twisted python separation of data between protocols

Question

According to twisted documentation a new Protocol instance is created every time a connection is made however i am observing some sort of data sharing between the two.

Briefly I have defined a class that each protocol will use for state

class JSONCollector():
    char_buffer = StringIO.StringIO()
    ...    

    def process_data(self, data):
        ...       
        self.char_buffer.write(char)

The protocol instantiates it

class JSONProtocol(protocol.Protocol):
    def __init__(self):
        self.json_collector = JSONCollector()

    def dataReceived(self, data):
        self.json_collector.process_data(data)
        self.transport.write(str(self))

However each connection seems to get the same instance of JSONCollector, when I added the following statements to dataReceived self.transport.write(str(self.json_collector.char_buffer)) self.transport.write(str(self))

I get the following:

connection 1: StringIO.StringIO instance at 0x968ae2c>< main .JSONProtocol instance at 0x969036c>

connection 2: StringIO.StringIO instance at 0x968ae2c>< main .JSONProtocol instance at 0x969068c>

Also everytime I type in text, text that was typed in from other connection(s) gets displayed. So it seems that for some strange reason StringIO() instances are shared, am I missing something. I suppose I can use a factory to separate buffers by addr and make sure each Protocol only uses its own buffer, but without having the need for shared storage I would rather not jump through hoops

Thank you.

Answer 1

This:

class JSONCollector():
    char_buffer = StringIO.StringIO()

Seems like an error, and the source of your troubles. Instead, try this:

class JSONCollector():
    def __init__(self):
        self.char_buffer = StringIO.StringIO()

Otherwise, you are making one char_buffer for the entire class type, as opposed to one per instance.