Thread: Getting FILE * to a string in memory

hi

I have a code library that was written to read from a file stream (FILE *) doing mostly getc ungetc, and processing that data. However I actually need to run against a string already in memory. So I can convert all the code to work over my string array, but what I'd like is somehow get a FILE * to my string.

I don't see any way to do this though, and this code does need to be fast so I'd like to avoid anything that requires writing out the data to disk to get an fd. If I have to I will just convert it all to work on an array vs getc ungetc but there's a fair amount of code to convert.

Any suggestions? (btw this is linux platform)

thanks

thanks, I think mmap will require me to write it to disk first, then map that to mem which I'm trying to avoid, since I already have it in memory.

Loran

Sure, anything is possible:

Code:

#include <string.h>
#include <stdlib.h>
#include <stdio.h>

int main()
{
   int c;
   FILE *str_fp;
   char *str = "your string data";
   char *buff = malloc(strlen(str) +8);
   sprintf(buff, "echo '&#37;s'", str);
   str_fp = popen(buff, "r");
   while((c = fgetc(str_fp)) != EOF)
   {
      putchar(c);
   }
   pclose(str_fp);
   free(buff);
}

It has limitations -- you can't have (unescaped) single quote characters or null bytes in the string. It's more a fun toy than a real method.

Personally, I'd rewrite your code to abstract out the I/O. Then you could just drop in different methods.

Originally Posted by brewbuck

Sure, anything is possible:

It has limitations -- you can't have (unescaped) single quote characters or null bytes in the string. It's more a fun toy than a real method.

Personally, I'd rewrite your code to abstract out the I/O. Then you could just drop in different methods.

thanks, yeh that would be a bit much overhead, and this is some xml data that can be in many different charsets also.

Yeh i may end up having to do that

Originally Posted by xuftugulus

Yes it was an interesting question and i am doing research on it. But the real solution which is to map the memory to a FILE *, i don't suppose it could be done by tinkering with the FILE * structure?

Yes, it can be. This is how popen() does what it does. You'd basically be writing something akin to popen() but which gets data from a memory buffer instead. This is going to be extremely platform-specific, down to the C library and header files. In other words, not portable at all. But possible.

My popen() example is at least a click more portable because it works everywhere popen() is available, which is a lot of places.

And if efficiency is important, then using an external program is out of the question right?

Like I said, my preferred choice would be rewriting it so it didn't directly use FILE* any more. But no, involving another program is not necessarily less efficient (although in this case it probably is).

Originally Posted by xuftugulus

Yes it was an interesting question and i am doing research on it. But the real solution which is to map the memory to a FILE *, i don't suppose it could be done by tinkering with the FILE * structure?
I would probably redefine the used IO functions to some simple implementation, but still i am trying to find a way to abstract it out.
And if efficiency is important, then using an external program is out of the question right?

Thanks, yeh I agree mapping to FILE * is the ideal thing and I'm sure there's a way. Maybe I'll dig into its internals or see if there's anything in Steven's UP

Originally Posted by loran32

Thanks, yeh I agree mapping to FILE * is the ideal thing and I'm sure there's a way. Maybe I'll dig into its internals or see if there's anything in Steven's UP

Mapping it to FILE* is probably the worst way imaginable actually. All it takes is one small detail in the C library to change, and you get to rewrite your code again.

Do it once, do it right, and forget about it.

Originally Posted by brewbuck

Mapping it to FILE* is probably the worst way imaginable actually. All it takes is one small detail in the C library to change, and you get to rewrite your code again.

Do it once, do it right, and forget about it.

thanks

Yeh i don't mean doing something unsupported, I may just take a look at the underlying struct. Now that I think about it i may just try redefing getc and ungetc for that file

Originally Posted by xuftugulus

I hacked a quick solution to the problem...


This way though you lose the original functionality of <stdio.h>. But you can always #undef the macros when the job is done. It also means that you will have to implement the rest of the <stdio.h> routines, or at least the most basic ones.

hey thanks very much, that's great!

really appreciate your help

Loran

Originally Posted by xuftugulus

This way though you lose the original functionality of <stdio.h>. But you can always #undef the macros when the job is done. It also means that you will have to implement the rest of the <stdio.h> routines, or at least the most basic ones.

This is slightly less hackish than actually munging around in the FILE structures. I applaud your effort