Generic swap

From Rosetta Code
Generic swap
You are encouraged to solve this task according to the task description, using any language you may know.

Write a generic swap function or operator which exchanges the values of two variables (or, more generally, any two storage places that can be assigned), regardless of their types.

If your solution language is statically typed please describe the way your language provides genericity.

If variables are typed in the given language, it is permissible that the two variables be constrained to having a mutually compatible type, such that each is permitted to hold the value previously stored in the other without a type violation. That is to say, solutions do not have to be capable of exchanging, say, a string and integer value, if the underlying storage locations are not attributed with types that permit such an exchange.

Generic swap is a task which brings together a few separate issues in programming language semantics.

Dynamically typed languages deal with values in a generic way quite readily, but do not necessarily make it easy to write a function to destructively swap two variables, because this requires indirection upon storage places or upon the syntax designating storage places.

Functional languages, whether static or dynamic, do not necessarily allow a destructive operation such as swapping two variables regardless of their generic capabilities.

Some static languages have difficulties with generic programming due to a lack of support for (Parametric Polymorphism).

Do your best!


F swap(&a, &b)
   (a, b) = (b, a)

360 Assembly

Three consecutive exclusive OR's swap variable contents

SWAP     CSECT ,                   control section start 
         BAKR  14,0                stack caller's registers 
         LR    12,15               entry point address to reg.12 
         USING SWAP,12             use as base 
         MVC   A,=C'5678____'      init field A 
         MVC   B,=C'____1234'      init field B 
         LA    2,L                 address of length field in reg.2 
         WTO   TEXT=(2)            Write To Operator, results in: 
*                                  +5678________1234 
         XC    A,B                 XOR A,B 
         XC    B,A                 XOR B,A 
         XC    A,B                 XOR A,B. A holds B, B holds A 
         WTO   TEXT=(2)            Write To Operator, results in: 
*                                  +____12345678____ 
         PR    ,                   return to caller 
         LTORG ,                   literals displacement 
L        DC    H'16'               halfword containg decimal 16 
A        DS    CL8                 field A, 8 bytes 
B        DS    CL8                 field B, 8 bytes 
         END   SWAP                program end

6502 Assembly

There are no atomic swap operations in 6502 Assembly but there are a few slower ways of doing it, using the stack. The 65c02 and 65816 revisions added a few commands which make swapping easier. The need for atomic swapping isn't really an issue since most 6502-based computers didn't use parallel processing; the biggest concern for registers or memory getting clobbered was from interrupts, which can be easily avoided by restricting the registers and/or memory they use, and using the stack to back up and restore their values at the beginning and end.

NMOS 6502

The original 6502 can't push X or Y onto the stack directly, and must do so through the accumulator. The most common way of swapping values is by pushing them onto the stack and popping them in the "wrong" order on purpose.

;swap X with Y
pha        ;push accumulator
   pha     ;push x
       pha ;push y
       tax ;pop y into x
   tay     ;pop x into y
pla        ;pop accumulator

Swapping A with X or Y is easiest with using zero page memory as a temporary store, or with self-modifying code.

sta temp     ;a label representing a zero page memory address
pha          ;push x
    ldx temp
pla          ;pop x into accumulator

Swapping memory locations is fairly straightforward on all versions of the 6502:

temp  equ $00
temp2 equ $01 ;these values don't matter.

lda temp
pha          ;push temp
lda temp2
pha          ;push temp2

sta temp     ;pop temp2 into temp
sta temp2    ;pop temp into temp2

CMOS 65c02

This revision of the 6502 can push X and Y onto the stack directly. The above code can be condensed quite a bit. This is a different example that better illustrates how swapping can be done:

lda #$33
ldx #$44
ldy #$55


pla ;a=#$55
ply ;y=#$44
plx ;x=#$33

68000 Assembly

Two registers can be swapped with EXG:

EXG D0,D1 ;swap the contents of D0 and D1
EXG A0,A1 ;swap the contents of A0 and A1

However, these commands cannot be used with memory. If you want to do that you'll need to use the stack or a data register.

MOVE.L ($00FF0000),D0
MOVE.L ($00FF1000),D1
MOVE.L D0,($00FF1000)
MOVE.L D1,($00FF0000)

The SWAP switches the high and low words of a data register. It can't be used with memory either.

MOVE.L #$1234ABCD,D0
SWAP D0              ;now D0 = #$ABCD1234

8086 Assembly

The XCHG command swaps two registers' values. It can also work with memory that is pointed to by an index register such as SI, DI,or BP. Both registers must be the same size, which enforces type matching. You cannot XCHG AX,CL for example since AX is 16-bit and CL is only 8-bit.

Here are some examples of the XCHG command in action:

xchg ax,bx    ;exchanges ax with bx

xchg ah,al    ;swap the high and low bytes of ax

;XCHG a register with memory
mov dx,0FFFFh
mov word ptr [ds:userRam],dx
mov si,offset userRam
mov ax,1234h
xchg ax,[si]   ;exchange ax with the value stored at userRam. Now, ax = 0FFFFh and the value stored at userRam = 1234h

;XCHG a register with a value on the stack.
mov ax,1234h
mov bx,4567h
push bx
push bp
mov bp,sp      ;using [sp] as an operand for XCHG will not work. You need to use bp instead.

xchg ax,[2+bp] ;exchanges AX with the value that was pushed from BX onto the stack. Now, AX = 4567h,
               ;and the entry on the stack just underneath the top of the stack is 1234h.



Or to swap between the stack and a var:

dup @ -rot !


(defun swap (pair)
   (cons (cdr pair)
         (car pair)))

(let ((p (cons 1 2)))
  (cw "Before: ~x0~%After: ~x1~%" p (swap p)))


INCLUDE "D2:REAL.ACT" ;from the Action! Tool Kit

PROC Swap(BYTE POINTER ptr1,ptr2 INT size)
  BYTE tmp
  INT i

  FOR i=0 TO size-1
    tmp=ptr1^ ptr1^=ptr2^ ptr2^=tmp
    ptr1==+1 ptr2==+1

PROC Main()
  BYTE b1=[13],b2=[56]
  INT i1=[65234],i2=[534]
  REAL r1,r2
  CHAR ARRAY s1="abcde",s2="XYZ"

  Put(125) PutE() ;clear the screen
  ValR("32.5",r1) ValR("-0.63",r2)

  Print("Swap bytes: ")
  PrintB(b1) Put(32) PrintB(b2) Print(" -> ")
  PrintB(b1) Put(32) PrintBE(b2)

  Print("Swap integers: ")
  PrintI(i1) Put(32) PrintI(i2) Print(" -> ")
  PrintI(i1) Put(32) PrintIE(i2)

  Print("Swap floats: ")
  PrintR(r1) Put(32) PrintR(r2) Print(" -> ")
  PrintR(r1) Put(32) PrintRE(r2)

  Print("Swap strings: ")
  Print(s1) Put(32) Print(s2) Print(" -> ")
  Print(s1) Put(32) PrintE(s2)

Screenshot from Atari 8-bit computer

Swap bytes: 13 56 -> 56 13
Swap integers: -302 534 -> 534 -302
Swap floats: 32.5 -0.63 -> -0.63 32.5
Swap strings: abcde XYZ -> XYZ abcde


The generic parameters for an Ada generic procedure are defined in a procedure specification, while the algorithm is defined in a procedure body. The first code snippet is the procedure specification. The second code snippet is the procedure body.

   type Swap_Type is private; -- Generic parameter
procedure Generic_Swap (Left, Right : in out Swap_Type);

procedure Generic_Swap (Left, Right : in out Swap_Type) is
   Temp : constant Swap_Type := Left;
   Left := Right;
   Right := Temp;
end Generic_Swap;


To use the generic swap procedure, you need to instantiate the procedure for each type that you intend to use.

with Generic_Swap;
type T is ...
procedure T_Swap is new Generic_Swap (Swap_Type => T);
A, B : T;
T_Swap (A, B);


Aime is statically typed. A generic swap utility may nonetheless be defined in terms of parameters of unspecified type and pass by reference.

__swap(&, &,,)
    set(0, $3);
    set(1, $2);

swap(&, &)
    xcall(xcall, __swap);


A generic swap operator =:= was proposed in ALGOL Bulletin for standard ALGOL 68 so that the compiler could optimise the operation. However such an operator was not adopted and needs to be manually defined for each mode required.

Works with: ALGOL 68 version Revision 1 - no extensions to language used
Works with: ALGOL 68G version Any - tested with release 1.18.0-9h.tiny
Works with: ELLA ALGOL 68 version Any (with appropriate job cards) - tested with release 1.8-8d

GENMODE v1:="Francis Gary Powers", v2:="Vilyam Fisher";

PRIO =:= = 1;

OP =:= = (REF GENMODE v1, v2)VOID: (
  GENMODE tmp:=v1; v1:=v2; v2:=tmp

v1 =:= v2;

print(("v1: ",v1, ", v2: ", v2, new line))
v1: Vilyam Fisher, v2: Francis Gary Powers

Special option

The B6700 Algol compiler offered access to a special machine operation via a function called ReadLock(a,b) that could be invoked on a variety of operands. By using the ability to #define this = that; one could define Swap(a,b) to be a:=ReadLock(a,b) to attain the appearance of a Swap operation. This all relied on the working of magnetic core memory, specifically that to read a word, the word is made zero and in the process the memory hardware notes which bits were thereby flipped. Thus it passes on the value in the word and meanwhile, rewrites that content back to the word so as to preserve its value on reading. Similarly, to write a value to a word, the word is first zeroed.

ReadLock(a,b) functioned by reading a and writing its value to b, but also, recovering the value that was in b which it returns as the result of the function - which is written to a by the assignment, completing the swap. The ReadLock part is "atomic" or not interruptable, so it is used in semaphores and the like, but was available for other use. It swapped a single word, so could swap types such as integers or floating-point numbers (single precision) thus being somewhat generic.

Amazing Hopper

#include <flow.h>

   DIM(10) AS-INT-RAND( 10, random array )
   SET( single var, 0.5768 )
   PRNL( "SINGLE VAR: ", single var, "\nRANDOM ARRAY: ", random array )
   single var <-> random array
   PRNL( "SINGLE VAR: ", single var, "\nRANDOM ARRAY: ", random array )
$ hopper swap.flw 
SINGLE VAR: 0.5768
RANDOM ARRAY: 4,2,5,10,8,3,4,8,3,5
SINGLE VAR: 4,2,5,10,8,3,4,8,3,5


The simpler way to write a swap is to use the Amiga E ability to return multiple values. All basic data type in Amiga E can be held by its LONG type, and complex data type (like lists) are indeed pointers (which fits into a LONG too); so, because of the fact that Amiga E is not strongly typed, this solution works for any type.

