument,  because ocaml supports partial application *) (* A prototype declaration for a function that utilizes named parameters *) val named_arg : param1:int -> param2:int -> unit (* An explanation and example of any special forms of prototyping not covered by the above *) (* A prototype declaration for a function that requires a function argument *) val fun_arg : (int -> int) -> unit (* A prototype declaration for a function with polymorphic argument *) val poly_arg : 'a -> unit

Oforth

Oforth can only forward declare methods (see Mutual Recursion task). A method can be declared without any class implementation :

Method new: myMethod

This creates a new method object with name myMethod (or does nothing if this object already exists). It says nothing about method implementations (number of parameters, return value, ...).

A method object is not directly related to classes :

- A method object is created.

- Classes are created.

- Into classes, you can create implementations for particular methods. If, at this point, the method object does not exist yet, it is created.

Ol

Ol have no function prototypes.

Ol functions is a first-class functions with dynamic arguments translation so no function prototypes is required.

OxygenBasic

'DECLARE FUNCTION' ABBREVIATED TO '!' ! f() ' a procedure with no params ! f(int a) ' with 1 int param ! f(int *a) ' with 1 int pointer param ! f(int a, int b, inc c) ' with 3 int params ! f(int a,b,c) ' compaction with 3 int params ! f(string s, int a,b) ' with 1 string and 2 int params ! f() as string ' function returning a string ! f(string s) as string ' with 1 string param ! *f(string s) as string ' as a function pointer: @f=address ! f(string s, optional i) ' with opptional param ! f(string s = "Hello") ' optional param with default value ! f(int n, ...) ' 1 specific param and varargs ! f(...) ' any params or none 'TRADITIONAL BASIC DECLARATIONS declare sub f( s as string, i as long, j as long) ' byref by default declare function f( byref s as string, byval i as long, byval j as long) as string 'C-STYLE DECLARATIONS void f(string *s, long i, long j) string f(string *s, long i, long j) 'BLOCK DIRECTIVES FOR FUNCTION PROTOTYPES: extern ' shareable stdcall functions extern lib "xyz.dll" ' for accessing functions in xyz Dynamic Link Library extern export ' functions to be exported if this is a DLL extern virtual ' for accssing interfaces and other virtual classes end extern ' revert to internal function mode

PARI/GP

GP does not use function prototypes.

PARI uses C prototypes. Additionally, gp2c parser codes are essentially function prototypes. They must be placed in the file called by gp2c, not in a file included by it, and they must appear as a GP; comment. For a function

long foo(GEN a, GEN b)

which takes two (required) t_INT arguments, returns a small integer (a C long) and appears as bar to the gp interpreter, the following command would be used:

/* GP;install("foo","LGG","bar","./filename.gp.so"); */

If its arguments were optional it could be coded as

/* GP;install("foo","LDGDG","bar","./filename.gp.so"); */

although other parser codes are possible; this one sends NULL if the arguments are omitted.

A code like

/* GP;install("foo","s*","bar","./filename.gp.so"); */

can be used to take a variable (0 or more) number of arguments. Note that the return type in this case is implicitly GEN

Other special forms are described in the User's Guide to the PARI library, section 5.7.3.

Perl

The perl scripting language allows prototypes to be checked during JIT compilation. Prototypes should be placed before subroutine definitions, declarations, or anonymous subroutines. The sigil special symbols act as argument type placeholders.

sub noargs(); # Declare a function with no arguments sub twoargs($$); # Declare a function with two scalar arguments. The two sigils act as argument type placeholders sub noargs :prototype(); # Using the :attribute syntax instead sub twoargs :prototype($$);

Phix

Library: Phix/basics

Use explicit forward definitions. Optional, unless (for instance) you want to use named parameters in a forward call.
Should be identical to the actual definition, but preceded by "forward" and with no body.

forward function noargs() -- Declare a function with no arguments forward procedure twoargs(integer a, integer b) -- Declare a procedure with two arguments forward procedure twoargs(integer, integer /*b*/) -- Parameter names are optional in forward (and actual) definitions forward function anyargs(sequence s) -- varargs are [best/often] handled as a (single) sequence in Phix forward function atleastonearg(integer a, integer b=1, ...); -- Default makes args optional (== actual defn)

No special syntax is needed on actual or forward function definitions for named parameters, and calls are identical whether still forward or now fully defined.
Defaults on optional parameters in forward definitions can also be dummy (but type compatible) values should the actual not yet be defined.

PL/I

declare s1 entry; declare s2 entry (fixed); declare s3 entry (fixed, float); declare f1 entry returns (fixed); declare f2 entry (float) returns (float); declare f3 entry (character(*), character(*)) returns (character (20));

PureBasic

PureBasic defines both functions and procedures by using the keyword Procedure. For the purposes of this task I will describe both 'procedures' and 'functions' using only the term 'procedure'. PureBasic uses a one-pass compiler. All procedures need to be defined before use. The prototypes referred to in the task description are performed with forward declarations in PureBasic.

