I read an exec’d program’s stdout using a pipe:
int pipes[2];
pipe(pipes);
if (fork() == 0) {
dup2(pipes[1], 1);
close(pipes[1]);
execlp("some_prog", "");
} else {
char* buf = auto_read(pipes[0]);
}
To read from stdout, I have a function auto_read which automatically allocates more memory as needed.
char* auto_read(int fp) {
int bytes = 1000;
char* buf = (char*)malloc(bytes+1);
int bytes_read = read(fp, buf, bytes);
int total_reads = 1;
while (bytes_read != 0) {
realloc(buf, total_reads * bytes + 1);
bytes_read = read(fp, buf + total_reads * bytes, bytes);
total_reads++;
}
buf[(total_reads - 1) * bytes + bytes_read] = 0;
return buf;
}
The reason I do it this way is I don’t know how much text the program is going to spew out ahead of time, and I don’t want to create an overly large buffer and be a memory hog. I’m wondering if there is:
- A cleaner way to write this.
- A more memory or speed-efficient way of doing this.
Use
popenif you only need to read from a process and are on a *NIX platform:EDIT: You asked for a way to map a programs output to a file, so here you go: