Add hadoop snappy format support alongside with framing format

[aleks-v-k]

See #61

[martindurant]

Thank you for the PR, @aleks-v-k . I won't have the time immediately to look through this, I hope others chime in. Am I right in thinking that the intent is to apply decompress on any data/stream, and detect the framing format in order to call the right method?

[aleks-v-k]

Am I right in thinking that the intent is to apply decompress on any data/stream, and detect the framing format in order to call the right method?

No, the patch doesn't introduce any format detection. The default format is the original framing format implemented in this python-snappy library. The patch just adds an optional support for hadoop snappy format. To use it in (de)compression a user should specify it explicitly:

• -t hadoop_snappy option for (de)comressing via python -m snappy
• call snappy.hadoop_stream_compress or snappy.hadoop_stream_decompress inside a python script

[martindurant]

OK, I guess that's a safe way to proceed. Is there a way to determine the framing?

[aleks-v-k]

I think there it is a way. At least it's implemented in snzip project. Ok, if you will this (format detection) to be implemented, then I will do it till end of this week.

[aleks-v-k]

I've implemented format autodetection, please review the last commit in the PR.

[martindurant]

Nice tests in test_formats.py.
I'll merge this now, thank you for the work.

[aleks-v-k]

Thanks!