Support for async functions with mock_callable()

[fornellas]

Support for async functions with mock_callable().

Here's a half working patch for that, as a starting point.

To prove that any solution works, all relevant tests at tests/mock_callable_testslide.py must be updated to to cover async cases (eg: async function at module, async instance method, async class method etc).

PS: previous PR #8

[david-caro]

Hi @fornellas, I have a question about this, in order to support asyncio, the whole module would have to drop python2 support right?

[fornellas]

I suppose you're correct. Python 2 is dying anyway, it's support can be dropped at this point.

[fornellas]

PS: I created a milestone for async stuff https://github.com/facebookincubator/TestSlide/milestone/1, as there's more stuff to implement.

[fornellas]

@fried, what do you think about the approach of using inspect.iscoroutinefunction to detect if the callable (module function, instance/class/static method) is async, and keep the self.mock_callable( interface? The other option would be to have self.mock_async_callable(. I understand that inspect has its limitations, I'm not experience with async to judge if this might lead to users shooting themselves in the foot in those cases.

[jivid]

I've put up #60 to provide mock_async_callable. I do believe having an explicit mock_async_callable is better than having the mock_callable interface detect whether it's received a coroutine or a regular function. The latter just feels like it would hide too much from the users.

[fried]

unittest.IsolatedAsyncioTestCase in 3.8 uses that. So to have async test methods they have to explicitly be coroutines. No sync methods that return awaitables, since that's pretty hairy.
Also the AsyncMock in 3.8 is also a really good interface for async mocks, in that it looks at the spec and if its a coroutine the mocks that return when that is called is another AsyncMock, otherwise you get a normal MagicMock.