How to swap two variables without using a third variable?

Question

Can you swap the values of two variables without using a third variable as a temporary placeholder?

Answer 1

The canonical way to swap two variables in Python is

a, b = b, a

Please note than this is valid whatever the "type" of a or b is (numeric, string, tuple, object, ...). Of course, it works too if both variables reference values of different types.

---

As many imperative languages, Python evaluates assignments right to left. Conceptually all behave like if a tuple was build for the right hand part of the expression, and then deconstructed to perform the affectation to the left hand part. This has already been explained more clearly than I can here: https://stackoverflow.com/a/14836456/2363712

The real details are implementation dependent though. For example, to build on a comment by @undefined is not a function below, the CPython virtual machine has a ROT_TWO opcode that swap the two top-level items on the stack, and so allow to optimize such affectation. See this previous answer for a detailed explanation: https://stackoverflow.com/a/21047622/2363712

Answer 2

This is the main snippet of code:

 x = x + y;  // x now becomes 15
 y = x - y;  // y becomes 10
 x = x - y;  // x becomes 5

This is what you friend meant.

Answer 3

If your friend is a "math-snob", he may have in mind a particular trick, which you can use in languages where you can apply the XOR function to the bitstring representation of numbers.

Say we have variables X and Y, with starting values of a and b respectively. Perform the following assignments (the values of the variables which result are shown as comments):

(start)      # X == a; Y == b
X = X XOR Y  # X == a XOR b;  Y == b
Y = X XOR Y  # X == a XOR b;  Y == b XOR (a XOR b)
X = X XOR Y  # X == (a XOR b) XOR b XOR (a XOR b);  Y == b XOR (a XOR b)

Because XOR is associative, we can regroup the resulting equations as follows:

X == (a XOR a) XOR (b XOR b) XOR b
Y == (b XOR b) XOR a

Because x XOR x == 0 and x XOR 0 == x, we can simply remove all those pairs of variables XOR'ed with themselves, and what's left is:

X == b
Y == a

which is what we wanted, to switch the values without using a third variable.

It's been quite a while since I did any bit manipulation in Python, so I cannot tell you whether this trick works in Python, but there are languages where it works. I also can't tell you whether it actually has sufficient benefits to balance out its non-obviousness.

Answer 4

You can directly try for:

a, b = b, a

It is a canonical way to exchange the value without using a third variable.

Answer 5

A seemingly simple question. In retrospect, presumably designed to determine whether or not you think mathematically. I pondered, it's not a simple problem but not out of reach.

As research reveals this is a fairly common question with many good and bad answers. I believe I've found an illustrative solution:

#!/usr/bin/env python
# -*- coding: utf-8 -*-

# MODULES


# VARIABLES
x = 20
y = 10


# FUNCTIONS
# Swap 2 vars: longhand method, the portable way:
def swapNosL(val1, val2):
    print("BEFOR: val1: %r,  val2: %r") % (val1, val2)
    val1 = val1 + val2
    val2 = val1 - val2
    val1 = val1 - val2
    print("AFTER: val1: %r,  val2: %r") % (val1, val2)
    return(val1, val2)


# Swap 2 vars: shorthand method, the non/less-portable way:
def swapNosC(val1, val2):
    print("BEFOR: val1: %r and val2: %r") % (val1, val2)
    val1, val2 = val2, val1
    print("AFTER: val1: %r and val2: %r") % (val1, val2)
    return(val1, val2)


# MAIN PROGRAM
print("")
print("The Problem:")
print("We need to swap 2 variables without using a 3rd.")
print("The values: 'x' is %r and 'y' is %r.") % (x, y)
print("")

(retVal1, retVal2) = swapNosC(x, y)
print("")
print("Now values: 'x' is %r and 'y' is %r.") % (retVal1, retVal2)

print"\n"

While there is some unnecessary repetition, the logic is solid. To baseline:

1) It works with all integers both positive and negative; I haven't tested floats yet.

2) The same memory is used 1 way or the other; only 2 variables are used either way.

3) Portability is (or should) always be a goal. In the case that programming languages go away and you need to port to a new one, research indicates that handling this problem mathematically will allow for greater portability.

In the "bad" example, this method is language-specific and porting to a language would (some day) require a different solution. I hope Python never goes the way of COBOL but the future is a big place.

However, in the "good" example the math is handled in a similar way in the C language. In fact, research also indicates math is generally handled the same way in most languages.

Therefore leaving the math in tact and only negotiating the syntax modification is the more portable method.

At some point I will concede to my friend that this was a good learning opportunity for me. But not for a while. I found the answer but failed by cheating with research.

TT