String in your Colon – The Daily WTF

Anders sends us an example which isn’t precisely a WTF. It’s just C handling C strings. Which, I guess, when I say it that way, is a WTF.

while(modPath != NULL) 
    p = strchr(modPath, ':');
    if(p != NULL) 
      *p++ = '\0';
    
    dvModpathCreate(utSymCreate(modPath));
    modPath = p;
 while(modPath != NULL);

We start with a variable called modPath which points to a C string. So long as modPath is not null, we’re going to search through the string.

The string is in a : separated format, e.g., foo:bar:goo. We want to split it. This function does this by being very C about it.

It uses strchr to find the address of the first colon. If we think about this in C strings, complete with null terminators, it looks something like this:

"foo:bar:goo\0"
 ^  ^
 |  |
 |  p
modPath

We then replace : with \0 and increment p, doing a “wonderful” blend of using the dereference operator and the post-increment operator and an assignment to accomplish a lot in one line.

"foo\0bar:goo\0"
 ^    ^
 |    |
 |    p
modPath

So now, modPath points at a terminated string foo, which we then pass down through some functions. Then we set it equal to p.

"foo\0bar:goo\0"
      ^
      |
      p
      modPath

This repeats until strchr doesn’t find a :, at which point it returns NULL. Our loop is guarded by a check that modPath (which gets set equal to p) can’t be null, so that breaks us out of the loop.

And enters us immediately into another, single line loop with no body, which immediately exits as well. I suspect that originally this was written as a dowhile, and then someone realized that it could just be a plain while, and completely forgot to remove the second while clause.

This is, honestly, a pretty common idiom in C. It’s arguably wrong to even put it here; aside from the bad while clause, you’ll see this kind of string handling all the time. But, maybe, that is the WTF.