PROC swap(a,b) IS b,a

PROC main()
  DEF v1, v2, x
  v1 := 10
  v2 := 20
  v1, v2 := swap(v1,v2)
  WriteF('\d \d\n', v1,v2)           -> 20 10
  v1 := [ 10, 20, 30, 40 ]
  v2 := [ 50, 60, 70, 80 ]
  v1, v2 := swap(v1,v2)
  ForAll({x}, v1, `WriteF('\d ',x))  -> 50 60 70 80
  ForAll({x}, v2, `WriteF('\d ',x))  -> 10 20 30 40


AppleScript has built-in support for swapping. This is generic and works for all combinations of data types.

set {x,y} to {y,x}


(mac myswap (a b)
     (w/uniq gx
             `(let ,gx a
                   (= a b)
                   (= b ,gx))))

(with (a 1
       b 2)
      (myswap a b)
      (prn "a:" a #\Newline "b:" b))


Translation of: Ruby

Using multi-assignment

swap: function [a,b]-> @[b,a]

c: 1
d: 2
print [c d]

[c,d]: swap c d
print [c d]
1 2 
2 1

In-place modification

swap: function [a,b].exportable [ 
    tmp: var b
    let b var a
    let a tmp 

c: 1
d: 2
print [c d]

swap 'c 'd
print [c d]
1 2 
2 1


Swap(ByRef Left, ByRef Right)
    temp := Left
    Left := Right
    Right := temp


    printf("%s version %s\n",ARGV[0],PROCINFO["version"])
    foo = 1
    bar = "a"
    printf("\n%s %s\n",foo,bar)
    rc = swap("foo","bar") # ok
    printf("%s %s %s\n",foo,bar,rc?"ok":"ng")
    printf("\n%s %s\n",foo,bar)
    rc = swap("FOO","BAR") # ng
    printf("%s %s %s\n",foo,bar,rc?"ok":"ng")
function swap(a1,a2,  tmp) { # strings or numbers only; no arrays
    if (a1 in SYMTAB && a2 in SYMTAB) {
      if (isarray(SYMTAB[a1]) || isarray(SYMTAB[a2])) {
      tmp = SYMTAB[a1]
      SYMTAB[a1] = SYMTAB[a2]
      SYMTAB[a2] = tmp
gawk version 4.1.0

1 a
a 1 ok

a 1
a 1 ng


The Exch() command can swap data of any size at any two addresses. This example swaps two 2-byte variables.



Applesoft BASIC

A=43:B=47:H=A:A=B:B=H:?" A="A" B="B;
 A=47 B=43


x = 1
y$ = "hello"

SWAP x, y$


global a, b
a = "one"
b = "two"

print a, b
call swap(a, b)
print a, b

subroutine swap(a, b)
  temp = a : a = b : b = temp
end subroutine


Built-in function

      a = 1.23 : b = 4.56
      SWAP a,b
      PRINT a,b
      a$ = "Hello " : b$ = "world!"
      SWAP a$,b$
      PRINT a$,b$

Custom function

      a = 1.23 : b = 4.56
      PROCswap(^a,^b, 5)
      PRINT a,b
      a$ = "Hello " : b$ = "world!"
      PROCswap(^a$,^b$, 6)
      PRINT a$,b$
      DEF PROCswap(a%, b%, s%)
      LOCAL i%
      FOR i% = 0 TO s%-1
        SWAP a%?i%,b%?i%
      4.56      1.23
world!    Hello

Chipmunk Basic

The GW-BASIC solution works without any changes.


FreeBASIC already has a built-in generic Swap procedure but a macro can be used to build another one:

' FB 1.05.0
#Macro Declare_Swap(T)
Sub Swap_##T(ByRef t1 As T, ByRef t2 As T)
  Dim temp As T = t2
  t2 = t1
  t1 = temp
End Sub

Dim As Integer i, j
i = 1 : j = 2

Declare_Swap(Integer) ' expands the macro
Swap_Integer(i, j)
Print i, j

Dim As String s, t
s = "Hello" : t = "World"

Swap_String(s, t)
Print s, t

Print "Press any key to exit"
 2             1
World         Hello


window 1, @"Generic Swap", (0,0,480,270)

text ,,,,, 60

long        i, j
double      x, y
CFStringRef a, b

i = 1059 : j = 62
print i, j
swap i, j
print i, j


x = 1.23 : y = 4.56
print x, y
swap x, y
print x, y


a = @"Hello" : b = @"World!"
print a, b
swap a, b
print a, b

 1059            62
 62              1059

 1.23            4.56
 4.56            1.23

Hello           World!
World!          Hello


Click this link to run this code

Public Sub Main()
Dim vA As Variant = " World"
Dim vB As Variant = 1

Swap vA, vB

Print vA; vB



1 World


Works with: Applesoft BASIC
Works with: Chipmunk Basic
Works with: MSX_BASIC
Works with: PC-BASIC version any
Works with: QBasic
10 CLS : REM  10 HOME for Applesoft BASIC
20 A = 1 : B = 2
30 PRINT " A=";A," B=";B
40 T = A : A = B : B = T
50 PRINT " A=";A," B=";B
60 END
 A=1     B=2
 A=2     B=1


120 END DEF 
130 LET A=1:LET B=2

Minimal BASIC

Works with: QBasic
Works with: QuickBasic
Works with: Applesoft BASIC
Works with: BASICA
Works with: Chipmunk Basic
Works with: GW-BASIC
Works with: IS-BASIC
Works with: MSX_Basic
Works with: Just BASIC
Works with: Liberty BASIC
Works with: Run BASIC
10 LET A = 1
20 LET B = 2
30 PRINT " A=";A;"  B=";B
40 LET T = A
50 LET A = B
60 LET B = T
70 PRINT " A=";A;"  B=";B
80 END

MSX Basic

The GW-BASIC solution works without any changes.


  macro Swap(a,b,    c)
    typeof(a) c
  end macro

  'demo with compound types:
  type point { float x,y}
  point p={1,2}
  point q={3,4}
  swap p,q
  print "p: "  p.x "," p.y    'p: 3,4
  print "q: "  q.x "," q.y    'q: 1,2


Built in function:

Swap a, b


QBasic already has a generic swap procedure built in, but a new subroutine can be defined:

SUB nswap (a, b)
    temp = a: a = b: b = temp

a = 1
b = 2

PRINT a, b
CALL nswap(a, b)
PRINT a, b


The Minimal BASIC solution works without any changes.


Run BASIC does not have support for swapping built in:

a = 1
b = 2
'----- swap ----
tmp = a
a   = b
b  = tmp


Generic function, swap the content of two variables.

Swap Var1, Var2


TI-89 BASIC is dynamically typed, so the genericity is implicit. It has no pass by reference, so we must pass the variable names as strings. It is dynamically scoped, so we must choose hopefully distinct names for the variables.

Define swap(swapvar1, swapvar2) = Prgm
  Local swaptmp
  #swapvar1 → swaptmp
  #swapvar2 → #swapvar1
  swaptmp → #swapvar2

1 → x
2 → y
swap("x", "y")


    LET a = 11
    LET b = 22
    PRINT a, "  ", b
    GOSUB 100
    PRINT a, "  ", b
100 REM swap(a, b)
    LET t = a 
    LET a = b 
    LET b = t


SUB swap(a, b)
    LET temp = a
    LET a = b
    LET b = temp

LET a = 1
LET b = 2

PRINT a, b
CALL swap(a, b)
PRINT a, b


Since uBasic/4tH has a stack (just like Forth) and it is an integer BASIC only, this is quite trivial. However, making a function or procedure with the same functionality is impossible, because there is no way to pass variables by reference.

a = 5 : b = 7
Print a,b
Push a,b : a = Pop() : b = Pop()
Print a,b


This works for everything: strings, dates, booleans ... The fact is, with everything being a Variant, it's always generic.

sub swap( byref x, byref y )
	dim temp
	temp = x
	x = y
	y = temp
end sub


dim a 
a = "woof"
dim b
b = now()
swap a,b
wscript.echo a
wscript.echo b
5/02/2010 2:35:36 PM

Visual Basic

Visual Basic can use the VBScript example above, with the caveat that it won't work if any DEFtype (except DefVar) has been used. (The default data type is Variant, which can be used as a stand-in for any variable type.)

Also, the sub will fail if one arg is a string containing non-numeric data and the other arg is numeric.

Visual Basic .NET

Semantically identical to C#

Sub Swap(Of T)(ByRef a As T, ByRef b As T)
    Dim temp = a
    a = b
    b = temp
End Sub


Dim a = 1, b = 2
Swap(a, b)


Works with: Windows XBasic

XBasic already has a generic swap procedure built in, but a new subroutine can be defined:

PROGRAM  "nswap"
VERSION  "0.0000"


FUNCTION  Entry ()
	a = 1
	b = 2

	PRINT a, b
	GOSUB nswap
	PRINT a, b

SUB nswap
    temp = a: a = b: b = temp


Yabasic already has a generic swap procedure built in.

a = 1
b = 2

print a, b
//----- swap ----
temp = a : a = b : b = temp
print a, b

Batch File

Swap using pass-by-name

@echo off
setlocal enabledelayedexpansion
set a=1
set b=woof
echo %a%
echo %b%
call :swap a b
echo %a%
echo %b%
goto :eof

set temp1=!%1!
set temp2=!%2!
set %1=%temp2%
set %2=%temp1%
goto :eof


beads 1 program 'Generic swap'

	a = [1 2 "Beads" 3 4]
	b = [1 2 "Language" 4 5]

calc main_init
	swap a[4] <=> b[3]
A = [1, 2, "Beads", 3, 4]
B = [1, 2, "Language", 4, 5]
A = [1, 2, "Beads", "Language", 4]
B = [1, 2, 3, 4, 5]

Binary Lambda Calculus

From we get the 25-bit swap function



Two variables can be exchanged with modified assignment:


With no right-hand side, this expands to a one-argument call a‿b ↩ ⌽ a‿b. BQN doesn't make variable names or references first-class, so it wouldn't make sense to make this generic over variable names.





Stack-based swap.


This has a restriction that a and b must be the same size.

void swap(void *va, void *vb, size_t s)
  char t, *a = (char*)va, *b = (char*)vb;
    t = a[s], a[s] = b[s], b[s] = t;
Works with: gcc

If you have gcc, you can write a preprocessor macro with __typeof__.

  • Caution: __typeof__ is a gcc extension, not part of standard C. __typeof__ does not conflict with C89 because the standard allows compilers to add keywords with underscores like __typeof__.
#define Swap(X,Y)  do{ __typeof__ (X) _T = X; X = Y; Y = _T; }while(0)

Usage examples are:

#include <stdio.h>

#define Swap(X,Y)  do{ __typeof__ (X) _T = X; X = Y; Y = _T; }while(0)

struct test
  int a, b, c;

int main()
  struct test t = { 1, 2, 3 };
  struct test h = { 4, 5, 6 };
  double alfa = 0.45, omega = 9.98;
  struct test *pt = &t;
  struct test *th = &h;
  printf("%d %d %d\n", t.a, t.b, t.c );
  Swap(t, h);
  printf("%d %d %d\n", t.a, t.b, t.c );
  printf("%d %d %d\n", h.a, h.b, h.c );
  printf("%lf\n", alfa);
  Swap(alfa, omega);
  printf("%lf\n", alfa);
  printf("%d\n", pt->a);
  Swap(pt, th);
  printf("%d\n", pt->a);

This is tested with GCC with -std=c89 option.


C#: Using a generic method

Works with: C# version 2.0+

C# 2.0 introduced the concept of generics to the language. Generics are outwardly similar to C++ templates, but are implemented quite differently: generics are maintained generically at runtime rather than being substitued with definite types by the compiler. Generics are intended to promote reusable, efficient, type-safe code, and are used widely throughout the .NET framework and 3rd party libraries, especially in collections. C# generics are less flexible than C++ templates, but are more strongly typed and arguably easier to work with.

static void Swap<T>(ref T a, ref T b)
    T temp = a;
    a = b;
    b = temp;


int a = 1;
int b = 2;
Swap(ref a, ref b); // Type parameter is inferred.

