Remove cython

[ceball]

My opinion about #166. Note that the only people to have been getting cythonized param would be those installing from pip while also having cython and a functioning c compiler.

• Closes python setup.py develop is silently broken for param #161 (makes it redundant).
• Hopefully allows noarch conda packages, making Clean up release procedure #167 simpler.

If merged, I would open an issue about considering cythonizing once:

• there's higher test coverage of param (e.g. resurrecting topographica's tests picked up a problem with cythonized param)
• there are performance tests

Meanwhile, it might be useful to cythonize param as part of testing (e.g. can be useful to catch certain classes of error/bad practice). We could maybe add a command/argument to setup.py to optionally cythonize param.

[coveralls]

Coverage remained the same at 74.904% when pulling ac4287e on remove_cython into 486a935 on master.

[ceball]

Note that the only people to have been getting cythonized param would be those installing from pip while also having cython and a functioning c compiler.

Or people installing from http://www.lfd.uci.edu/~gohlke/pythonlibs/#param

[philippjfr]

Looks fine to me. We should go back and look at what benefits cythonizing actually provides. I think it may be worth it in some cases but the current situation where certain installs get the cythonized version and others don't is clearly not right.

resurrecting topographica's tests picked up a problem with cythonized param

Could you remind me what that issue was?

[jlstevens]

I think I approve of this change as well. Maybe @jbednar would like to merge?

[jbednar]

I guess I'll go ahead and merge. I generally think that if there are performance problems caused by param then people are not using param the way it is intended, and this should simplify things overall.

[ceball]

@philippjfr:

Could you remind me what that issue was?

#166 (comment)