On Mon, 11 Aug 2008 19:13:56 +0200, Gabriel Rossetti wrote:

Gabriel Rossetti wrote:

...

Hello,

I'm needing to send small packets to update a progress bar. My update msgs were all being sent in big packets, so I tried setting setTcpNoDelay() in connectionMade() but it doesn't change anything, is it broken?

Thank you, Gabriel I an using code that I wrote using threads.deferToThread(), I wondered if it was blocking the reactor maybe, so I tried just using a plain deferToThread() call. Here's the code (without all the protocol stuff, it's executed when data come in.

def loop(self): import time m = utils.createMessage("toto", "hello") while(True): print "Sending %s" % m.toXml() self.sendMessage(m) print "Sent!" time.sleep(2)

threads.deferToThread(self.loop)

the TCP no delay option is turned on like I said before, the weird thing is that I have the same thing, all my msgs get sent in big blocks. Either the TCP no delay doesn't work, or the reactor is getting tied up. I don' t get it since I'm running the blocking code in a thread, the reactor should have nothing to do.

Since you didn't include the definition of `sendMessage´, it's hard to say what's really going on here. However, my guess is that you're calling the transport.write method in a non-reactor thread, which is not allowed. If this isn't the case, including a full, minimal example in your next message would be a good idea. :) Jean-Paul

On Mon, 11 Aug 2008 20:18:43 +0200, Gabriel Rossetti wrote:

Jean-Paul Calderone wrote:

...

On Mon, 11 Aug 2008 19:13:56 +0200, Gabriel Rossetti wrote:

...

Gabriel Rossetti wrote:

...

Hello,

I'm needing to send small packets to update a progress bar. My update msgs were all being sent in big packets, so I tried setting setTcpNoDelay() in connectionMade() but it doesn't change anything, is it broken?

Thank you, Gabriel I an using code that I wrote using threads.deferToThread(), I wondered if it was blocking the reactor maybe, so I tried just using a plain deferToThread() call. Here's the code (without all the protocol stuff, it's executed when data come in.

def loop(self): import time m = utils.createMessage("toto", "hello") while(True): print "Sending %s" % m.toXml() self.sendMessage(m) print "Sent!" time.sleep(2)

threads.deferToThread(self.loop)

the TCP no delay option is turned on like I said before, the weird thing is that I have the same thing, all my msgs get sent in big blocks. Either the TCP no delay doesn't work, or the reactor is getting tied up. I don' t get it since I'm running the blocking code in a thread, the reactor should have nothing to do.

Since you didn't include the definition of `sendMessage´, it's hard to say what's really going on here. However, my guess is that you're calling the transport.write method in a non-reactor thread, which is not allowed.

If this isn't the case, including a full, minimal example in your next message would be a good idea. :)

Jean-Paul Yes, self.sendMessage() calls transport.write. So there's no way of sending msgs from a non-reactor thread, even if it has a reference to the protocol?

There is a way. The way is to use reactor.callFromThread to schedule a function to be called in the reactor thread. eg, reactor.callFromThread(self.sendMessage) Jean-Paul

On Mon, Aug 11, 2008 at 2:18 PM, Gabriel Rossetti wrote:

Jean-Paul Calderone wrote:

...

Since you didn't include the definition of `sendMessage´, it's hard to say what's really going on here. However, my guess is that you're calling the transport.write method in a non-reactor thread, which is not allowed.

If this isn't the case, including a full, minimal example in your next message would be a good idea. :)

Jean-Paul

Yes, self.sendMessage() calls transport.write. So there's no way of sending msgs from a non-reactor thread, even if it has a reference to the protocol?

use reactor.callFromThread, which takes as an argument a function to call in the reactor thread. That function can use Twisted APIs like transport.write. Most Twisted APIs are not thread-safe, and cannot be called from a non-reactor thread. -- Christopher Armstrong International Man of Twistery http://radix.twistedmatrix.com/ http://twistedmatrix.com/ http://canonical.com/