C#: Using tuple syntax

Works with: C# version 7.0+

C# 7.0 introduced language support for tuples, which are implemented using the ValueTuple family of structs. The example below creates a tuple with the values of b and a and uses deconstructing assignment to assign the members of the tuple back to the variables.

int a = 1;
int b = 2;
(a, b) = (b, a);


Generic programming in C++ is provided through templates. Templates in C++ are quite powerful: They form a Turing-complete compile-time sub-language. However, that power isn't needed for swap. Note that the C++ standard library already provides a swap function which contains optimized implementations for standard library types; thus it's advisable to use that instead of a self-written variant like the one below.

While the standard allows to separate declaration and definition of templates into different files using the export keyword, most compilers (including the most used ones) don't implement that. Therefore in practice, templates declared in header files also have to be defined there.

The implementation of the swap function template is straightforward:

template<typename T> void swap(T& left, T& right)
  T tmp(left);
  left = right;
  right = tmp;

Note that this function requires that the type T has an accessible copy constructor and assignment operator.

The standard utility 'swap' can be used to swap two values:


It will work with any types.


C++11 adds move constructors which can be more efficient than copy constructors.

template<class T>
void swap(T &lhs, T &rhs){
  T tmp = std::move(lhs);
  lhs = std::move(rhs);
  rhs = std::move(tmp);


Chapel includes a swap operator:

a <=> b

and supports swapping directly via tuples and destructuring:

(a, b) = (b, a)

Both variables must be of the same type. The Fibonnacci implementation contains an example.


(defn swap [pair] (reverse pair))    ; returns a list
(defn swap [[a b]] '(b a))           ; returns a list
(defn swap [[a b]] [b a])            ; returns a vector

The latter two implementations use destructured binding to define local names for the two elements.


CMake has only one data type: the string.

function(swap var1 var2)
  set(_SWAP_TEMPORARY "${${var1}}")
  set(${var1} "${${var2}}" PARENT_SCOPE)
  set(${var2} "${_SWAP_TEMPORARY}" PARENT_SCOPE)
set(x 42)
set(y "string")
swap(x y)
message(STATUS ${x})  # -- string
message(STATUS ${y})  # -- 42

Because of limitations in CMake, there are a few specific situations where swap() will fail to swap the variables.

  1. When _SWAP_TEMPORARY is the name of the second variable:
    set(x 42)
    set(_SWAP_TEMPORARY "string")
    swap(x _SWAP_TEMPORARY)
    message(STATUS ${x})                # -- 42
    message(STATUS ${_SWAP_TEMPORARY})  # -- 42
    Inside swap(), its local variable _SWAP_TEMPORARY shadows the original _SWAP_TEMPORARY from the parent scope, preventing access to the original value.
  2. When value of either variable is "CACHE" or "PARENT_SCOPE":
    string(TOUPPER CACHE x)
    set(y "string")
    swap(x y)  # CMake Error... set given invalid arguments for CACHE mode.
    swap() can never set a variable to "CACHE" or "PARENT_SCOPE", because these are keywords of set() command.


       AUTHOR.  Bill Gunshannon.
       INSTALLATION.  Home.
       DATE-WRITTEN.  16 December 2021.
      ** Program Abstract:
      **   A simple program to demonstrate the SWAP subprogram.
       01  Val1                 PIC X(72).
       01  Val2                 PIC X(72).

          DISPLAY 'Enter a Value: ' WITH NO ADVANCING.
          ACCEPT Val1.
          DISPLAY 'Enter another Value: ' WITH NO ADVANCING.
          ACCEPT Val2.
          DISPLAY ' ' .
          DISPLAY 'First value: ' FUNCTION TRIM(Val1) .
          DISPLAY 'Second value: ' FUNCTION TRIM(Val2) .


          DISPLAY ' '.
          DISPLAY 'After SWAP '.
          DISPLAY ' '.
          DISPLAY 'First value: ' FUNCTION TRIM(Val1).
          DISPLAY 'Second value: ' FUNCTION TRIM(Val2).

           STOP RUN.
       AUTHOR.  Bill Gunshannon.
       INSTALLATION.  Home.
       DATE-WRITTEN.  16 December 2021.
      ** Program Abstract:
      **   SWAP any Alphanumeric value.  Only limit is 72
      **     character size.  But that can be adjusted for
      **     whatever use one needs.



       01  TEMP                  PIC X(72).


       01  Field1                PIC X(72).
       01  Field2                PIC X(72).

               USING BY REFERENCE Field1, BY REFERENCE Field2.

       MOVE Field1 to TEMP.
       MOVE Field2 to Field1.
       MOVE TEMP to Field2.


Enter a Value: 33
Enter another Value: 77
First value: 33
Second value: 77
After SWAP 
First value: 77
Second value: 33


Enter a Value: Hello World
Enter another Value: Good Bye
First value: Hello World
Second value: Good Bye
After SWAP 
First value: Good Bye
Second value: Hello World


This is another standard swap.

<cfset temp = a />
<cfset a = b />
<cfset b = temp />

Common Lisp

(rotatef a b)

(psetq a b b a)

Computer/zero Assembly

LDA 29
STA 31
LDA 30
STA 29
LDA 31
STA 30
org 29
byte $0F
byte $E0
byte $00


Emulator with program loaded Hexdump of address 29-31 after running:

E0 0F 0F


Crystal directly supports swapping:

a, b = b, a


import std.algorithm: swap; // from Phobos standard library

// The D solution uses templates and it's similar to the C++ one:
void mySwap(T)(ref T left, ref T right) {
    auto temp = left;
    left = right;
    right = temp;

void main() {
    import std.stdio;

    int[] a = [10, 20];

    // The std.algorithm standard library module
    // contains a generic swap:
    swap(a[0], a[1]);

    // Using mySwap:
    mySwap(a[0], a[1]);
[10, 20]
[20, 10]
[10, 20]


We use two registers to swap in POSIX dc.

1 2 SaSbLaLb f
=2 1

Reverse (r) is a built-in stack command available as a GNU extension for dc.

1 2 r f
=2 1


symbols do not have to be declared, they can be integers or strings, they can change type on the fly

$ a1 = 123
$ a2 = "hello"
$ show symbol a*
$ gosub swap
$ show symbol a*
$ exit
$ swap:
$  t = a1
$  a1 = a2
$  a2 = t
$ return
$ @generic_swap
  A1 = 123   Hex = 0000007B  Octal = 00000000173
  A2 = "hello"
  A1 = "hello"
  A2 = 123   Hex = 0000007B  Octal = 00000000173


Delphi does not have generics as such. The following code must be copied for each type that a swap is required. T should be changed to the required type.

procedure Swap_T(var a, b: T);
  temp: T;
  temp := a;
  a := b;
  b := temp;

Generics were introduced with Delphi 2009

program GenericSwap;

  TSwap = class
    class procedure Swap<T>(var left, right: T);

class procedure TSwap.Swap<T>(var left, right: T);
  temp : T;
  temp := left;
  left := right;
  right := temp;

  a, b : integer;
  a := 5;
  b := 3;
  writeln('Before swap: a=', a, ' b=', b);
  TSwap.Swap<integer>(a, b);
  writeln('After swap: a=', a, ' b=', b);

Déjà Vu

To swap the two top-most items on the stack:


To swap two variables without needing a third name, using the stack for temporary storage:

set :a set :b @a @b



def swap(&left, &right) {
  def t := left
  left := right
  right := t


def swap([left, right]) {
  return [right, left]


;; 1)
;; a macro will do it, as shown in Racket (same syntax)
(define-syntax-rule (swap a b)
    (let ([tmp a])
    (set! a b)
    (set! b tmp)))

(define A 666)
(define B "simon")
(swap A B)
A  "simon"
B  666

;; 2) 
;; The list-swap! function allows to swap two items inside a list, regardless of their types
;; This physically alters the list

(define L ' ( 1 2 3 4 🎩 ))
(list-swap! L 1 ' 🎩 )
     (🎩 2 3 4 1)


ELENA 4.1 :

import extensions;
swap(ref object v1, ref object v2)
    var tmp := v1;

    v1 := v2;
    v2 := tmp
public program()
    var n := 2;
    var s := "abc";
    console.printLine(n," ",s);
    swap(ref n, ref s);
    console.printLine(n," ",s)
2 abc
abc 2


Elixir provides a robust mechanism of pattern matching; the = operator is actually the match operator. Using the match operator, values can be assigned and variables can be bound or unbound, but only on the left (=:).

x = 4
y = 5

{y,x} = {x,y}
y # => 4
x # => 5

[x,y] = [y,x]
x # => 4
y # => 5

Data structures can be used both for matching and for generally destructuring complex data. We can use anonymous functions to create a generic swap in iex. Note: using multiple value requires a data construct on which to match (as opposed to, say, Ruby's a,b = 1,2), but we can use a list:

swap = fn x,y -> [y|x] end
[x|y] = swap.(1,2)
x # => 2
y # => 1

Variables can be bound and rebound regardless of type

swap_tuple = fn {x,y} -> {y,x} end
{a,b} = swap_tuple.({1,:ok})
a # => :ok
b # => 1

swap_list  = fn [x,y] -> [y,x] end
[a,b] = swap_list.([1,"2"])
a # => "2"
b # => 1

Emacs Lisp

(defun swap (a-sym b-sym)
  "Swap values of the variables given by A-SYM and B-SYM."
  (let ((a-val (symbol-value a-sym)))
    (set a-sym (symbol-value b-sym))
    (set b-sym a-val)))
(swap 'a 'b)

A macro can take variable names unquoted. Here prog1 eliminates the temporary variable above so as to avoid any chance of a name clash between the two variables and the temporary.

(defmacro swap (a b)
  `(setq ,b (prog1 ,a (setq ,a ,b))))

A macro could use the cl-lib psetf which can store to various kinds of expressions as locations, for example list elements. psetf evaluates all its values before storing (like the Common Lisp example).

(require 'cl-lib)
(defmacro swap (a b)
  `(cl-psetf ,a ,b
             ,b ,a))

(setq lst (list 123 456))
(swap (car lst) (cadr lst))
;; now lst is '(456 123)


fun swap = Pair by var a, var b do return var%var(b => a) end
int a = 1
int b = 2
writeLine("before swap: a=" + a + ", b=" + b)
Pair pair = swap(a, b)
a = pair[0]
b = pair[1]
writeLine(" after swap: a=" + a + ", b=" + b)
before swap: a=1, b=2
 after swap: a=2, b=1


Erlang variables are single assignment and Erlang is dynamically typed, so this task doesn't really apply.

The closest thing would be to swap the items in a list (shown in the shell).

1> L = [a, 2].
2> lists:reverse(L).

Or swap the items in a tuple (also shown in the shell).

1> T = {2,a}.
2> list_to_tuple(lists:reverse(tuple_to_list(T))).


include std/console.e -- for display

object x = 3.14159
object y = "Rosettacode"

{y,x} = {x,y}

display("x is now []",{x})
display("y is now []",{y})
x is now Rosettacode
y is now 3.14159

objects are generic, and can accept any type.

More strongly-typed variables can be swapped, if their types match or if the value is "acceptable" to the declared type.

In the above example, if we declared y as an integer, the swap would fail, pi is not an integer. If we declare y as an atom, it would succeed, as atoms can contain any valid number. If we declare y as a sequence or string, it would fail, because x is atomic.


let swap (a,b) = (b,a)


Depending on how you look at it: this task doesn't apply, or it's trivial:



a = 1
b = 2
a,b = arr = b,a

Reading right to left: Assign b & a into an array variable called arr, then assign into a & b


Swap the top two values on the stack:



Since the Forth stack can contain pointers to any data type all we need is...



