Removed unicode encoding for strings that might not be unicode

[njo]

Reverted where the code was attempting to unicode encode a value pulled from redis. The to_unicode would cause an exception to be thrown if there was a multibulk reply (eg. HMGET myhash field1 field2) and the values stored in redis were not unicode data - ie. gzip data.

[leporo]

Please check out the unit-test output: https://travis-ci.org/leporo/tornado-redis/builds/14397546
I can't merge your change at the moment.

BTW, tornado-redis' performance with MGET/MSET commands is really poor. May I suggest you to try using redis-py (synchronous client) in your application for MGET/MSET commands? Please benchmark your code to be sure it performs better. Feel free to ask me if you need any assistance.

See the issue #36 for details on poor MGET/MSET performance.

I think we may get far better performance by using the hiredis parser for reply processing. There are just too many callbacks in multi-bulk reply handling.

[gmr]

The unicode bit is biting me with binary data as well. Any chance we can see this addressed? @leporo is the main issue the tests needing updates?

[leporo]

The main issue with the proposed fix is that it may ruin existing applications.

I'll review client's code when I have spare time.

[rudolphfroger]

Please fix this, putting binary data into Redis is quite common I believe (like struct packed integers, hash digests etc). Contact me if there's anything I can do to help with this.

[zjjott]

+1