ACE Tutorial 010
Passing chunks of data through an ACE_Message_Queue


We'll look first at main().


// page02.html,v 1.9 1999/09/22 03:13:44 jcej Exp

/* To illustrate the ACE_Message_Queue, we use a derivative of
   ACE_Task<>.  We also derive from ACE_Message_Block to show that we
   don't have memory leaks.  */
#include "task.h"
#include "block.h"

int
run_test (int iterations,
          int threads)
{
  /* Create and start an instance of our Task object.  */
  Task task (threads);

  if (task.open () == -1)
    ACE_ERROR_RETURN ((LM_ERROR,
                       "%p\n",
                       "open"),
                      -1);

  /* Give the threads a moment to open.  This isn't really necessary
    but if we don't we find that all of our blocks are constructed and
    enqueued before any of the threads get created. Basically, the
    sleep() makes the output look more interesting.  */
  ACE_OS::sleep (ACE_Time_Value (1));

  int i;
  for (i = 0; i < iterations; ++i)
    {
      /* Create a new message block to hold our data.  Here, we ask
        for a block that has 128 bytes of data space.  */
      Block *message;
      ACE_NEW_RETURN (message,
                      Block (128),
                      -1);

      /* Grab the "write pointer".  This is a pointer into the data
        area where we can write our data.  After writting the data you
        have to increment the wr_ptr() so that subsequent writes won't
        clobber what you've put there.  */
      ACE_OS::sprintf (message->wr_ptr (),
                       "This is message %d.",
                       i);
      message->wr_ptr (ACE_OS::strlen (message->rd_ptr ()));

      /* Put the message block into the queue.  One of the threads in
        the Task object will pick up the block and "do work" on it.  */
      if (task.putq (message) == -1)
        break;
    }

  /* Once we're done, we have to signal the Task objects to shut
    down. There are several choices including: - Send a message of
    zero length - Send a message with a special content I don't like
    these choices because they're likely to interfere with application
    logic.  Instead, I use the message type feature to send a message
    of type "hangup".  The default type is MB_DATA, so when the tasks
    get a MB_HANGUP type, they know to go away.  */
  Block *message;

  ACE_NEW_RETURN (message,
                  Block (),
                  -1);
  message->msg_type (ACE_Message_Block::MB_HANGUP);
  task.putq (message);

  /* Wait for the threads in our task object to go away.  */
  task.wait ();

  return 0;
}

int
main (int argc, char *argv[])
{
  /* Set the number of iterations through our putq() loop and the
    number of threads to use in our Task<> derivative.  */
  int iterations = argc > 1 ? atoi (argv[1]) : 9;
  int threads = argc > 2 ? atoi (argv[2]) : 2;

  run_test (iterations, threads);

  ACE_DEBUG ((LM_DEBUG,
              "(%P|%t) Application exiting\n"));

  return 0;
}

This looks a lot like our thread-pool server and it even does some things better. In particular, I've scoped the Task object so that it's destructor will have a chance to get called before the application exits. Notice how we write actual data into the message block though. In the thread-pool server we just provided a pointer. Writting the data is actually a more correct way of doing things since you don't get into strange pointer casting situations. What if you want to put complex objects into the message block though? We'll do that in the next tutorial, let's stick with the basics first.

On the next page we'll take a look at our Block object...


[Tutorial Index] [Continue This Tutorial]