ACE Tutorial 015
Building a protocol stream


And now the implementation of the Protocol_Stream. There are more lines of code here than we've seen so far but it still isn't complicated. The basic idea is to construct the ACE_Stream with our set of protocol objects that will manipulate the data. Our primary concern in this file is to get everything in the correct order!
// page11.html,v 1.13 1999/09/22 03:13:53 jcej Exp

#include "Protocol_Stream.h"
#include "Protocol_Task.h"

#include "Xmit.h"
#include "Recv.h"

#include "Compressor.h"
#include "Crypt.h"

#include "ace/Stream_Modules.h"

/* You can choose at compile time to include/exclude the protocol
   pieces.
*/
#define ENABLE_COMPRESSION
#define ENABLE_ENCRYPTION

// The usual typedefs to make things easier to type.
typedef ACE_Module<ACE_MT_SYNCH> Module;
typedef ACE_Thru_Task<ACE_MT_SYNCH> Thru_Task;

/* An ACE_Stream is a collection of ACE_Modules.  You can think of it
   as a doubly-linked list if you like.  Each Module contains two
   ACE_Task derivatives.  One of these tasks is used when sending data
   "upstream", the other is used for "downstream" operation.  In some
   cases, you'll only need to move data in one direction.  To provide
   a placeholder for the other direction, ACE_Thru_Task can be used.
   ACE_Thru_Task responds to the put() by simply invoking put_next()
   to send the data to the next module.
 */

/* Do-nothing constructor and destructor
 */

Protocol_Stream::Protocol_Stream (void)
{
}

Protocol_Stream::~Protocol_Stream (void)
{
}

/* Even opening the stream is rather simple.  The important thing to
   remember is that the modules you push onto the stream first will be
   at the tail (eg -- most downstream) end of things when you're
   done.
 */
int
Protocol_Stream::open (ACE_SOCK_Stream &peer,
                       Protocol_Task *reader)
{
  // Initialize our peer() to read/write the socket we're given
  peer_.set_handle (peer.get_handle ());

  // Construct (and remember) the Recv object so that we can read from
  // the peer().
  ACE_NEW_RETURN (recv_,
                  Recv ( this->peer ()),
                  -1);

  // Add the transmit and receive tasks to the head of the stream.  As
  // we add more modules these will get pushed downstream and end up
  // nearest the tail by the time we're done.
  if (stream ().push (new Module ("Xmit/Recv",
                                  new Xmit ( this->peer ()),
                                  recv_)) == -1)
    ACE_ERROR_RETURN ((LM_ERROR,
                       "%p\n",
                       "stream().push(xmit/recv)"),
                      -1);

  // Add any other protocol tasks to the stream.  Each one is added at
  // the head.  The net result is that Xmit/Recv are at the tail.
  if (this->open () == -1)
    return -1;

  // If a reader task was provided then push that in as the upstream
  // side of the next-to-head module.  Any data read from the peer()
  // will be sent through here last.  Server applications will
  // typically use this task to do the actual processing of data.
  // Note the use of Thru_Task.  Since a module must always have a
  // pair of tasks we use this on the writer side as a no-op.
  if (reader)
    {
      if (stream ().push (new Module ("Reader",
                                      new Thru_Task (),
                                      reader)) == -1)
        ACE_ERROR_RETURN ((LM_ERROR,
                           "%p\n",
                           "stream().push(reader)"),
                          -1);
    }

  return 0;
}

/* Add the necessary protocol objects to the stream.  The way we're
   pushing things on we will encrypt the data before compressing it.
*/
int
Protocol_Stream::open (void)
{
#if defined (ENABLE_COMPRESSION)
  if (stream ().push (new Module ("compress",
                                  new Compressor (),
                                  new Compressor ())) == -1)
    ACE_ERROR_RETURN ((LM_ERROR,
                       "%p\n",
                       "stream().push(comprssor)"),
                      -1);
#endif /* ENABLE_COMPRESSION */

#if defined (ENABLE_ENCRYPTION)
  if (stream ().push (new Module ("crypt",
                                  new Crypt (),
                                  new Crypt ())) == -1)
    ACE_ERROR_RETURN ((LM_ERROR,
                       "%p\n",
                       "stream().push(crypt)"),
                      -1);
#endif /* ENABLE_ENCRYPTION */
  return 0;
}

// Closing the Protocol_Stream is as simple as closing the ACE_Stream.
int
Protocol_Stream::close (void)
{
  return stream ().close ();
}

// Simply pass the data directly to the ACE_Stream.
int
Protocol_Stream::put (ACE_Message_Block *&message,
                      ACE_Time_Value *timeout)
{
  return stream ().put (message,
                        timeout);
}

/* Tell the Recv module to read some data from the peer and pass it
   upstream.  Servers will typically use this method in a
   handle_input() method to tell the stream to get a client's request.  */

int
Protocol_Stream::get(void)
{
  // If there is no Recv module, we're in big trouble!
  if (recv_ == 0)
    ACE_ERROR_RETURN ((LM_ERROR,
                       "(%P|%t) No Recv object!\n"),
                      -1);

  // This tells the Recv module to go to it's peer() and read some
  // data.  Once read, that data will be pushed upstream.  If there is
  // a reader object then it will have a chance to process the data.
  // If not, the received data will be available in the message queue
  // of the stream head's reader object (eg --
  // stream().head()->reader()->msg_queue()) and can be read with our
  // other get() method below.
  if (recv_->get () == -1)
    ACE_ERROR_RETURN ((LM_ERROR,
                       "(%P|%t) Cannot queue read request\n"),
                      -1);

  // For flexibility I've added an error() method to tell us if
  // something bad has happened to the Recv object.
  if (recv_->error ())
    ACE_ERROR_RETURN ((LM_ERROR,
                       "(%P|%t) Recv object error!\n"),
                      -1);

  return 0;
}

/* Take a message block off of the stream head reader's message queue.
   If the queue is empty, use get() to read from the peer.  This is
   most often used by client applications.  Servers will generaly
   insert a reader that will prevent the data from getting all the way
   upstream to the head.  */
int
Protocol_Stream::get (ACE_Message_Block *&response,
                      ACE_Time_Value *timeout )
{
  if (stream ().head ()->reader ()->msg_queue ()->is_empty ()
      && this->get () == -1)
    ACE_ERROR_RETURN ((LM_ERROR,
                       "(%P|%t) Cannot get data into the stream.\n"),
                      -1);

  return stream ().head ()->reader ()->getq (response,
                                             timeout);
}


[Tutorial Index] [Continue This Tutorial]