Question

The following code gives an “invalid initializer” error:

     int a[]=(1,2,3);

But the following compiles successfully although it considers ‘,’ as comma OPERATOR and not SEPARATOR:

    int a[][2]={(1,2),(3,4)};

So why is () invalid for a 1D array and not for a 2D array?

Answer 1

In the first example:

int a[]=(1,2,3);

the initializer is a (rather odd) expression of type int. (It contains two comma operators, and yields the value 3.) The object is an array. The initialization is invalid because it’s a type mismatch.

The second:

int a[][2]={(1,2),(3,4)};

is equivalent to:

int a[][2] = { 2, 4 };

which is valid because it’s permissible to omit nested curly braces in an initializer; elements are used to initialize successive elements of the object. The first and third commas are comma operators; the second is a delimiter.

The outermost curly braces are optional if the initializer is simply an expression of the target type, whether it’s a scalar, struct, or union. For example, you can write:

int x = 42;
int y = { 42 };

The outermost curly braces are required for an initializer that specifies element values (for an array, struct, or union object).

For example:

struct foo {
    int x;
    int y;
};

 struct foo arr[2] = { 1, 2, 3, 4 };

is valid — but it’s more clearly written as:

 struct foo arr[2] = { { 1, 2 }, { 3, 4 } };

Apart from the first example being invalid, both are poor style. The first may have been intended to be:

int a[] = { 1, 2, 3 };

and the second either:

int a[][2] = { 2, 4 };

or

int a[][2] = {{1, 2}, {3, 4}};

depending on the intent.