Works with: Fortran version 90 and later
MODULE Genericswap

    MODULE PROCEDURE Swapint, Swapreal, Swapstring


  SUBROUTINE Swapint(a, b)
    INTEGER :: temp
    temp = a ; a = b ; b = temp

  SUBROUTINE Swapreal(a, b)
    REAL, INTENT(IN OUT) :: a, b
    REAL :: temp
    temp = a ; a = b ; b = temp

  SUBROUTINE Swapstring(a, b)
    CHARACTER(*), INTENT(IN OUT) :: a, b
    CHARACTER(len(a)) :: temp
    temp = a ; a = b ; b = temp
END MODULE Genericswap

  USE Genericswap
  INTEGER :: i1 = 1, i2 = 2
  REAL :: r1 = 1.0, r2 = 2.0
  CHARACTER(3) :: s1="abc", s2="xyz"

  CALL Swap(i1, i2)
  CALL Swap(r1, r2)
  CALL Swap(s1, s2)

  WRITE(*,*) i1, i2   ! Prints 2 and 1
  WRITE(*,*) r1, r2   ! Prints 2.0 and 1.0
  WRITE(*,*) s1, s2   ! Prints xyz and abc

Free Pascal

{$ifdef fpc}{$mode delphi}{$H+}{$endif}
{ note this is compiled with delphi mode but will only compile in Free Pascal }
{ Delphi doesn't support this syntax                                          }
procedure swap<T>(var left,right:T);
  a:string = 'Test';
  b:string = 'me';
    Test    me
    me  Test


The following example will work on all Frink data types:

[b,a] = [a,b]


Fōrmulæ programs are not textual, visualization/edition of programs is done showing/manipulating structures but not text. Moreover, there can be multiple visual representations of the same program. Even though it is possible to have textual representation —i.e. XML, JSON— they are intended for storage and transfer purposes more than visualization and edition.

Programs in Fōrmulæ are created/edited online in its website.

In this page you can see and run the program(s) related to this task and their results. You can also change either the programs or the parameters they are called with, for experimentation, but remember that these programs were created with the main purpose of showing a clear solution of the task, and they generally lack any kind of validation.


Fōrmulæ supports assignments of several symbols (provided as a list) from an expression that reduces to a list of the same cardinality (the expression is first reduced before the actual assignment).

This can be used to do a generic swap as follows:

Test case


1 !0 2 !1

Now tape[0] and tape[1] are set to 1 and 2, respectively.

&0 &1 !0 pop !1

This pushes the value of tape[0] to the stack, tape[1] to the stack, sets tape[0] to the top element, and then pops it, then tape[1] to the top element.


Built in

Not a valid solution, since the task requires writing a function or operator, but it is worth mentioning that Go's built in assignment operator does generic swap. The following swaps the values of a and b as long as they are of identical type.

a, b = b, a

Pass interfaces

A generic swap function can easily be written however, if you require the caller to use variables of the empty interface type. The empty interface can hold a value of any type.

package main

import "fmt"

func swap(a, b *interface{}) {
    *a, *b = *b, *a

func main() {
    var a, b interface{} = 3, "four"
    fmt.Println(a, b)
    swap(&a, &b)
    fmt.Println(a, b)
3 four
four 3

Pass pointers

Somewhat less restrictive, this version allows pointers of any type to be passed, as long as they are the same type.

package main

import (

func swap(a, b interface{}) error {
    ta := reflect.TypeOf(a)
    tb := reflect.TypeOf(b)
    if ta != tb {
        return fmt.Errorf("swap args are different types: %v and %v", ta, tb)
    if ta.Kind() != reflect.Ptr {
        return fmt.Errorf("swap args must be pointers")
    ea := reflect.ValueOf(a).Elem()
    eb := reflect.ValueOf(b).Elem()
    temp := reflect.New(ea.Type()).Elem()
    return nil

func main() {
    a, b := 3, "cats"
    fmt.Println("a b:", a, b)
    err := swap(a, b)
    fmt.Println(err, "\n")

    c, d := 3, 4
    fmt.Println("c d:", c, d)
    err = swap(c, d)
    fmt.Println(err, "\n")

    e, f := 3, 4
    fmt.Println("e f:", e, f)
    swap(&e, &f)
    fmt.Println("e f:", e, f, "\n")

    type mult struct {

    g, h := mult{3, "cats"}, mult{4, "dogs"}
    fmt.Println("g h:", g, h)
    swap(&g, &h)
    fmt.Println("g h:", g, h)
a b: 3 cats
swap args are different types: int and string 

c d: 3 4
swap args must be pointers 

e f: 3 4
e f: 4 3 

g h: {3 cats} {4 dogs}
g h: {4 dogs} {3 cats}


Putting & in a call makes the parameter "call by reference", giving a command the opportunity to modify a variable in the caller. So to swap variables,

`Swap Vars &.a. &.b.'
    new .temp.
    .temp. = \.word2.
    \.word2. = \.word3.
    \.word3. = .temp.
    delete .temp.

.foo. = 123
.bar. = 456
Swap Vars &.foo. &.bar.

show .foo. " " .bar.
# prints "456 123"

Or similar to swap synonyms (strings),

`Swap Syns &\a &\b'
    new \temp
    \temp = "\.word2."
    \.word2. = "\.word3."
    \.word3. = "\temp"
    delete \temp

\quux = "one"
\xyzzy = "two"
Swap Syns &\quux &\xyzzy

show "\quux \xyzzy"
# prints "two one"


Groovy has support for swapping built in:

(a, b) = [b, a]

But the task calls for a "generic swap method" to be written, so here it is:

def swap(a, b) {
    [b, a]

This function doesn't mutate anything, but simply returns a new list with the order of the elements switched. It can be used like shown below:

def (x, y) = swap(1, 3)
assert x == 3
assert y == 1

Some examples here show an in-place swap of indexed elements in an array or collection, so for completeness here is an in-place swap of arbitrary indexed elements in a list:

def listSwap = { a, i, j ->
    assert (0..<(a.size())).containsAll([i,j]);
    a[[j,i]] = a[[i,j]]

def list = [2,4,6,8]
listSwap(list, 1, 3)
assert list == [2,8,6,4]


Harbour has no build-in swap function, however, implemention of a UDF is a fairly easy task. And since it's a dynamic typed lang, there won't be problem while swapping different value types.

Implementation: (We exploit "pass by reference" method on the parameters used)

PROCEDURE Swap( /*@*/v1, /*@*/v2 )
   LOCAL xTmp
   xTmp := v1
   v1 := v2
   v2 := xTmp

Sample code:

   LOCAL v1 := "World!", v2 := "Hello"
   ? v1, v2 // --> World! Hello
   Swap( @v1, @v2 )
   ? v1, v2 // --> Hello World!


Pure swap

Usually Haskellers prefer to work with immutable data. The following function doesn't mutate anything, but simply returns a new pair with the order of the elements switched.

The type signature, the first line, is optional; it may be inferred.

swap :: (a, b) -> (b, a)
swap (x, y) = (y, x)

This swap function is available in the Data.Tuple standard library module in GHC 7.0+

Swap mutable variables

The following function swaps the contents of two mutable references. Again the type signature is optional.

import Control.Monad.Ref
swap :: MonadRef r m => r a -> r a -> m ()
swap xRef yRef = do 
   x<-readRef xRef
   y<-readRef yRef
   writeRef xRef y
   writeRef yRef x

Icon and Unicon

Icon provides a :=: operator for this. Additionally, there is a reversible exchange operator <-> that reverses the exchange if resumed.

procedure main()
   x := 1
   y := 2
   x :=: y
   write(x," ",y)
   # swap that will reverse if surrounding expression fails
   if x <-> y & x < y then write(x, " ", y)


IDL is dynamically typed and array-centric, so swapping is quite easy for any data type. The TEMPORARY function sets its argument to "undefined", and allows us to swap without any large copying.

pro swap, a, b
  c = temporary(a)
  a = temporary(b)
  b = temporary(c)


J is dynamically typed and J's cycle primitive (C.) will swap elements of an arbitrary list. See also J's reference documentation on C.

Shown here are a list of prime numbers and the result of J's parser on some random text (inverting the parsing process on the swapped result):

   (<2 4) C. 2 3 5 7 11 13 17 19
2 3 11 7 5 13 17 19
   (<0 3)&C.&.;:'Roses are red. Violets are blue.'
Violets are red. Roses are blue.

Also, if the argument list can be guaranteed to be a pair, J's reverse primitive will swap the pair.

   |.2 3
3 2
   |.&.;:'one two'
two one

A generic destructive swap of named values would instead require reference to the locations being destroyed. Here's an implementation of that:

destructiveSwap=: {{ EMPTY[ (m;n)=: n ,&<&do m }}

Example use:

   V1=: 'cat'
   V2=: 7
   'V1' destructiveSwap 'V2'


fn swap<T>(anon a: &mut T, anon b: &mut T) {
    let temporary = *a
    *a = *b
    *b = temporary

fn main() {
    mut a = "Hello"
    mut b = "World"

    println("{} {}", a, b)
    swap(&mut a, &mut b)
    println("{} {}", a, b)

    mut c = 1
    mut d = 2

    println("{} {}", c, d)
    swap(&mut c, &mut d)
    println("{} {}", c, d)


Works with: Java version 1.5+

Java uses references, so it can't swap the values of two variables that don't belong to a class.

class Pair<T> {
    T first;
    T second;
public static <T> void swap(Pair<T> p) {
   T temp = p.first;
   p.first = p.second;
   p.second = temp;


JavaScript uses references, but if a function reassigns a parametric reference, the new object only has a local reference. However, if we wrap the variables to be switched in some other structure, like an object or an array, we can easily swap the values.

There's no actual "generics", since all variables are just that, variables of some kind.

The below function expects an array of length 2 (or longer), and switches the first two values in place, in the same array. This is closely related to how the Java solution works.

function swap(arr) {
  var tmp = arr[0];
  arr[0] = arr[1];
  arr[1] = tmp;

Also there is metaprogramming solution. It uses code generation and eval. To avoid naming conflicts(user can pass 'tmp', which causes var tmp = tmp) it uses buildin, per activation context (thats why it is enclosed into self executing lambda), var arguments for temp storage.

function swap(aName, bName) {
  eval('(function(){ arguments[0] = aName; aName = bName; bName = arguments[0] })()'
    .replace(/aName/g, aName)
    .replace(/bName/g, bName)
var x = 1
var y = 2
swap('x', 'y')

Solution without eval(), assuming that the code is running in the browser (window is the global object)

function swap(a, b) {
  var tmp = window[a];
  window[a] = window[b];
  window[b] = tmp;
var x = 1;
var y = 2;
swap('x', 'y');

Another solution for swapping array items using destructing assignment:

const arr = [1, 2, 3, 4, 5];
[arr[0], arr[1]] = [arr[1], arr[0]]


Provided that the stack contains at least two elements:


changes the order of those elements.


jq is a functional language, so one is more likely to want to swap the two elements of an array than to swap the values of two variables, but jq does have variables and their values can be swapped, for example, using an intermediate variable, say $tmp, as illustrated here:

jq -n '1 as $a | 2 as $b | $a as $tmp | $b as $a | $tmp as $b | [$a,$b]'

Here is a filter that will swap the elements of a two-element array:


And here is a filter that, if presented with an array, will in effect copy it and then swap the i-th and j-th items, it being understood that if a is an array and k < 0 or k >= (a|length), then a[k] will evaluate to null:

def swap(i;j): .[i] as $t | .[i] = .[j] | .[j] = $t;


Similar to Python, Julia has built-in support for swapping:

a, b = b, a


As Kotlin does not support passing parameters by reference and tuples cannot be destructured automatically to pre-existing variables, it's just as easy to swap variable values 'inline' rather than using a function. However, here's one of way of doing it generically using the latter:

fun <T> swap(t1: T, t2: T) = Pair(t2, t1)

fun main() {
    var a = 3
    var b = 4
    val c = swap(a, b) // infers that swap<Int> be used
    a = c.first
    b = c.second
    println("a = $a")
    println("b = $b")
    var d = false
    var e = true
    val f = swap(d, e) // infers that swap<Boolean> be used
    d = f.first
    e = f.second
    println("d = $d")
    println("e = $e")
a = 4
b = 3
d = true
e = false

You can also explicitly create a container class and swap the value that it contains (but this is a bad idea in practice):

data class Ref<T>(var value: T) {
    fun swap(other: Ref<T>) {
        val tmp = this.value
        this.value = other.value
        other.value = tmp
    override fun toString() = "$value"

fun main() {
    val a = Ref(1)
    val b = Ref(2)


1) using an Immediately Invoked Function Expression:
{{lambda {:x :y} :y :x} hello world}
-> world hello

2) or user defined function
{def swap {lambda {:x :y} :y :x}}
-> swap

3) applied on words (which can be numbers)
{swap hello world}
-> world hello

{swap hello brave new world}
-> brave new hello world

{swap {cons hello brave} {cons new world}}
-> (new world) (hello brave)


# Swap function with call-by-pointer
fp.swap = ($[aPtr], $[bPtr]) -> {
	$tmp = $*aPtr
	$*aPtr = $*bPtr
	$*bPtr = $tmp

$a = 42
$b = A short text

fn.println($a, $b)

fp.swap($a, $b)

fn.println($a, $b)
42, A short text
A short text, 42


swap        # stack
reverse     # array


Values can be swapped using multi-variable assignment.

var .abc = [1, 2, 3]
var .def = [5, 6, 7]

.abc[3], .def[3] = .def[3], .abc[3]

writeln .abc
writeln .def
[1, 2, 7]
[5, 6, 3]


define swap(a, b) => (: #b, #a)

local(a) = 'foo'
local(b) = 42

local(a,b) = swap(#a, #b)

Using Decompositional Assignment

local(a)   = 'hair'
local(b)   = 'moose'
local(a,b) = (: #b, #a)


Lhogho is very similar except that it does not have a localmake opcode.

to swap :s1 :s2
  local "t
  make "t thing :s1
  make :s1 thing :s2
  make :s2 :t

make "a 4
make "b "dog
swap "a "b        ; pass the names of the variables to swap
show list :a :b  ; [dog 4]


A generic swap function is not possible in Lingo, since scalar values are passed by value. But the following solution shows how such generic swapping still can be achieved by executing a single line of code:

on swap (x, y)
  return "tmp="&x&RETURN&x&"="&y&RETURN&y&"=tmp"


x = 1
y = 2
put x, y
-- 2 1


(a, b) := (b, a);


put "first" into a1
put "last" into b2
swap a1,b2
put a1 && b2

command swap @p1, @p2
    put p2 into p3
    put p1 into p2
    put p3 into p1
end swap

to swap :s1 :s2
  localmake "t thing :s1
  make :s1 thing :s2
  make :s2 :t

make "a 4
make "b "dog
swap "a "b        ; pass the names of the variables to swap
show list :a :b  ; [dog 4]


:- object(paws).

    :- public(swap/4).
    swap(First, Second, Second, First).

:- end_object.

Usage examples:

| ?- paws::swap(apples, oranges, X, Y).
X = oranges
Y = apples

| ?- paws::swap(3.14, ext(lgt), X, Y).
X = ext(lgt)
Y = 3.14



LOLCODE's dynamic typing makes generic swapping trivial. In addition, the special IT variable‒which contains the most recently evaluated expression‒permits doing so without explicitly creating a temporary variable.

HAI 1.3

I HAS A foo ITZ "kittehz"
I HAS A bar ITZ 42

foo, foo R bar, bar R IT

VISIBLE bar BTW, kittehz



Lua evaluates the values on the right-hand side before assigning them to the variables on the left-hand side. This behaviour allows the following notation to be used to swap two values:

x, y = y, x                -- swap the values inside x and y
t[1], t[2] = t[2], t[1]    -- swap the first and second values inside table t

Usage example:

x, y = 3, 4
print(x, y)                --> 3 4
x, y = y, x                -- swap
print(x, y)                --> 4 3

M2000 Interpreter

Swap is a statement in M2000 which get two identifiers, variables or array items. Variables and Array items are all internal type of Variant. Normally a numeric variable hold the first type we assign to it. Numeric types are: Double, Single, Decimal, Currency, Decimal, Long, Integer. Boolean is also a type but true and false are not boolean, they are double -1 and 0). When we use Swap internal only variant swap happen, without check of type of variant.

Here we make a local Swap and pass by reference, numbers and strings. References created without testing what type of variant we use. So calling swap we make a swap moving bytes, and for strings this means moving pointers to BSTR type of strings.

\\ programming again Swap (for local use)
Module Swap (&a, &b) {
      \\ this call internal command - by default is by reference without using character &
      Swap a, b
Swap &x, &y
Print X, Y, Type$(X)="Double",Type$(Y)="Double"
Swap &A$, &B$
Print A$="B$", B$="A$"

Using Swap (internal command), for variables, groups (only for variables inside groups), pointers to groups, pointers to containers, etc.

Swap a,b
Print a, b
Swap A$, B$
Print A$, B$
Dim A(4)
Swap  A(3), A(2)
Print A(3), A(2)

\\ Groups are Values
Group alfa {
      x=10, y=20
Group Beta {
      x=40, y=50
\\ with List we show the public variables
\\ so among other variables there are:
\\ alfa[Group], alfa.x=10, alfa.y=20, beta[group], beta.x=40, beta.y=50
\\ So Alfa.x and Beta.x are simple variables, we can use swap
Swap Alfa.x, Beta.x
Print Alfa.x, Beta.x
Swap Alfa.x, Beta.x
\\ We have to use a third variable to hold value
For This {
      \\ Local always make a new variable, and shadow any same local variable
      Local M=alfa
\\ Now M erased (defined in For This block)
Print Alfa.x=40, Alfa.y=50
Print Beta.x=10, Beta.y=20

\\ Using -> we make pointers to Alfa, and Beta
\\ These pointers are valid until Alfa and Beta erased, or get Empty Group (->0)
Print pA=>x=40, pA=>y=50
Print pB=>x=10, pB=>y=20
Swap pA,pB
Print pA=>x=10, pA=>y=20   ' pA point to beta
Print pB=>x=40, pB=>y=50   'pB point to alfa
Print type$(pA)="Group",Valid(pA=>X)=True
Print type$(pA)="Group",Valid(pA=>X)=False
\\ These pointers are valid until get Empty Group (->0), they point to a copy of Alfa  and Beta
\\ both are in heap as "closed groups"
Print pA=>x, pA=>y
\\ swap need variables or arrays
\\ pA=>x are closed to object so we have to open the object, and use the open one, where all public variables of group can be used
For pA, pB {
      Swap .x, .y
Print pA=>x, pA=>y
For pA, pB {
      Swap .x, .y
Print pA=>x, pA=>y
Print pB=>x, pB=>y
Swap pA,pB
Print pA=>x, pA=>y
Print pB=>x, pB=>y

L1=lambda x=1->{=x : x++}
L2=lambda x=100->{=x : x--}
Print L1()=1, L2()=100
Swap L1, L2
Print L1()=99, L2()=2
Swap L1, L2
Print L1()=3, L2()=98
\\ swap change pointers to containers (here pointers to arrays)
Swap A, B
Print A
Print B
\\ Arrays with () in names are values
Dim A(10)=1, B(10)=2
For This {
      Dim C()
Print A()
Print B()


define(`def2', `define(`$1',`$2')define(`$3',`$4')')dnl
define(`swap', `def2(`$1',defn(`$2'),`$2',defn(`$1'))')dnl
a b
a b
x y

y x


The assignment operator in Maple can swap values, since the right hand side is evaluated before the assignment occurs.

> a, b := 2, "foo":
> a;

> b;

> a, b := b, a: # SWAP
> a;           

> b;           

Mathematica / Wolfram Language

Mathematica functions are generic by default; however, it has to be told not to evaluate the arguments before executing the function.

swap[a_, b_] := {a, b} = {b, a}
SetAttributes[swap, HoldAll]

MATLAB / Octave

Numercial swaps are trivial operations. In fact, they are so natural to the language that multiple swaps can be performed simultaneously.


>> a = [30 40 50 60 70]

a =

    30    40    50    60    70

>> a([1 3]) = a([3 1]) %Single swap

a =

    50    40    30    60    70

>> a([1 2 4 3]) = a([2 3 1 4]) %Multiple swap, a.k.a permutation.

a =

    40    30    60    50    70

A generic swap (compatible with any variable type) can be performed with the deal command:

>> a = 12
a =  12

>> b = 'foo'
b = foo

>> [b, a] = deal (a, b)
b =  12
a = foo


a: 10$
b: foo$

/* A simple way to swap values */
[a, b]: [b, a]$

a; /* foo */
b; /* 10 */

/* A macro to hide this */
swap(x, y) ::= buildq([x, y], ([x, y]: [y, x], 'done))$

swap(a, b)$

a; /* 10 */
b; /* foo */


swap a b


In Metafont, only numeric declarations can be omitted; any other type, must be explicitly given. So our swap, in order to declare and use a proper temporary variable(? in this code), must check the type of the variable passed (we check only for a; if b is of another kind, an error will occur)

vardef swap(suffix a, b) =
  save ?; string s_;
  if boolean a: boolean ?
    elseif numeric a: numeric ? % this one could be omitted
    elseif pair a: pair ?
    elseif path a: path ?
    elseif pen a: pen ?
    elseif picture a: picture ?
    elseif string a: string ?
    elseif transform a: transform ? fi;
  ? := a; a := b; b := ?


j := 10;
i := 5;
show j, i;
show j, i;

boolean truth[];
truth1 := true;
truth2 := false;
show truth1, truth2;
show truth1, truth2;


Works with: min version 0.19.3

Like many other stack languages, this is trivial.



Like many other languages, MiniScript passes references by value, so a straightforward swap is impossible. However, there is a trick: given the map the variales are in (e.g. locals) and the names of the variables, we can swap them.

swap = function(map, a, b)
    temp = map[a]
    map[a] = map[b]
    map[b] = temp
end function

x = 1
y = 2
print "BEFORE: x=" + x + ", y=" + y
swap(locals, "x", "y")
print "AFTER:  x=" + x + ", y=" + y
BEFORE: x=1, y=2
AFTER:  x=2, y=1



PROCEDURE Swap(VAR left: Elem.T; VAR right: Elem.T);

END GenericSwap.
GENERIC MODULE GenericSwap(Elem);

PROCEDURE Swap(VAR left: Elem.T; VAR right: Elem.T) =
  VAR temp: Elem.T := left;
    left := right;
    right := temp;
  END Swap;

END GenericSwap.

Here is an example usage for integers:

INTERFACE IntSwap = GenericSwap(Integer) END IntSwap.
MODULE IntSwap = GenericSwap(Integer) END IntSwap.

IMPORT IntSwap, IO, Fmt;

VAR left := 10;
    right := 20;

  IO.Put("Left = " & Fmt.Int(left) & "\n");
  IntSwap.Swap(left, right);
  IO.Put("Left = " & Fmt.Int(left) & "\n");
END Main.
Left = 10
Left = 20


For pairs, namespace Nemerle.Utility.Pair contains Swap():

def coords    = (1, -1);
def invcoords = Swap(coords);

Or to swap two mutable variables of the same type:

a <-> b;

But, enough about built in functionality, let's demonstrate using generics:

Swap[T, U] (a : T, b : U) : U * T
    (b, a)


Values stored in the default Rexx data type are treated as typeless data; context is based on the contents. Swapping the contents of variables stored in Rexx object can be achieved via the PARSE instruction.

/* NetRexx */
options replace format comments java crossref symbols nobinary

  -- Simple values with no spaces can be swapped without the use of a parse template
  lval = 27
  rval = 5
  say 'Before - <lval>'lval'</lval> <rval>'rval'</rval>'
  parse (lval rval) rval lval
  say 'After  - <lval>'lval'</lval> <rval>'rval'</rval>'

  -- More complex data needs to use some form of parsing template
  lval = 'This value started on the left'
  rval = 'This value started on the right'
  dlm  = 12x80facebead01 -- some delimiting value that is unlikely to occur in the LVAL to be swapped
  say 'Before - <lval>'lval'</lval> <rval>'rval'</rval>'
  parse (lval || dlm || rval) rval (dlm) lval
  say 'After  - <lval>'lval'</lval> <rval>'rval'</rval>'

Before - <lval>27</lval> <rval>5</rval>
After  - <lval>5</lval> <rval>27</rval>

Before - <lval>This value started on the left</lval> <rval>This value started on the right</rval>
After  - <lval>This value started on the right</lval> <rval>This value started on the left </rval>


(swap a b)


Like J

|reverse 1 2
=2 1


Builtin procedure swap. Example usage:

swap(a, b)

OASYS Assembler

You can swap variable %A# with %B# by writing:


A method which can be called to implement it can be written like:


To call such method:



Tuples are immutable in OCaml. This function doesn't mutate anything, but simply returns a new pair with the order of the elements switched.

let swap (x, y) = (y, x)

If the arguments are constrained to be reference values, a swap function is simple:

let swapref x y =
  let temp = !x in
    x := !y;
    y := temp




Oz variables are dataflow variables and cannot be changed once a value has been assigned. So a swap operation on dataflow variables does not make sense.

We can write a swap procedure for cells, though. Cells are mutable references.

  proc {SwapCells A B}
     Tmp = @A
     A := @B
     B := Tmp

Or shorter, if we exploit the fact that the assignment operator := returns the old value of the cells:

  proc {SwapCells A B}
     B := A := @B

A functional swap, operating on pairs:

  fun {SwapPair A#B}


Pari is near-typeless—everything is a GEN.

Works with: PARI/GP version 2.6.0 and above


Works with: Free_Pascal version 2.6.0

Standard Pascal does not have generics, but FreePascal has a start:

program generictest;

{$mode objfpc}

  generic TSwap<T> = procedure (var a, b: T);

procedure Proc1(var a, b: integer);
    temp: integer;
    temp := a;
    a := b;
    b := temp;

  S, T: integer;
  SwapInt: specialize TSwap<integer>; 

  S := 4;
  T := 3;
  SwapInt := @Proc1;
  writeln(S, T:2);
  SwapInt(S, T);
  writeln(S, T:2);
4 3
3 4

since FreePascal version 3.2.0:

program generic_test;
{$mode objfpc}{H+}

generic procedure GSwap<T>(var L, R: T);
  Tmp: T;
  Tmp := L;
  L := R;
  R := Tmp;
  I, J: Integer; 
  I := 100;
  J := 11;
  WriteLn('I = ',  I, ', J = ', J);
  specialize GSwap<Integer>(I, J);
  WriteLn('I = ',  I, ', J = ', J);
I = 100, J = 11
I = 11, J = 100


Perl has support for swapping built-in

($y, $x) = ($x, $y);

Here's a generic swap routine:

sub swap {@_[0, 1] = @_[1, 0]}


Library: Phix/basics

The following applies to any types. Subscripting and nesting may also be used freely on either side.

{a,b} = {b,a}


Applies to any types.

a b var a var b


function swap(&$a, &$b) {
    list($a, $b) = array($b, $a);


Picat supports re-assignment (:=), but it is not possible to swap two variables like this:

A = 1,
B = 2,
[A,B] := [B,A],
% ...

(This throws an error: "invalid left-hand side of assignment").

The following works, but it's probably not general enough for this task:

A = 1, 
B = 2,
T = A, % Assuming that T is not used elsewhere in the clause
A := B,
B := T,
% ...

Instead swaps has to be done by some other means.

Inline swapping in a list

One way is to place the values in a list (or array) and use inline swapping of the list.

% Swapping positions in a list/array
swap2(L) =>

% Swap two elements in a list
swap_list(L,I,J) =>
  T = L[I],
  L[I] := L[J],
  L[J] := T.


L = [1,2],
% ...

L is now [2,1]

Note: Placing the variables A, B in swap2


will NOT change the values of A and B.

Using a second list

Another way is to place the swapped values in a second list:



A = 1,
B = 2,
% ...

X is now [2,1]. A is still 1 and B is still 2.


xchg works with any data type

(let (A 1  B 2)
   (xchg 'A 'B)
   (println A B) )

(let (Lst1 '(a b c)  Lst2 '(d e f))
   (xchg (cdr Lst1) (cdr Lst2))
   (println Lst1 Lst2) )
2 1
(a e c) (d b f)


mixed first = 5;
mixed second = "foo";
array pair = ({ 5, "foo" });

void swapvars(string a, string b)
    [this[a], this[b]] = ({ this[b], this[a] });

void swaparray(array swapit)
    [swapit[1], swapit[0]] = ({ swapit[0], swapit[1] });

void main()
    write("swap variables:\n");
    write("%O, %O\n", first, second);
    // we could just use [first, second] = ({ second, first });
    swapvars("first", "second");
    write("%O, %O\n", first, second);

    write("swap array:\n");
    write("%{ %O %}\n", pair);
    // we could just use [pair[1], pair[0]] = ({ pair[0], pair[1] });
    // directly, but since arrays are called by reference,
    // it also works through a function
    write("%{%O %}\n", pair);
swap variables:
5, "foo"
"foo", 5
swap array:
 5  "foo" 
"foo" 5 


Using the preprocessor

%swap: procedure (a, b);
   declare (a, b) character;
   return ( 't=' || a || ';' || a || '=' || b || ';' || b '=t;' );
%end swap;
%activate swap;

The statement:-

  swap (p, q);

is replaced, at compile time, by the three statements as in-line code:

  t = p; p = q; q = t;

Using generic procedures

declare swap generic (
   swapf when (float, float),
   swapc when (char, char));

swapf: proc (a, b);
   declare (a, b, t) float;
   t = a; a = b; b = t;
end swapf;
swapc: proc (a, b);
   declare (a, b) character(*);
   declare t character (length(b));
   t = a; a = b; b = t;
end swapc;

declare (r, s) character (5);
call swap (r, s);

Both of the above are not completely generic, but depend on either the presence of

  • a temporary variable with the same attributes of the variables to be swapped, OR
  • data-attribute specific procedures for the swap

The following code is completely generic, but, in line with the usual safety offered by PL/I, swaps only the contents up to the storage occupied by the smallest of the two variables: Prino 01:24, 11 February 2011 (UTC)

Completely generic code using the pre-processor

%swap: proc(x,y);
dcl (x, y) char;

x = trim(x); /* Just for neatness sake */
y = trim(y);

ans('begin;                                                ') skip;
ans('  dcl c  char       (1);                              ') skip;
ans('  dcl sx char       (1) based(px);                    ') skip;
ans('  dcl sy char       (1) based(py);                    ') skip;
ans('  dcl i  fixed bin (31);                              ') skip;
ans('  dcl px ptr            init (addr(' || x || '));     ') skip;
ans('  dcl py ptr            init (addr(' || y || '));     ') skip;
ans('  do i = 1 to min(stg(' || x || '), stg(' || y || '));') skip;
ans('    c  = sx;                                          ') skip;
ans('    sx = sy;                                          ') skip;
ans('    sy = c;                                           ') skip;
ans('    px = px + 1;                                      ') skip;
ans('    py = py + 1;                                      ') skip;
ans('  end;                                                ') skip;
ans('end;                                                  ') skip;
%end swap;
%act swap;

dcl c1 char (10) init ('1234567890');
dcl c2 char (10) init ('ABCDEFGHIJ');
dcl f1 fixed bin (31) init (12345);
dcl f2 fixed bin (31) init (98765);

put data(c1, c2, f1, f2);
swap(c1, c2);
swap(f1, f2);
put data(c1, c2, f1, f2);
f1 = -656877352; /* '5a5a5a5a'x, aka 'QQQQ' */
swapper(c1, f1);
put data(c1,f1);

The code generated by 'swap(c1, c2);' looks like

  dcl c  char       (1);
  dcl sx char       (1) based(px);
  dcl sy char       (1) based(py);
  dcl i  fixed bin (31);
  dcl px ptr            init (addr(C1));
  dcl py ptr            init (addr(C2));
  do i = 1 to min(stg(C1), stg(C2));
    c  = sx;
    sx = sy;
    sy = c;
    px = px + 1;
    py = py + 1;

and, because declarations in PL/I begin blocks are local to that block, generating several blocks with the same variables will not cause any problems.

The result of compiling, linking and executing the above code:

F1=         12345
F2=         98765;
F1=         98765
F2=         12345;
F1=   -1044200508;

Or, using "Like"

The key problem is that a temporary storage area is needed (there alas being no compiler-recognised "swap" statement), and the waystation variable must have the correct type, nor can there be reliance on a suitable "t" variable being available for use. Devising a different Swap for each type of parameter would be tedious and tiresome to use, however one could employ the "generic" facility as above demonstrated, and use the pre-processor to generate a collection of Swap routines by it employing a template and stepping through a list of accommodated types.

Instead, the first example can be generalised via two steps. Firstly, it is possible to declare a variable to be of a type "like" some named variable (otherwise a third parameter naming the type could be supplied), and secondly, placing the in-line code between Begin ... End; means that any declaration is local to within that block only. Further, this bracketing allows a Swap to be invoked via an if-statement, as in If ... then Swap(x,y); - otherwise there would be a mess. Thus:

   declare (a,b) character; /*These are proper strings of arbitrary length, pre-processor only.*/
   return ('Begin; declare t like '|| a ||'; t='|| a ||';'|| a ||'='|| b ||';'|| b ||'=t; End;');
%End Swap;

Whereupon a Swap(this,that); would generate a rather longer text of in-line source code. This and other text expansions caused odd difficulties, because the 1980s compiler replaced the invocation by the expansion and then reformatted the result into lines of 71 characters (not 72) as necessary, and then, since any additional lines were given the same source sequence number as the original line, added 100000 as needed to generate strictly increasing sequence numbers. If many lines overflowed, eventually the sequence field (eight digits) overflowed, and all following source lines thereby acquired the same source sequence number. For this and other reasons, one approach was two-stage compilation: the output from the pre-processor stage could be saved and further compilation cancelled. That file could then be resequenced and fed to the pl/i compiler afresh.

