Some Python 3 compatibility changes made python-memcached import pickle
rather than cPickle, since the latter isn't available on Python 3:
45403325e0
This is fine on Python 3, since importing pickle will automatically use
the faster C implementation if available, however with Python 2 this
caused a 400% slowdown in memcache gets between python-memcached v1.53
and v1.54, when dealing with objects other than bytes.
This mostly fixes #71, though there is still a 10-20% performance drop
using current master plus this change compared to v1.53 (using Python
2.7.10).
* Port set_multi() to Python 3
* Port delete_multi() to Python 3
* Fix _get_server() on Python 3 when the connection to the first server
fails: encode to ASCII before calling serverHashFunction.
* Fix expect(): don't decode line on Python 3, return the raw line
* Add more unit tests
* tox now also runs unit tests
* Explicit the encoding when calling str.encode(): use 'utf-8'
* test_memcache: close sockets in tearDown()
* test_get_unknown_value(): delete the key to ensure that it doesn't exist
* travis: make python 3 tests voting (cannot fail anymore)
* setup.py: add Python 3 classifiers
* Encode unicode key to UTF-8: add _encode_key() method
* Add _encode_cmd() helper method to format a memcache command as a byte
string (bytes%args will only be supported in Python 3.5)
* Rewrite _map_and_prefix_keys() code converting keys
* _val_to_store_info() now accepts Unicode: Unicode is encoded to UTF-8
* _set('cas') doesn't call _val_to_store_info() anymore when it's not
needed: begin by checking if the key is in the cas_ids dictionary
* Process server reply as bytes
* _recv_value() now clears the _FLAG_COMPRESSED flag after decompressing
to simplify the code
* On Python 3, _recv_value() now decodes byte strings from UTF-8
* Simplify check_key(), _encode_key() now encodes Unicode to UTF-8
* Replace u'...' with six.u('...') in tests for Python 3.2
Sean Reifschneider