Question

I’m reviewing a C++ project and see effectively the following:

std::vector<SomeType> objects;

//then later
int size = (int)objects.size();
for( int i = 0; i < size; ++i ) {
    process( objects[i] );
}

Here’s what I see. std::vector::size() returns size_t that can be of some size not related to the size of int. Even if sizeof(int) == sizeof(size_t) int is signed and can’t hold all possible values of size_t. So the code above could only process the lower part of a very long vector and contains a bug. The right way would be to use size_t for both the size variable and the loop index.

That said I’m curious of why the author might have written this?

My only guess is that first he omitted the (int) cast and the compiler emitted something like Visual C++ C4018 warning:

warning C4018: '<' : signed/unsigned mismatch

so the author though that the best way to avoid the compiler warning would be to simply cast the size_t to int thus making the compiler shut up.

Is there any other possible sane reason for that C cast?

Answer 1

No, that is probably the reason. Plus the fact that the vector would probably never be so long that it would risk truncating the size (the app. developer would know that).

And… maybe in some parts of the program he actually compared “size” with something else that was int-typed, so making size a “size_t” type would fix it in one place, but break it somewhere else.