Why does zmq.setsockopt_string complain about default 'ascii' code?

Question

I'm trying to figure out if my dev environment is somehow screwed up, since "it works on [my colleagues] computer" but not mine. Instead of tackling the 'meat' of the problem, I'm working on the first funny thing I've spotted.

I have a bit of code that doesn't make sense why one call would work, and the other not:

import sys
import zmq

if __name__ == "__main__":
    print sys.getdefaultencoding()  # Displays 'ascii'

    zContext = zmq.Context()
    zSocket = zContext.socket(zmq.SUB)

    # This works.
    zSocket.setsockopt_string( zmq.SUBSCRIBE, "Hello".decode('ascii'))

    # This fails with error... why?
    # raise TypeError("unicode strings only")
    #
    # Shouldn't the default encoding for "Hello" be ascii?
    # zSocket.setsockopt_string( zmq.SUBSCRIBE, "Hello")

    zSocket.connect( "tcp://localhost:5000")

I'm assuming for the working call to setsockopt_string, that I am passing an array of ascii characters. In the broken code, I must be sending something not ascii, but not unicode. How would I know what is getting passed to setsockopt_string?

Maybe this isn't even the questions to ask. I'm just rather confused.

Any help would be great.

Here's my environment:

python --version
Python 2.7.3
#1 SMP Debian 3.2.57-3+deb7u2 x86_64 GNU/Linux

thanks.

Answer 1

The problem encountered caused by the unicode/str value set to the setsockopt_string method, a quick fix would be:

zSocket.setsockopt_string(zmq.SUBSCRIBE, b"Hello")

This will pass bytes instead of string, or if you have variable there, then you should write it this way:

zSocket.setsockopt_string(
    zmq.SUBSCRIBE,
    bytes("Hello", encoding="latin-1")
)

This would work on both python 2 and 3