Thread: Need regexp help

Originally Posted by MK27

okay, okay, "linein" is probably the first function I ever wrote. Or close to it. And the "Teach Yourself C in 21 Days" I got from the library (1990) doesn't mention fgets, but it does warn against the use of "gets".

I've never read the book, so that's entirely possible.

Originally Posted by Mk27

Excuses aside, however, I will defend myself by noting that with fgets "You must supply count characters worth of space in [the char pointer]" (GNU documentation). So linein seems more versatile to me -- ie. without testing it looks like you have to do your own malloc using fgets.

So long live linein. Or else I'm still wrong.

There are two advantages to fgets in this situation (from my perspective): 1. Since I am managing my own memory, I know exactly where everything is. I know when my memory is created, and when I need to free it. I can't do a memory leak every time through a for-loop, as in the example above with a memory leak when reading every single line of a file, without making it tremendously obvious that malloc is being called 10000000 times and free once. I get to see the matching malloc/free calls, rather than malloc being literally three-times-removed from my code. I know that I have a temporary buffer that I have to strcpy with, but I can then reuse immediately for the next line. (I realize that programming has Moved On from any of this sort of thing with memory management, gc, etc. Fortunately I have a second point.) 2. Using fgets (or whichever other standard function you choose to use) means that any computer programmer off the street -- such as, I don't know, someone who's having trouble with a regex, for instance -- knows how it works, what's necessary to use it, and exactly what it does and does not do. No surprises like "oh and it calls malloc for you too".

Edit: The advantage of getline is what I alluded to at the end of point one -- since each call gets its own slice of memory, you can just store pointers in an array or whatever you want -- no strcpying! Unfortunately, that isn't what we ended up doing.

Obviously, I'm not trying to convince the world that linein is really the future, esp. since it relies on getline which is GNU specific I think.

Still, just because you didn't call malloc yourself doesn't mean you can't call free if you know malloc is being called -- which is why in post #21 I corrected myself by adding a "free" call in linefile. Which only needs to be called once anyway, since even if the while loop does iterate 1000000 times, it just reassigns to the same local variable (line).

As for our friend needing regexp help, I think he got it...

Originally Posted by tabstop

Eh, you only have to type it once, but it needs to be inside the loop (if getline happens 1000000 times, free needs to happen 1000000 times, or else you've leaked the first 999999 of them).

Well. This is probably due to my misunderstanding of this (boldface) statement in The GNU C Library Reference Manual:

Do not expect to find any data (such as a pointer to the next block in a chain of blocks) in the block after freeing it.

to mean that free just plain eliminates the variable. So I thought this would fail

Code:

int main() {
        char *this;
        while (1) {
                this=malloc(5);
                strcpy(this,"that");
                puts(this);
                free(this);
        }   
}

which it doesn't. More thinking for me to do...thanx tabstop

ps. by "leaking" memory you mean using excessively unnessesarily -- or could something genuinely bad happen, like an overwrite? And isn't that counter-intuitive -- shouldn't multiple malloc's on the the same variable deal with the same starting block? Or is that impossible because the size will vary in the stack (or wherever)?