Thread: Pipes question

I am trying to write a C application, that utilizes a perl script.

the Perl script i have reads from standard input, modifies the input, and outputs the modified version to stdout.

here is a code sample, it will do a lot more than this, but im just trying to get it to work before i really get into it.

Code:

while (<STDIN>) {

chomp;
my $name = $_;
print "here is some modified output ".$name."\n";

}

What i would ultimately like to do, is have that perl script always running as a subprocess, using execv(), or similar. that way, whenever i need to modify a string in my C program, i can simply pipe it to that perl script, and somehow retrieve the output.

Iv read a ton about two way pipes, but it seems like they all deal with programs that are fully written in C and can utilize both sides of the pipe. I learned about dup2() in college, but cant seem to get it to work right for whatever reason.. So far, iv successfully been able to pipe a string to the perl script, have it modify it, and print it to stdout like it should. my problem, is that i need to read the scripts output from the C program, and iv had a hell of a time trying to do that.

Here is what iv come up with so far, and i just doesnt work very well. sorry for the messy code, this is only temporary and wasnt planning on making it public..

Code:

int readpipe[2];
int writepipe[2];

#define	PARENT_READ	readpipe[0]
#define	CHILD_WRITE	   readpipe[1]
#define CHILD_READ	    writepipe[0]
#define PARENT_WRITE	writepipe[1]

  
   if (pipe (readpipe)){
           fprintf (stderr, "Pipe failed.\n");
           return EXIT_FAILURE;
    }
       
   if (pipe (writepipe)){
           fprintf (stderr, "Pipe failed.\n");
           return EXIT_FAILURE;
   }
    
   if ( (pid = fork()) < 0){
	printf("could not create subprocess!\n");
   }
   else if ( pid == 0 )	{
		
	/* this is the child */

	close(PARENT_READ);
	close(PARENT_WRITE);
	
	dup2(CHILD_READ,  0);  		 close(CHILD_READ);
	dup2(CHILD_WRITE, 1);		close(CHILD_WRITE); 

	execv("./dirmap.pl", argv);    /* start the perl script
    }
   else{

	/* this is the parent *
	
	char string[100];
	char *newstring;
	int nbytes = 100;
	newstring = (char *) malloc (nbytes + 1);
	int rc;
	
	close(CHILD_READ);
	close(CHILD_WRITE);
	
	sprintf(string, "%s", "here is a test string");

	while( rc = write(PARENT_WRITE, string, strlen(string)) ){

   /* EVERYTHING WORKS UP TO THIS POINT, BUT TRYING TO READ IN THE OUTPUT  IS GIVING ME TROUBLE */

		getline(&newstring, &nbytes, PARENT_READ);
		fprintf(stderr, "FROM C: %s\n", newstring);
         }
    }

If someone could provide a logic outline of the best way to do this, or a code sample, i would be forever greatful.

and i cant "just do the string modification in the C program" because i will be utilizing some pre-made perl modules .

thanks!

you're going about it the right way
just let the child know when its time

heres an exaple I wrote up....it takes input, feeds input into child echo with stdin, then prints output from stdout with parent

Code:

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

void child_handler(int input_pipe[], int output_pipe[])
{
        int ret_val;
        char buffer;

        /* close unneeded */
        close(input_pipe[1]);
        close(output_pipe[0]);

        ret_val = read(input_pipe[0], &buffer, 1);

        if (ret_val <= 0)
        {
                perror("child read");
                close(input_pipe[0]);
                close(output_pipe[1]);
                exit(1);
        }

        if (buffer != '-')
        {
                printf("unknown data\n");
                close(input_pipe[0]);
                close(output_pipe[1]);
                exit(1);
        }

        dup2(input_pipe[0],  0);
        dup2(output_pipe[1], 1);

        ret_val = execl("/bin/echo", (char *)0);

        if (ret_val == -1)
        {
                perror("execl");
                close(input_pipe[0]);
                close(output_pipe[1]);
                exit(1);
        }
}

void parent(int input_pipe[], int output_pipe[])
{
        int ret_val;
        char buffer[32];

        close(input_pipe[1]);
        close(output_pipe[0]);

        printf("Enter line to echo (starting with - for valid input)\n>");

        fgets(buffer, sizeof(buffer), stdin);

        ret_val = write(output_pipe[1], &buffer, sizeof(buffer));

        if (ret_val <= 0)
        {
                perror("parent write");
                close(input_pipe[0]);
                close(output_pipe[1]);
                exit(1);
        }

        ret_val = read(input_pipe[0], &buffer, sizeof(buffer));

        if (ret_val <= 0)
        {
                perror("parent read");
                close(input_pipe[0]);
                close(output_pipe[1]);
                exit(1);
        }

        printf("Echo Back: %s", buffer);
}

int main(void)
{
        int parent_to_child[2];
        int child_to_parent[2];
        int pid;
        int ret_val;

        ret_val = pipe(parent_to_child);
        if (ret_val == -1)
        {
                perror("PIPE p2c");
                exit(1);
        }

        ret_val = pipe(child_to_parent);
        if (ret_val == -1)
        {
                perror("PIPE p2c");
                exit(1);
        }

        pid = fork();

        switch (pid)
        {
                case -1: perror("fork"); exit(1);
                case 0: child_handler(parent_to_child, child_to_parent);
                default: parent(child_to_parent, parent_to_child);
        }

        return 0;
}

output:
Enter line to echo (starting with - for valid input)
>invalid input
unknown data
parent read: Success

2:
Enter line to echo (starting with - for valid input)
>-valid input
Echo Back:
valid input


so basically changing echo to your script and using loops (carefull no deadlock) to stay open and take multiple lines of data

::edit:: woops forgot to close the pipes in functions, just visualize that I did

I ALMOST have this working...it looks like my C program is blocking at the read() function.