Such a file would have the expansion of Swap(this,that) as follows (but with added layout here):

 declare t like this;
  t = this;
  this = that;
  that = t;

There would however be trouble if the type of this differed from the type of that, and a pl/i compiler may not generate a warning because it handles many type conversions in an assignment without complaint. There are no pre-processor enquiry functions to inspect the types at pre-processor time - if there were, a more accomplished Swap procedure could produce suitable error reports, which can be classed as "warning" or "severe", etc. The "storage" function produces its result at run time, but, each invocation of Swap being compiled would have its actual parameters known as the compiler dealt with the code produced by that invocation of Swap, and so for each invocation, the results of "storage" would be constants - except for items that were allocated at run time.

The bracketing could be via DO; ... END; instead of BEGIN; ... END; but in that case the declared temporary variable would be visible outside its fragment and there could be conflicts, either of differing type for the same name or of multiple declaration. This could be solved by adjusting Swap to generate a different name each time. One could try a prefix (or suffix) to the name of the first parameter (thus generating say SwapTemp_this or similar), but there would still be difficulty if there were multiple swaps involving the same variable. Instead, Swap could count its invocations and generate a name involving that. Temporary variables would then litter the storage area, and they could consume a lot of space. On the other hand, the BEGIN; ... END; arrangement, though typically involving temporary space on the data stack, could have its own constraints. In the 1980s, the IBM mainframe pl/i compiler had a limit of no more than 240 (or so) BEGIN; ... END; blocks, plus procedure blocks, plus a few other items, in any one compilation otherwise there would be a failure "in phase PI". Separate compilation and the linking of pieces introduced its own oddities, as when pieces had been compiled with different compiler options.

Plain English

Plain English includes a routine for swapping two values in its noodle.

Swap [a value] with [another value].


Swap is easily done via multiple assignment:

(a, b) -> (b, a);

Pop11 is dynamically typed, so the code above is "generic".


Works with anything you can put on the operand stack:



PowerShell allows swapping directly, through tuple assignment:

$b, $a = $a, $b

But one can also define a function which swaps the values of two references:

function swap ([ref] $a, [ref] $b) {
    $a.Value, $b.Value = $b.Value, $a.Value

When using this function the arguments have to be explicitly given as references:

swap ([ref] $a) ([ref] $b)



?- swap(1,2,X,Y).
X = 2,
Y = 1.


Python has support for swapping built in:

a, b = b, a

But the task calls for a "generic swap method" to be written, so here it is:

def swap(a, b):
    return b, a

Note that tuples are immutable in Python. This function doesn't mutate anything, but simply returns a new pair with the order of the elements switched.


Quackery objects reside on the stack while in use. The word swap exchanges the top and second item on the stack. The closest thing Quackery has to variables are ancillary stacks. The word exchange defined here exchanges the top items on two ancillary stacks.

[ over take over take 
  2swap dip put put ] is exchange ( s s --> )


R function arguments are passed by value, not by reference. You can work around this, however, by using their names and environment:

swap <- function(name1, name2, envir = parent.env(environment()))
    temp <- get(name1, pos = envir)
    assign(name1, get(name2, pos = envir), pos = envir)
    assign(name2, temp, pos = envir)


> x <- 1
> y <- 2
> swap('x', 'y')
> cat(x, y)
2 1


A swap operation can be easily written as a macro in Racket. The macro will even work as expected in Typed Racket.

#lang racket/load

(module swap racket
  (provide swap)

  ;; a simple macro to swap two variables
  (define-syntax-rule (swap a b)
    (let ([tmp a])
      (set! a b)
      (set! b tmp))))

;; works fine in a statically typed setting
(module typed typed/racket
  (require 'swap)

  (: x Integer)
  (define x 3)

  (: y Integer)
  (define y 4)

  (swap x y)
  (printf "x is ~a~n" x)
  (printf "y is ~a~n" y))


(formerly Perl 6)

Similar to Perl 5. Raku supports type constraints for variables and subroutines, unlike Perl 5, but the default is still to permit all values.

Alternatively, you can write it like this:

($x, $y) .= reverse;


	Title: "Generic Swap"
	Reference: []

swap: func [
	"Swap contents of variables."
	a [word!] b [word!] /local x
	x: get a  
	set a get b  
	set b x

answer: 42  ship: "Heart of Gold"
swap 'answer 'ship ; Note quoted variables.
print rejoin ["The answer is " answer ", the ship is " ship "."]
The answer is Heart of Gold, the ship is 42.




REXX has no primitive for swapping, but it can easily be performed using a temporary variable.
(This is the slowest of the three versions.)

using temp

a = 'I see you.'
b = -6

_temp_ = a                           /*swap ···     */   
     a = b                           /*     A ···   */
     b = _temp_                      /*  and  B     */

using VALUE

This version will work with any values.

a = "bull feathers"
b = 10

a= value('b', a)                     /*swap A and B */

using PARSE

If it's known that there are

  • no blanks
  • no null values
  • (maybe) no whitespace (such as tabs)

in the values, the following method can be used:
(This is the fastest of the three versions.)

a = -199e-12
b = 12.

parse value  a  b    with    b  a    /*swap A and B */

Note that some REXX interpreters handle whitespace differently, some honor whitespace other than blanks,
others don't   (particularly the older versions).


a = 1
b = 2
temp = a
a = b
b = temp
see "a = " + a + nl
see "b = " + b + nl


RLaB does not have a built-in function for swapping the content of two variables. However, there is a workaround which comes from the fact that the global variable space $$ contains all the variables var1, var2 and so forth as $$.var1, ...

Let we want to swap the content of two variables, which names are a and b, then the following function would do the trick

swap = function(x,y)
  if (!exist($$.[x]))
  { return 0; }
  if (!exist($$.[y]))
  { return 0; }
  local (t);
  t = $$.[x];
  $$.[x] = $$.[y];
  $$.[y] = t;
  return 1;

>>  a=1
>>  b = "fish"
>> swap( "a" , "b" );
>>  a
>>  b


Works with: Halcyon Calc version 4.2.7
Code Comments

≪ OVER TYPE 6 == OVER TYPE 6 == 2 * + 1 + 
  { ≪ SWAP ≫ 

":#" means "stack level #"
 calculate case index upon types of data at :1 and :2
 case 1: swap :2 with :1
 case 2: swap content of variable named at :2 with :1
 case 3: swap content of variable named at :1 with :2
 case 4: swap content of variables named at :1 and :2
 select program from case index and execute it


Ruby has support for swapping built in:

a, b = b, a

But the task calls for a "generic swap method", so here it is:

def swap(a, b)
    return b, a

This method does not swap the original variables, because Ruby passes parameters by value. Instead, this method returns simply a new array with the order of the elements switched. The caller may assign the original variables with the return value:

x = 42
y = "string"
x, y = swap x, y
puts x  # prints string
puts y  # prints 42


Rust does not allow for swapping the value of two variables with different types, but if the types are the same it can be done using generic types and lifetimes.

fn generic_swap<'a, T>(var1: &'a mut T, var2: &'a mut T) {
    std::mem::swap(var1, var2)

This function can be used in e.g. the following ways:

fn main() {
    let mut a: String = "Alice".to_owned();
    let mut b: String = "Bob".to_owned();
    let mut c: i32 = 1;
    let mut d: i32 = 2;

    generic_swap(&mut a, &mut b);
    generic_swap(&mut c, &mut d);

    println!("a={}, b={}", a, b);
    println!("c={}, d={}", c, d);
a=Bob, b=Alice
c=2, d=1


A possible way that needs the type of the objects to be specified:

class SWAP{T} is
  swap(inout a, inout b:T) is
    t ::= a;
    a := b;
    b := t;
class MAIN is
  main is
    x ::= 10;
    y ::= 20;
    SWAP{INT}::swap(inout x, inout y);
    #OUT + x + ", " + y + "\n";


Scala has type parameters and abstract types (not to be confused with abstract data types). The swap example is about as simple as such things can be, with no variance or high-order type parameters.

The return type need not be declared in the example below, but it is shown for clarity. However, as Scala does not pass parameters by reference, it cannot swap values in-place. To make up for that, it receives two values, and returns a tuple with the values inverted.

def swap[A,B](a: A, b: B): (B, A) = (b, a)


; swap elements of a vector
; vector-swap! is not part of r5rs, so we define it
(define (vector-swap! v i j)
(let ((a (vector-ref v i)) (b (vector-ref v j)))
(vector-set! v i b)
(vector-set! v j a)))

(let ((vec (vector 1 2 3 4 5)))
  (vector-swap! vec 0 4)
; #(5 2 3 4 1)

; we can swap also in lists
(define (list-swap! v i j)
(let* ((x (list-tail v i))
       (y (list-tail v j))
       (a (car x))
       (b (car y)))
(set-car! x b)
(set-car! y a)))

(let ((lis (list 1 2 3 4 5)))
   (list-swap! lis 0 4)
; (5 2 3 4 1)

; using macros (will work on variables, not on vectors or lists)
(define-syntax swap!
(syntax-rules ()
((_ a b)
   (let ((tmp a))
   (set! a b)
   (set! b tmp)))))

; try it
(let ((a 1) (b 2)) (swap! a b) (list a b))
; (2 1)


There are no variables in sed. Just a "pattern space" and a "hold space" — that's all. The x command exchanges the contents of both.


A generic template to generate swap functions is defined with:

const proc: generate_swap (in type: aType) is func

    const proc: swap (inout aType: left, inout aType: right) is func
        var aType: temp is aType.value;
        temp := left;
        left := right;
        right := temp;
      end func;

  end func;

An instance of a swap function can be generated with:


A swap function can be called with:

swap(a, b);


set [x,y] to [13,"Hello"] -- assign values to two variables
put x,y
set [x,y] to [y,x] -- swap the variable values
put x,y





func swap(Ref a, Ref b) {
    var tmp = *a;
    *a = *b;
    *b = tmp;


func swap(Ref a, Ref b) {
    (*a, *b) = (*b, *a);


func swap(Ref a, Ref b) {
    [*a, *b] » (b, a);

The swap functions must be called with variable references.

var (x, y) = (1, 2);
swap(\x, \y);


This must be done with a macro method in Slate, but is in the standard library:

x@(Syntax LoadVariable traits) swapWith: y@(Syntax LoadVariable traits) &environment: env
"A macro that expands into simple code swapping the values of two variables
in the current scope."
  env ifNil: [error: 'Cannot swap variables outside of a method'].
  tmpVar ::= env addVariable.
  {tmpVar store: x variable load.
   x variable store: y variable load.
   y variable store: tmpVar load} parenthesize


a `swapWith: b


Works with: GNU Smalltalk

An OrderedCollection can collect any kind of objects; so this swap implementend extending the OrderedCollection class is really generic.

OrderedCollection extend [
    swap: a and: b [
	t := self at: a.
	self at: a put: (self at: b).
	self at: b put: t


The "canonical" version from M. Emmers tutorial:

* SWAP(.V1, .V2) - Exchange the contents of two variables.
*  The variables must be prefixed with the name operator
*  when the function is called.

        DEFINE('SWAP(X,Y)TEMP')              :(SWAP_END)
        $X = $Y
        $Y = TEMP                            :(RETURN)

Standard ML

Tuples are immutable in Standard ML. This function doesn't mutate anything, but simply returns a new pair with the order of the elements switched.

fun swap (x, y) = (y, x)

If the arguments are constrained to be reference values, a swap function is simple:

fun swapref (x, y) =
    let temp = !x in x := !y; y := temp end


The Mata swap function is built-in.

b="ars longa vita brevis"
swap(a, b)

Notice that swap only works with variables, not with indexed arrays. For instance, swap(a[i],a[j]) does not work. One would instead write a[(i,j)]=a[(j,i)].


func swap<T>(inout a: T, inout b: T) {
  (a, b) = (b, a)

Note: The Swift standard library has already a swap function.


