Re: on allowing "int a" definition everywhere

Ben Bacarisse <ben@bsb.me.uk> writes:

Keith Thompson <Keith.S.Thompson+u@gmail.com> writes:

The original code:
>
    if (FILE* f = fopen("file.txt", "r"))

>
I think the proposal would be a declaration to precede an expression:
>
  if (FILE* f = fopen("file.txt", "r"); f)

I don't think so, or at least I don't think it's only that.

In C++17, the syntax for an if statement is (leaving some stuff out):

    if ( init-statement[opt] condition ) statement

where an init-statement is an expression statement or a simple
declaration (which includes a trailing semicolon), and a condition is an
expression or a declarator with an initialization.  So these are
equivalent in C++:

    if (FILE* f = fopen("file.txt", "r"))
    if (FILE* f = fopen("file.txt", "r"); f)

    {
       /*...*/ fclose(f);
    }
>
doesn't allow for taking some non-trivial action of the fopen() call
fails, but if a declaration in an if condition is visible in the else
clause, you could write something like:
>
    if (FILE* f = fopen("file.txt", "r")) {
       /*...*/
       fclose(f);
    }
    else {
        perror("file.txt");
        exit(EXIT_FAILURE); // or try something else
    }

>
What's the significance of the declared name being in scope in the else
clause here?

My brain seems to have taken a detour while I was writing that.

fopen() returns either a valid pointer or NULL; if it fails, there's no
extra information in the pointer itself.  But if a function's result on
failure includes information about how it failed, accessing the value in
the else clause is useful.  (And C++ supports this.)

--
Keith Thompson (The_Other_Keith) Keith.S.Thompson+u@gmail.com
void Void(void) { Void(); } /* The recursive call of the void */