Code:

ret_val = read(PARENT_READ, buffer, sizeof(buffer) + 1) ;

/* WONT GET PAST THIS POINT */
		printf("parent has read\n");

        if (ret_val <= 0)
        {
                perror("parent read");
                close(PARENT_READ);
                close(PARENT_WRITE);
                exit(1);
        }

        printf("Echo Back: %s", buffer);

this is my perl script

Code:

while (<STDIN>) {
chomp;
print  "here is some modified output ".$_."\n";


/*IF I PUT AN EXIT STATEMENT HERE, EVERYTHING WORKS */
}

I noticed if i put an 'exit' statement inside that while loop, everything works perfect, except for the fact that the script exits. without the 'exit' statement, the C program blocks at read(); i would really like to find a way to keep the script running, to avoid the overhead of creating a process every time i need to use it (this program will be handling 10,000+ file names that need to be modified.

please help!

I'm gonna need to see more, it looks like you've changed it a bit. If it's blocking there, its extremely likly that nothing is being writen into the pipe it is trying to read from. So it sits and waits to read. so maybe you have your pipes mixed up.

::edit

ahhh I see now...it blocks on a 2nd pass.....If you are still using execl to start your perl script, that may be the problem...don't quote me on this but I think the exec function is transfering your scripts image with the childs image, this means, your script will run through..but the child doesn't go on beyond the exec. so it never produces 2nd output, and never writes that output to the pipe, so the parent never reads the non-ouput.
if that makes any sense to you. exec functions may not be the way to go. but then again I may be totally wrong about this copying the process image over.

::edit edit
nope I'm right, man execl, and I thaught I forgot most everything I learned in uni, that I don't still use

here is the whole code.

I KNOW for a fact that the perl script is both getting the input from the C program, and writing it back to the pipe the correct way. i know that, because when i do put the exit statement in the perls while loop, everything in the C program works perfectly. so its only when the perl script doesnt exit that the read() function blocks. which tells me that the read() function is missing some kind of EOF marker or something similar...i just cant figure it out.

I think your code worked well because the 'echo' was exiting after it output, while this perl script is not.
here is the code

Code:

GLOBAL int readpipe[2];
GLOBAL int writepipe[2];

#define	PARENT_READ		readpipe[0]
#define	CHILD_WRITE		readpipe[1]
#define 	CHILD_READ		writepipe[0]
#define 	PARENT_WRITE	writepipe[1]

void child_handler()
{
        int ret_val;
        char buffer;

        /* close unneeded */
        close(PARENT_READ);
        close(PARENT_WRITE);

        dup2(CHILD_READ,  0);
        dup2(CHILD_WRITE, 1);

        ret_val = execl("./dirmap.pl", (char *)0);
	
        if (ret_val == -1)
        {
                perror("execl");
                close(CHILD_READ);
                close(CHILD_WRITE);
                exit(1);
        }
}

void parent(char * dir_string){

        int ret_val;
        char buffer[100];
        
        sprintf(buffer,"%s\n", dir_string);

		printf("Parent is writing %s\n", buffer);
        ret_val = write(PARENT_WRITE, &buffer, sizeof(buffer));
        
        if (ret_val <= 0)
        {
                perror("parent write error\n");
                close(PARENT_READ);
                close(PARENT_WRITE);
                exit(1);
        }
		
		printf("parent is reading\n");
		
        ret_val = read(PARENT_READ, buffer, sizeof(buffer) + 1) ;

		printf("parent has read\n");

        if (ret_val <= 0)
        {
                perror("parent read");
                close(PARENT_READ);
                close(PARENT_WRITE);
                exit(1);
        }

        printf("Echo Back: %s", buffer);
}

if (pipe (readpipe)){
		fprintf (stderr, "Pipe failed.\n");
		return EXIT_FAILURE;
	}
	
	if (pipe (writepipe)){
		fprintf (stderr, "Pipe failed.\n");
		return EXIT_FAILURE;
	}
	
	/* pipes started */
	    
	/* fork the child process for dirmap.pl */
	pid = fork();
	
	if( pid < 0){
		printf("Could not create dirmap subprocess!\n");
		return EXIT_FAILURE;
	}
	else if ( pid == 0 )	/* in the child */
	{
		
		child_handler();
	}
	else{
	
        close(CHILD_READ);
        close(CHILD_WRITE);
        
	//char *dp = "This is a test\n";
    create_archive ();
    name_close ();
    	parent("This is a test");
    	sleep(1);
    	parent("This is another test");
 }

The input to your Perl script is likely buffered. This means that the output from your C program is queued until enough of it has been received, so that that larger chunk of data can be sent instead of a bunch of smaller chunks (larger chunks are more efficient). So your Perl script is waiting for data, even though your C program has already supplied the data.

The solution to this is to get the C program to flush the stream, possibly with fflush(). Supplying an EOF flushes the buffer automatically, but ends the perl script.

I may have misread your question but I think that's what's happening.

BTW, there's a good perl forum on DaniWeb (see my signature).

You could try putting

Code:

$| = 1;

In your perl script to make it flush the output buffer every time it does a write.