PureBasic allows two types of prototyping. The first uses the keyword Declare and describes the name, return value, and parameters of a procedure. It is identical in form with the first line of a procedure definition with the exception that the keyword Declare is used instead of the keyword 'Procedure. It must be placed before the first use of the procedure and must occur before the procedure's definition. The procedure declaration's parameteres need to match the order, type, and number of those in the procedure's definition, though their names may be different.

The keyword ProtoType may be used for pointers to procedures so that a definition of the parameters and return type for the function being pointed to are defined and that the pointer may be used to execute the function with type checking. The parameter names do not have to match in the 'ProtoType' definition but the order, type and optional parameters do. 'ProtoTypes' must be defined before their first use.

PureBasic does not allow either variable arguments or named parameters.

;Forward procedure declare defined with no arguments and that returns a string Declare.s booTwo() ;Forward procedure declare defined with two arguments and that returns a float Declare.f moo(x.f, y.f) ;Forward procedure declare with two arguments and an optional argument and that returns a float Declare.f cmoo(x.f, y.f, m.f = 0) ;*** The following three procedures are defined before their first use. ;Procedure defined with no arguments and that returns a string Procedure.s boo(): ProcedureReturn "boo": EndProcedure ;Procedure defined with two arguments and that returns an float Procedure.f aoo(x.f, y.f): ProcedureReturn x + y: EndProcedure ;Procedure defined with two arguments and an optional argument and that returns a float Procedure.f caoo(x.f, y.f, m.f = 1): ProcedureReturn (x + y) * m: EndProcedure ;ProtoType defined for any function with no arguments and that returns a string Prototype.s showString()  ;Prototype defined for any function with two float arguments and that returns a float Prototype.f doMath(x.f, y.f) ;ProtoType defined for any function with two float arguments and an optional float argument and that returns a float Prototype.f doMathWithOpt(x.f, y.f, m.f = 0)  Define a.f = 12, b.f = 5, c.f = 9 Define proc_1.showString, proc_2.doMath, proc_3.doMathWithOpt ;using defined ProtoTypes If OpenConsole("ProtoTypes and Forward Declarations")    PrintN("Forward Declared procedures:")  PrintN(boo())  PrintN(StrF(a, 2) + " * " + StrF(b, 2) + " = " + StrF(moo(a, b), 2))  PrintN(StrF(a, 2) + " * " + StrF(b, 2) + " + " + StrF(c, 2) + " = " + StrF(cmoo(a, b, c), 2))  PrintN(StrF(a, 2) + " * " + StrF(b, 2) + " = " + StrF(cmoo(a, b), 2))    ;set pointers to second set of functions  proc_1 = @boo()  proc_2 = @aoo()  proc_3 = @caoo()    PrintN("ProtoTyped procedures (set 1):")  PrintN(proc_1())  PrintN(StrF(a, 2) + " ? " + StrF(b, 2) + " = " + StrF(proc_2(a, b), 2))  PrintN(StrF(a, 2) + " ? " + StrF(b, 2) + " ? " + StrF(c, 2) + " = " + StrF(proc_3(a, b, c), 2))  PrintN(StrF(a, 2) + " ? " + StrF(b, 2) + " = " + StrF(proc_3(a, b), 2))    ;set pointers to second set of functions  proc_1 = @booTwo()  proc_2 = @moo()  proc_3 = @cmoo()    PrintN("ProtoTyped procedures (set 2):")  PrintN(proc_1())  PrintN(StrF(a, 2) + " ? " + StrF(b, 2) + " = " + StrF(proc_2(a, b), 2))  PrintN(StrF(a, 2) + " ? " + StrF(b, 2) + " ? " + StrF(c, 2) + " = " + StrF(proc_3(a, b, c), 2))  PrintN(StrF(a, 2) + " ? " + StrF(b, 2) + " = " + StrF(proc_3(a, b), 2))      Print(#CRLF$ + #CRLF$ + "Press ENTER to exit"): Input()  CloseConsole() EndIf ;*** If the forward Declaration above are not used then the following Procedure ;definitions each have to be placed before the call to the respective procedure. ;Procedure defined with no arguments and that returns a string Procedure.s booTwo()  ProcedureReturn "booTwo" EndProcedure ;Procedure defined with two arguments and that returns an float Procedure.f moo(x.f, y.f)  ProcedureReturn x * y EndProcedure ;Procedure defined with two arguments and an optional argument and that returns an float Procedure.f cmoo(x.f, y.f, m.f = 0)  ProcedureReturn (x * y) + m EndProcedure