Works with: Tcl
proc swap {aName bName} {
    upvar 1 $aName a $bName b
    lassign [list $a $b] b a
Works with: Tcl
proc swap {aName bName} {
    upvar 1 $aName a $bName b
    foreach {b a} [list $a $b] break


proc swap {aName bName} {
    upvar 1 $aName a $bName b
    set a $b[set b $a; list]
set a 1
set b 2
puts "before\ta=$a\tb=$b"
swap a b
puts "after\ta=$a\tb=$b"
before	a=1	b=2
after	a=2	b=1

An idiomatic method:

set a 1
set b 2
puts "before\ta=$a\tb=$b"
set a $b[set b $a;lindex {}]
puts "after\ta=$a\tb=$b"
before	a=1	b=2
after	a=2	b=1


As with other stack-based languages (e.g. Factor and Joy), the solution to this task is a trivial matter of swapping the top two operands on the stack:



TXR Lisp has a swap macro operator. However, an operator just like it can be user-defined (let us call it swp). Moreover, the user-defined version can be just as robust, ensuring once-only evaluation for both expressions.

Swapping can be achieved with pset and rotate also. We won't use these in the following examples.

Naive macro

This allows multiple evaluation of the argument expressions.

(defmacro swp (left right)
  (with-gensyms (tmp)
    ^(let ((,tmp ,left))
       (set ,left ,right
            ,right ,tmp))))

Using placelet

TXR Lisp's placelet macro allows the programmer to bind a lexically scoped alias for a syntactic place. The place can be accessed and stored through this alias. Yet, the place is evaluated only once. With placelet it is easy to write many kinds of place-manipulating macros very simply. We can write a robust swap which evaluates the left and right expressions just once:

(defmacro swp (left right)
  (with-gensyms (tmp lpl rpl)
    ^(placelet ((,lpl ,left)
                (,rpl ,right))
       (let ((,tmp ,lpl))
         (set ,lpl ,rpl
              ,rpl ,tmp)))))

Using place expanders

Finally, the following is closely based on how swap is actually implemented in TXR Lisp's library. This explicitly uses the general mechanism for handling places, on which placelet is based also:

(defmacro swp (left right :env env)
  (with-gensyms (tmp)
    (with-update-expander (l-getter l-setter) left env
      (with-update-expander (r-getter r-setter) right env
        ^(let ((,tmp (,l-getter)))
           (,l-setter (,r-getter))
           (,r-setter ,tmp))))))

with-update-expander is a macro which writes code for accessing and updating a place, and makes that code available as local macros. The result is wrapped around the body of code passed to the macro; the body can access these functions, using a backquote to insert the symbols which refer to them. For instance the macro call (,l-getter) expands to code which accesses the prior value of the left place, and (,r-setter ,tmp) stores the value of the temporary variable into the right place.

UNIX Shell

Works with: ksh93
$ swap() { typeset -n var1=$1 var2=$2; set -- "$var1" "$var2"; var1=$2; var2=$1; }
$ a=1 b=2                                                                         
$ echo $a $b
1 2
$ swap a b                                                                        
$ echo $a $b                                                                      
2 1
$ swap a b  
$ echo $a $b
1 2
Works with: bash version 4.2
$ swap() { local var1=$1 var2=$2; set -- "${!var1}" "${!var2}"; declare -g "$var1"="$2" "$var2"="$1"; }
$ a=1 b=2
$ echo $a $b
1 2
$ swap a b
$ echo $a $b
2 1
$ swap a b
$ echo $a $b
1 2


Most functions are polymorphic without any special provision to that effect. Swapping a pair is a very inexpensive operation because no actual copying or overwriting is performed.

pmgs("x","y") = ("y","x")    # the pattern matching way

ugs = ~&rlX                  # the idiosyncratic Ursala way

#cast %sWL

test = <pmgs ('a','b'),ugs ('x','y')>


Using the view to shuffle the stack.

[swap [a b : b a] view].

1 2 swap
= 2 1
'hello' 'hi' swap
='hi' 'hello'


// user-defined swap verb -- parms are passed by alias, not value, so they can be updated:

'<==> [_a] @FN [_b] { _a _b = _b _a } by_alias: ; 

// test out swap verb

@VAR a = 12345;
@VAR b = "*****";

@SAY "a=" a "   b=" b;

\b <==> \a;                    // "\" verb prevents evaluation of a and b here, 
                               // so they can be passed by alias to <==>    
@SAY "a=" a "   b=" b;

a b = b a;                     // swap them back, just using the usual  =  verb

@SAY "a=" a "   b=" b;

Visual FoxPro

Since Visual FoxPro is not strongly typed, this will work with any data types.
 *!* Swap two variables
 LOCAL a, b
 a = 1
 b = "Hallo"
 ? a, b
 *!* Pass a and b by reference
 Swap(@a, @b)
 ? a, b

PROCEDURE Swap(v1, v2)
dum = v1
v1 = v2
v2 = dum
 1 Hallo
 Hallo 1


There's a primitive for modifying bindings.

(swap! x y)

New bindings can be created in parallel.

let (x y) (list y x)


Wren is dynamically typed and, once a variable has been declared, a value of any type can be assigned to it. Generic programming is not therefore a problem here.

However, it is still not possible to write a generic swap function. This is because all variables are passed to functions or methods by value and prescribing that they should be passed by reference instead is not supported. Moreover, variables of simple types (numbers, bools and nulls) are never boxed and so a function cannot mutate the original variable.

Perhaps the nearest we can get to a generic swap function is to pass the variables in a list (the list's address is then passed by value under the hood), swap the list elements (lists do have a swap method) and then unpack the list to the original variables after the function returns.

Another approach would be to box simple variables (using a user defined class) so that they can be mutated. However, the problem with this is that they usually need to be 'unboxed' when the function returns.

Both approaches are illustrated below.

var swap = { |l| l.swap(0, 1) }

var a = 6
var b = 3
var c = [a, b] // pass a list instead of individual variables
a = c[0] // unpack
b = c[1] // ditto
System.print("a is now %(a)")
System.print("b is now %(b)")

// all user defined classes are reference types
class Box {
    construct new(v) { _v = v }
    v { _v }
    v=(value) { _v = value }

// by passing boxed arguments we can mutate them
var boxSwap = { |a, b|
    var t = a.v
    a.v = b.v
    b.v = t

var d =
var e =, e)
d = d.v // unbox
e = e.v // ditto
System.print("d is now %(d)")
System.print("e is now %(e)")
a is now 3
b is now 6

d is now 8
e is now 4


There is a keyword for swapping variables. Here is an example of 2 ways to swap variables

Swap Keyword

set x=1;
set y=2;
log("Before Swap");
swap x y;
log("After Swap");
Before Swap
After Swap

Swap Function

set x = 1;
set y = 2;
func Swap(a,b){
set z = Swap(x,y);


The name Swap is normally used to call an intrinsic routine that swaps bytes in an integer. Thus Exch is used here instead. A and B must both be the same size.

include c:\cxpl\codes;

proc Exch(A, B, S);
char A, B, S;
int  I, T;
for I:= 0 to S-1 do
        [T:= A(I);  A(I):= B(I);  B(I):= T];

real X, Y;
[X:= 3.0;  Y:= 4.0;
Exch(addr X, addr Y, 8);
RlOut(0, X);  RlOut(0, Y);  CrLf(0);
    4.00000    3.00000


Yorick has a built-in function swap for exchanging the contents of two variables without requiring a temporary copy. Example of use:

> a = 1
> b = "foo"
> swap, a, b
> a
> b

Swapping elements in an array can be accomplished using index lists. Arbitrary permutations of swaps are also straightforward. Example:

> foo = [10,20,30,40,50]
> foo([1,2]) = foo([2,1])
> foo
> foo([3,4,5]) = foo([4,5,3])
> foo

Z80 Assembly

Zilog Z80

The Z80 has a few commands for swapping register contents:

  • EX DE,HL will swap the contents of DE with the contents of HL. This is helpful because many instructions are only compatible with HL.
  • EXX will swap out BC, DE, and HL with their "shadow registers." To switch back to the standard registers use EXX again. The accumulator and flags have their own shadow registers, which can be swapped with EX AF,AF'.
  • EX (SP),HL will exchange HL with the top two bytes of the stack.
  • Any two registers that you can PUSH/POP can be exchanged by pushing both registers and popping them in the "wrong" order on purpose. For example, to swap the IX and IY registers:
push ix
push iy
pop ix   ;the value that was once in IY is now in IX
pop iy   ;the value that was once in IX is now in IY

Swapping two memory locations takes a bit more work. The Z80 can do an 8-bit swap or a 16-bit swap on memory. There are several ways to do this, but the methods differ slightly depending on which addressing modes you use. Only the accumulator can load/store a single byte directly from/to memory, but the register pairs BC, DE, HL, SP, IX, and IY can all load a 16-bit word directly from memory, or vice versa.

;8-bit swap using the stack.

ld a,(&C000)
push af
   ld a,(&D000)
   ld (&C000),a   ;store the byte at &D000 into &C000
pop af            ;now a = the byte at &C000
ld (&D000),a      ;now the byte at &D000 equals the byte that was originally at &C000
;16-bit swap:
ld hl,(&C000)     ;load the byte at &C000 into L and the byte at &C001 into H.
ld de,(&D000)     ;load the byte at &D000 into E and the byte at &D001 into D.
ld (&D000),hl     ;store the contents of L into &D000 and H into &D001.
ld (&C000),de     ;store the contents of E into &C000 and D into &C001.

Game Boy

The Game Boy is missing all the exchange commands that the Z80 and 8080 both have. There are no shadow registers, EX, or EXX, and you cannot load/store 16-bit register pairs to/from immediate memory addresses. The only register that can store directly to memory is the stack pointer (SP), and it can only store, not load! But not all is lost. EX DE,HL can be mimicked with the following sequence; although it's much slower than the Zilog's EX DE,HL, it's as quick as it possibly can be.

push hl
ld h,d
ld L,e  ;store de into HL. This is much faster than "push de pop hl."
pop de ;put old HL into DE

Swapping from memory must be done one byte at a time on the Game Boy. The Game Boy exclusive LDI/LDD commands can help us in this task.

ld hl,&C000
push bc
ldi a,(hl)  ;equivalent of "ld a,(hl) inc hl" but is faster than the two separately. Some assemblers call this "ld a,(hl+)"
ld c,a
ld a,(hl)   ;we don't need to increment hl this time. There's no wasted time or bytecode if we did however. 
ld b,a      ;on Zilog Z80 we would have just done "LD BC,(&C000)" but Game Boy can't do that.

; now we do the swap.
ld hl,&D000

ld a,(hl)        ;get the byte at &D000
ld (&C000),a     ;store it into &C000
ld (hl),c        ;store the old byte at &C000 into &D000

inc hl           ;inc HL to &D001

ld a,(hl)        ;get the byte at &D001
ld (&C001),a     ;store it at &C001
ld (hl),b        ;store the old byte at &C001 into &D001
pop bc


As a general case, no. Here a fake around for two known classes (which are the same for ease of example):

class C{var v; fcn init(n){v=n}}
var c1=C(1), c2=C(2);
println(c1.v," : ",c2.v);
fcn swap(ca,cb,name){
   ca.setVar(name,cb.resolve(name)); cb.setVar(name,tmp)
println(c1.v," : ",c2.v);
1 : 2
2 : 1