Remove AsyncMock.assert_awaited_*

[lisroach]

BPO	38136
Nosy	@cjw296, @ezio-melotti, @voidspace, @asvetlov, @1st1, @lisroach, @mariocj89, @tirkarthi
PRs	bpo-38136: Updates await_count and call_count to be different things #16192 [3.8] bpo-38136: Updates await_count and call_count to be different things (GH-16192) #16431

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/lisroach'
closed_at = <Date 2019-09-30.04:21:49.191>
created_at = <Date 2019-09-12.12:40:20.364>
labels = ['3.8', 'type-bug', 'library', '3.9']
title = 'Remove AsyncMock.assert_awaited_*'
updated_at = <Date 2019-09-30.04:21:49.191>
user = 'https://github.com/lisroach'

bugs.python.org fields:

activity = <Date 2019-09-30.04:21:49.191>
actor = 'lisroach'
assignee = 'lisroach'
closed = True
closed_date = <Date 2019-09-30.04:21:49.191>
closer = 'lisroach'
components = ['Library (Lib)']
creation = <Date 2019-09-12.12:40:20.364>
creator = 'lisroach'
dependencies = []
files = []
hgrepos = []
issue_num = 38136
keywords = ['patch']
message_count = 6.0
messages = ['352144', '352147', '352155', '352168', '353053', '353424']
nosy_count = 8.0
nosy_names = ['cjw296', 'ezio.melotti', 'michael.foord', 'asvetlov', 'yselivanov', 'lisroach', 'mariocj89', 'xtreak']
pr_nums = ['16192', '16431']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue38136'
versions = ['Python 3.8', 'Python 3.9']

[lisroach]

After some discussion about call_count vs await_count, I believe call_count should be counting when things are *awaited* (await foo()), not when they are *called* (foo()).

I think people expect "calling" to execute the code and give them a return_value, which for asyncio is what happens when you await, not when you call with (). If people disagree about this I am open to discussion, we can change the current functionality and leave in the assert_awaited_* calls.

Currently the code does count asyncio calls when they are awaited, but this makes the assert_awaited_* calls redundant.

We should remove these in favor of the call_count_* functions.

[tirkarthi]

IMO it gives a good segregation between other mocks and AsyncMock that using assert_awaited_* makes it explicit and to cause less confusion over whether this is an awaitable or a synchronous mock. I would favor in keeping the API. I also found it to better in conceptualizing while writing tests for other PRs over having assert_call_* and assert_await_*.

[voidspace]

I'm particularly concerned that we have call_count "sane" for AsyncMock and avoid adding "await_count" if possible. I think we've decided on that.

I'm more agnostic on the assert_await* methods. I agree they're a nice API. I don't mind those staying but I also don't mind them going.

[lisroach]

Going to try to recap an in-person conversation:

There are some cases where calls are made separate from when they are awaited, for example:
   >>> call = foo()
   >>> await call

This would be 1 call and 1 await:

Call List Await List
--------- ----------
[foo] [call]

Calls like:
   >>> await foo()

Should also be counted as 1 call and 1 await, there is no difference between this call and the call above (expect for slight differences in when the lists are updated):

Call List Await List
--------- ----------
[foo] [call]

If someone were to do this:

>>>call_1 = foo()
>>>call_2 = foo()

>>> await call_1
>>> await foo(x)

We should see 2 calls added to the call list, then 1 await added to the await list, then 1 call added to the call list and 1 await added to the await list. We would end up with 3 calls and 2 awaits.

Call List Await List
--------- ----------
[foo, foo, foo] [call_1, foo]

And a call without an await:
 
  >>> call = foo()

Call List Await List
--------- ----------
[foo] []

With a setup like this, we would keep the API the same (leaving in the assert_await*). There is some risk that users will incorrectly be using assert_call* when they really want to test awaits, but we can try to make to docs as clear as possible around this.

[lisroach]

New changeset ef04851 by Lisa Roach in branch 'master':
bpo-38136: Updates await_count and call_count to be different things (GH-16192)
ef04851

[lisroach]

New changeset 52bdd41 by Lisa Roach in branch '3.8':
[3.8] bpo-38136: Updates await_count and call_count to be different things (GH-16192) (GH-16431)
52bdd41