Sample output:

Forward Declared procedures: boo 12.00 * 5.00 = 60.00 12.00 * 5.00 + 9.00 = 69.00 12.00 * 5.00 = 60.00 ProtoTyped procedures (set 1): boo 12.00 ? 5.00 = 17.00 12.00 ? 5.00 ? 9.00 = 153.00 12.00 ? 5.00 = 0.00 ProtoTyped procedures (set 2): booTwo 12.00 ? 5.00 = 60.00 12.00 ? 5.00 ? 9.00 = 69.00 12.00 ? 5.00 = 60.00

Quackery

In Quackery a "word" corresponds to a function or subroutine. If you want to make a forward declaration (typically but not exclusively for recursive or mutually recursive words) you would use forward is, and resolve the forward declaration with resolves.

For example, the naive recursive Fibonacci function. (Note that the texts in parentheses are stack comments and can be omitted. Their inclusion is good practice. No declaration of parameters or arguments is required in Quackery.)

 forward is fibonacci ( n --> n ) [ dup 2 < if done dup 1 - fibonacci swap 2 - fibonacci + ] resolves fibonacci ( n --> n )

Racket

Most of the points are covered in this program

#lang racket (define (no-arg) (void)) (define (two-args a b) (void)) ;arguments are always named (define (varargs . args) (void)) ;the extra arguments are stored in a list (define (varargs2 a . args) (void)) ;one obligatory argument and the rest are contained in the list (define (optional-arg (a 5)) (void)) ;a defaults to 5

(void) is a function that returns "nothing", so this are prototypes that do nothing. Although standard Racket doesn't allow type declarations, it allows contracts, so we can add this to the previous declarations

(provide (contract-out  [two-args (integer? integer? . -> . any)]))

then any module that imports the function can only pass integers to two-args.

Another way is using the typed/racket language, like this

#lang typed/racket (: two-args (Integer Integer -> Any)) (define (two-args a b) (void))

Raku

(formerly Perl 6) There is no restriction on placement of prototype declarations. (Actually, we call them "stub declarations".) In fact, stub declarations are rarely needed in Raku because post-declaration of functions is allowed, and normal function declarations do not bend the syntax the way they sometimes do in Perl 5.

Note that the ... in all of these stub bodies is literally part of the declaration syntax.

A prototype declaration for a function that does not require arguments (and returns an Int):

sub foo ( --> Int) {...}

A prototype declaration for a function that requires two arguments. Note that we can omit the variable name and just use the sigil in the stub, since we don't need to reference the argument until the actual definition of the routine. Also, unlike in Perl 5, a sigil like @ defaults to binding a single positional argument.

sub foo (@, $ --> Int) {...}

A prototype declaration for a function that utilizes varargs after one required argument. Note the "slurpy" star turns the @ sigil into a parameter that accepts all the rest of the positional arguments.

sub foo ($, *@ --> Int) {...}

A prototype declaration for a function that utilizes optional arguments after one required argument. Optionality is conferred by either a question mark or a default:

sub foo ($, $?, $ = 42 --> Int) {...}

A prototype declaration for a function that utilizes named parameters:

sub foo ($, :$faster, :$cheaper --> Int) {...}

Example of prototype declarations for subroutines or procedures, which in Raku is done simply by noting that nothing is returned:

sub foo ($, $ --> Nil) {...}

A routine may also slurp up all the named arguments that were not bound earlier in the signature:

sub foo ($, :$option, *% --> Int) {...}

A routine may make a named parameter mandatory using exclamation mark. (This is more useful in multi subs than in stubs though.)

sub foo ($, :$option! --> Int) {...}

A routine may unpack an Array automaticly. Here the first element is stored in a scalar and the rest in an Array. Other buildin types can be unpacked as well.

sub foo ([$, @]) {...}

A routine may ask for a closure parameter to implement higher order functions. Typed or untyped signatures can be supplied.

sub foo (@, &:(Str --> Int)) {...}

REXX

Include How to use
Include Source code

In the REXX language, there is no difference between functions and subroutines, except that functions must return a value, even if that value is a "null" (empty string).

In REXX, if a function doesn't return a value, a syntax condition is raised.

Being an interpreted language, REXX has no need of declaring or prototyping for functions or subroutines, but there are facilities (in the form of BIFs) to assist the REXX programmer to easily determine the number of arguments passed (if any), and perform (and/or enforce) any necessary argument passing (including the type of values or variables passed), and also including checking for omitted arguments. In effect, the relaxation of requirements/rules for function or subroutine invocations has been moved from the compile stage (for REXX, the parsing/interpretive) stage) to the execution stage.

In module Math.inc you'll find hundreds of procedures showing all kinds of argument checking.

