Add support for lambda functions

[andrewcleveland]

Adds to create_function a check if func_name is '<lambda>', and if so builds a lambda expression instead of a regular function definition. This is passed in by @wraps(f) when f is a lambda function.
Without this, create_function builds a function definition 'def <lambda>(...):' which fails to compile.
Also adds a test to verify the above is working.
Resolves #81.

(lambda: None).__name__ is used in place of '<lambda>' in case a different Python implementation uses a different name.
Coverage.py complains about the lambda: None not being run. I don't know if that's an issue or not.

[smarie]

Thanks @andrewcleveland , this is definitely helpful ! I wrote a few comments, mostly details - the principle is good.

[andrewcleveland]

Hi @smarie,
Thanks for taking the time to comment.

Per your note about __code__.co_name, I've taken a different approach and added a new co_name parameter. A lambda is created only in the case that co_name is '<lambda>' or func_impl.__code__.co_name is not valid in a def statement. If that's too much I'll change it back to the original commit with your suggestions added.

[smarie]

Thanks a lot @andrewcleveland ! I have two minor comments, mostly to improve readability/ future maintenance, see above
(by the way sorry for the delay in reviewing !)

[smarie]

Hi @andrewcleveland , would you mind updating your PR according to the two comments above, so that we can merge ? Users of pytest-cases will be happy about your mod :) (see smarie/python-pytest-cases#278)

What is missing is just code comments in two sections.
Thanks in advance !

[andrewcleveland]

Hi @smarie, sorry for the delay. I did some work on this about a month ago and then kind of put it on the back burner.
I think I've made the comment changes you requested but I may have missed something.
I also added the co_name parameter to some of the other functions for consistency.

[smarie]

Thanks @andrewcleveland ! I had a few other comments, hopefully easy to cope with

[andrewcleveland]

Hi @smarie, I made the changes you suggested.
I changed the adding a space to func_signature_str to a check that uses a different body string for a lambda with/without arguments. It seems a bit easier to read.

Moving the conditional is_identifier import to the start of main.py seems to create a circular dependency.
The following error happens:

.nox/tests-2-7/lib/python2.7/site-packages/makefun/main.py:22: in <module>
     from makefun._main_legacy_py import is_identifier
 .nox/tests-2-7/lib/python2.7/site-packages/makefun/_main_legacy_py.py:8: in <module>
     from makefun.main import wraps
 E   ImportError: cannot import name wraps

I moved the import to the bottom of main.py, but I'm not sure if that's the best way to resolve it.

[smarie]

Thanks @andrewcleveland !
Concerning the is_identifier, can you do one of the two options below:

• option1: move the def is_identifier back to the main.py, in the else branch (basically replace the from makefun._legacy import is_identifier by the def
• option 2: in the legacy module, move the from makefun.main import wraps into the body of the function where it is used. I like this option less, to be honest

that way, the if/else can be moved to the top just after the imports, and there is no circular dependency anymore.

Thanks!

[andrewcleveland]

@smarie, option 1 is done.

[smarie]

All set, thanks a lot @andrewcleveland for your patience !
I'll tag and release asap