tasks and client_ids should be logical strings in the public API

[lentinj]

As in Yelp#35 there are still encoding/decoding issues in this branch, this is an attempt to fix at least the ones that stops our client and worker working.

I put the encoding of the task string at the same level as encoding the data for the job, this seemed to make sense in my head but meant I had to butcher the unit tests somewhat, as they inspect the internal job structures after sending commands. Does this seem a reasonable approach to you?

The tests pass here on Python 3.4

[svisser]

Thanks for looking at this branch.

I think the main thing is the use of ASCII here. Given that we're sending bytes across and given that the Gearman protocol specification doesn't indicate what encoding should be used, it seems to me that UTF-8 could be a better choice here? This would allow us to support a broader range of identifiers. Looking into this, the OpenStack gear client also does this: https://github.com/openstack-infra/gear/blob/master/gear/__init__.py#L705.

Regarding the unit tests, I think it's fair to adapt them to make them work - we can clean / refactor them at a later date. The code has been around for a while so I'm more concerned with getting the current codebase to work in Python 3.x and letting the unit tests adequately cover the code.

[lentinj]

IIRC I couldn't see anything in the gearman protocol, and copied .encode('ascii') from elsewhere in the code. I'd be happy to go utf8 instead.

Possibly the encoding should be factored out into a constant in the source somewhere? Will have a look at that.