SNOBOL4

In SNOBOL4, functions are actually a hack and are defined in an idiosyncratic way that is simultaneously like a prototype or not like one as the case may be.

Basics

To begin with, we look at the definition provided at the relevant task page:

 define('multiply(a,b)') :(mul_end) multiply multiply = a * b  :(return) mul_end * Test output = multiply(10.1,12.2) output = multiply(10,12) end

The key to this is the define() BIF which declares the actual function and the multiply label which is the entry point to the code that is executed. The key is that SNOBOL4 is an almost militantly unstructured language. There is absolutely nothing special about the multiply entry point that distinguishes it from the target of any other branch target. What happens instead is that the define() BIF associates a certain string pattern--the prototype, in effect--with an entry point. The :(mul_end) piece at the end, in fact, exists because were it not present the body of the multiply "function" would be executed: it is a branch to the label mul_end.

On execution, the SNOBOL4 runtime will execute line by line of the script. When it reaches the define BIF call it will do the stuff it needs to do behind the scenes to set up function-like access to the multiply branch target. It would then proceed to execute the next line were it not for the branch.

Separation of prototype and body

Of course this implies that you can separate the two pieces. Which you can, like this:

 define('multiply(a,b)') * * Assume lots more code goes here. *  :(test) * * MORE CODE! * multiply multiply = a * b  :(return) * * MORE CODE! * test output = multiply(10.1,12.2) output = multiply(10,12) end

With this structure the "function" is declared at the program, the implementation is somewhere down in the middle, and the mainline (test here) is at the end.

Full prototyping

The define() BIF is used for more than merely providing function-like access to a label with the same name. It is used to prototype all of these (with some default behaviour):

the function name (multiply in the examples);
the formal arguments to the function (a, b in the examples);
the entry point label for the function's code (defaults to the function name, mult_impl in the following example);
any local variables which should be protected in the function (defaults to none, acc1,acc2 in the following example).

Thus a highly-contrived example function that illustrates all of these would look like this:

 define('multiply(a,b)acc1,acc2','mult_impl') :(mult_end) mult_impl acc1 = a acc2 = b multiply = acc1 * acc2  :(return) mult_end * Test output = multiply(10.1,12.2) output = multiply(10,12) end

Wren

Firstly, Wren makes a distinction between functions and methods. The latter are always members of a class and never need to be prototyped regardless of the order in which they are declared or called.

On the other hand functions are standalone objects and cannot be called before they have been declared. Consequently, prototypes are required if a function calls itself recursively, if two (or more) functions are mutually recursive or if a function is simply called out of order for some reason.

A prototype is just a forward declaration of the function's name. Details of any parameters are not needed and cannot even be optionally specified as parameters are considered to be part of the function's body.

In the following example, the 'factorial' function is recursive and so needs a forward declaration. However, even though the function takes a single argument, no prior information about that is needed or possible. There is an example of mutual recursion protoyping in the Mutual_recursion#Wren task.

var factorial // forward declaration factorial = Fn.new { |n| (n <= 1) ? 1 : factorial.call(n-1) * n } System.print(factorial.call(5))

Output:

zkl

In zkl, all functions are var args. Prototypes provide some documentation and an overlay on the incoming args. Named parameters are not supported.

fcn{"Hello World"} // no expected args fcn(){"Hello World"} // ditto fcn{vm.arglist}(1,2) // ask the VM for the passed in args -->L(1,2) fcn f(a,b){a+b} // fcn(1,2,3) works just fine fcn f(args){}(1,2,3) //args = 1 fcn(args){vm.arglist.sum()}(1,2,3) //-->6 fcn(a=1,b=2){vm.arglist}() //-->L(1,2) fcn(a=1,b=2){vm.arglist}(5) //-->L(5,2) fcn(a=1,b){vm.arglist}() //-->L(1), error if you try to use b fcn(a,b=2){vm.arglist}(5) //-->L(5,2) fcn(a,b=2,c){vm.arglist}(1) //-->L(1,2) fcn(){vm.nthArg(1)}(5,6) //-->6 fcn{vm.numArgs}(1,2,3,4,5,6,7,8,9) //-->9 fcn{vm.argsMatch(...)} // a somewhat feeble attempt arg pattern matching based on type (vs value) // you can do list assignment in the prototype: fcn(a,[(b,c)],d){vm.arglist}(1,L(2,3,4),5) //-->L(1,L(2,3,4),5) fcn(a,[(b,c)],d){"%s,%s,%s,%s".fmt(a,b,c,d)}(1,L(2,3,4),5) //-->1,2,3,5 // no type enforcement but you can give a hint to the compiler fcn([Int]n){n.sin()} //--> syntax error as Ints don't do sin