Call-by-name, by-reference, by-sharing

stereobooster - Dec 8 '20 - - Dev Community

Value vs reference

There are a lot of parameter-passing strategies. Let's start from two:

  • call-by-value, the argument for a function parameter is a copy of the value of the argument
  • call-by-reference, the function is given the address of the argument

For example, C uses call-by-value strategy for all arguments except pointers, for which it uses call-by-reference

Variable assignment

Described strategies are always mentioned in the context of the variable binding for function calls, but they apply to all variable bindings e.g. assignments as well.

int a = 1;
int* b = malloc(sizeof(*b));
int* c = malloc(sizeof(*c));
Enter fullscreen mode Exit fullscreen mode

call-by-value

*b = a;
// it copies value, so change of variable `a` isn't reflected in `b`
a = 2;
printf("a = %d, b = %d\n", a, *b);
Enter fullscreen mode Exit fullscreen mode

call-by-refernce

b = &a;
// change of variable `a` is reflected in `b`
a = 3;
printf("a = %d, b = %d\n", a, *b);
Enter fullscreen mode Exit fullscreen mode


// and vice versa - change of variable `b` is reflected in `a`
*b = 4;
printf("a = %d, b = %d\n", a, *b);
Enter fullscreen mode Exit fullscreen mode

One more example of call-by-reference

b = c;
// change of variable `b` is reflected in `c`
*b = 5;
// but change of variable `a` isn't reflected in `b` anymore
a = 6;
printf("a = %d, b = %d, c = %d\n", a, *b, *c);
Enter fullscreen mode Exit fullscreen mode

We can think about the next example as call-by-reference, which creates a reference to a new "anonymous" variable

// allocates new memory cell for b
b = malloc(sizeof(*b));
*b = 7;
printf("a = %d, b = %d, c = %d\n", a, *b, *c);
Enter fullscreen mode Exit fullscreen mode

Function call

#include<stdio.h>

void modify(int p, int* q) {
    p = 10; // passed by value
    *q = 20; // passed reference
}

int main() {
    int a = 0;
    int b = 0;
    modify(a, &b);
    printf("a = %d, b = %d, c = %d\n", a, b); // a = 0, b = 20
    return 0;
}
Enter fullscreen mode Exit fullscreen mode

Call-by-sharing

Now the confusing part. Most popular languages, for example, Python, JavaScript, and Java use call-by-sharing. And people often confuse it with call-by-reference.

We call the argument passing technique call by sharing because the argument objects are shared between the caller and the called routine. ... it is similar to argument passing in LISP.
-- CLU Reference manual

In call-by-sharing:

  • primitive values, like numbers, booleans and similar behave, like in call-by-value
  • non-primitive values, like objects, arrays and similar behave, like in call-by-reference
    • when you mutate reference you mutate all "aliases"
    • when you reassign reference you change "alias" - variable will contain a new reference and you can't reassign values in other "aliases"
    • it is not possible to create a reference to a reference, like int**

In case of function calls it means that you can mutate variable in caller scope (because they are "shared"), but you can't reassign them in caller scope:

call-by-reference call-by-sharing call-by-value
can reassign in caller scope + (see note) - -
can mutate in caller scope + + -

Note:

  • to reassign, for example, integer or struct you need to use pointer int*, struct Test*
  • to reassign, for example, an array you need to use "double" pointer int**

Note 2: "primitive values" are immutable

Immutability and reassignment

If we would prohibit mutations and reassignment (like in many pure functional languages) 3 given strategies will look the same. Call-by-value is more expensive though because it needs to copy values.

Code examples

call-by-reference

#include<stdio.h>
#include <stdlib.h>

void mutate(int* a) {
    a[0] = 2;
}

// doesn't work
void reassign(int* a) {
    int b[1] = {3};
    a = b;
}

void reassign2(int** a) {
    int b[1] = {4};
    *a = b;
}

struct Test {
   int counter;
};

void reassign3(struct Test* c) {
    struct Test d = {counter: 5};
    *c = d;
}

int main() {
    // same as int a[1] = {1};
    int* a = malloc(sizeof(a));
    *a = 1;

    mutate(a);
    printf("a[0] = %d\n", a[0]);
    reassign(a);
    printf("a[0] = %d\n", a[0]);

    int** b = malloc(sizeof(b));
    b = &a;
    reassign2(b);
    printf("a[0] = %d\n", a[0]);

    struct Test c = {counter: 0};
    reassign3(&c);
    printf("c.counter = %d\n", c.counter);
    return 0;
}
Enter fullscreen mode Exit fullscreen mode

Call-by-sharing:

function mutate(a) {
  a[0] = 2;
}

// doesn't work
function reassign(a) {
  a = [3];
}

// not possible
// function reassign2(a) {}
// function reassign3(c) {}

let a = [1];

mutate(a);
console.log(a);
reassign(a);
console.log(a);
Enter fullscreen mode Exit fullscreen mode

PS

If you want to read more about other parameter-passing strategies, I recommend this resource.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .