Improvements to "compose"

[ariebovenberg]

Implements #397

some remarks:

• I've also implemented __hash__, under the assumption that Compose objects are not intended to be mutated. Perhaps it is best to make this explicit, or to remove __hash__.
• I have extended the current test method for compose. Perhaps it is best to split the test function up, considering its length.
• do there need to be any additions to the docs?
• I've implemented __signature__ for python3 only. Perhaps an elegant way of supporting python 2 is to import Signature from the funcsigs module, if it is installed. This way funcsigs is an optional dependency, keeping toolz itself lightweight and dependency-less.

edit: formatting

[ariebovenberg]

I am getting a test failure on python 3.3. Not sure how to resolve this:

======================================================================
FAIL: test_inspect_args.test_introspect_builtin_modules
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/virtualenv/python3.3.6/lib/python3.3/site-packages/nose/case.py", line 198, in runTest
    self.test(*self.arg)
  File "/home/travis/build/pytoolz/toolz/toolz/tests/test_inspect_args.py", line 443, in test_introspect_builtin_modules
    raise AssertionError(message + '\n\n'.join(messages))
AssertionError: Missing introspection for the following callables:
toolz.functoolz:
    Compose

[eriknw]

@ariebovenberg, thanks for following through and taking this on! I'll make time to give this a proper review soon.

Regarding the failure in Python 3.3, my guess is it's from __wrapped__ being defined. If you remove this, does the test pass?

We should probably drop support for Python 2.6 and 3.3 anyway (heh, it's a little funny/embarrassing that we still test and support these versions). I recall these versions (and even 3.4) caused some difficulty for me in the past, which is probably the main reason this work hasn't been done yet. I think inspection has stabilized for Python 3.5-3.7.

[ariebovenberg]

@eriknw, removing __wrapped__ does fix the test on python 3.3.

Dropping support for python 2.6 and 3.3 makes sense though, as they are both officially EOL.

[eriknw]

Hey, I've been afk for a few weeks. At a glance, this PR looks pretty good. I have a lot to catch up on, but I'm looking forward to giving this a more detailed review.

[eriknw]

This is in! I took care of a merge conflict. Thanks @ariebovenberg!

[eriknw]

Quick question: why did you define __wrapped__ attribute? When will this be used, and is it even valid given that compose is generally comprised of many functions?

[ariebovenberg]

@eriknw it's so that you can use compose to make decorators, which will then have a propper __wrapped__ set.

def int_output(func):
     return compose(func, int)

@int_output
def mult(a, b):
     return a * b

It's fine to remove __wrapped__ if you don't want to support this.

[ariebovenberg]

It's probably best to either remove __wrapped__, or comment/document why it is defined.