| Commit message (Collapse) | Author | Age | Lines |
|---|
| | |
|
| |
|
|
|
| |
Co-authored-by: Boris Muratov <[email protected]>
Co-authored-by: wookie184 <[email protected]>
|
| | |
|
| |
|
|
|
|
|
| |
region was removed from the guild object, so this has been replaced with features
add_cog is now async, so it is now an async_mock during tests
Two new required voice_channel attrs were added
channel.type is required to be set to ChannelType due to a new isinstance check in d.py
|
| |
|
|
| |
This reverts commit 960619c23300c56c8aaa454edc7241e2badf80ad.
|
| |
|
|
| |
All of the tag content is out of scope for this PR.
|
| |
|
|
|
|
| |
Python 3.8 introduced an `unittest.mock.AsyncMock` class that can be used to mock coroutines and other types of asynchronous operations like async iterators and async context managers. As we were using our custom, but limited, AsyncMock, I have replaced our mock with unittest's AsyncMock.
Since Python 3.8 also introduces a different way of automatically detecting which attributes should be mocked with an AsyncMock, I've changed our CustomMockMixin to use this new method as well. Together with a couple other small changes, this means that our Custom Mocks now use a lazy method of detecting coroutine attributes, which significantly speeds up the test suite.
|
| |
|
|
| |
Since we upgraded to Python 3.8, we can now use the new IsolatedAsyncioTestCase test class to use coroutine-based test methods instead of our own, custom async_test decorator. I have changed the base class for all of our test classes that use coroutine-based test methods and removed the now obsolete decorator from our helpers.
|
| |
|
|
|
|
|
| |
Previously, the coroutine object passed to `MockBot.loop.create_task`
would trigger a `RuntimeWarning` for not being awaited as we do not
actually create a task for it. To prevent these warnings, coroutine
objects passed will now automatically be closed.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Our custom `discord.py` now follow the specifications of the object
they are mocking more strictly by using the `spec_set` instead of the
`spec` kwarg to initialize the specifications. This means that trying
to set an attribute that does not follow the specifications will now
also result in an `AttributeError`.
To make sure we are not trying to set illegal attributes during the
default initialization of the mock objects, I've changed the way we
handle default values of parameters. This does introduce a breaking
change: Instead of passing a `suffix_id`, the `id` attribute should
now be passed using the exact name. `id`.
This commit also makes sure existing tests follow this change.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I have enhanced the custom mocks defined in `tests/helpers.py` in a
couple of important ways.
1. Automatically create AsyncMock attributes using `inspect`
Our previous approach, hard-coding AsynckMock attributes for all the
coroutine function methods defined for the class we are trying to
mock is prone to human error and not resilient against changes
introduced in updates of the library we are using.
Instead, I have now created a helper method in our `CustomMockMixin`
(formerly `GetChildMockMixin`) that automatically inspects the spec
instance we've passed for `coroutine functions` using the `inspect`
module. It then sets the according attributes with instances of the
AsyncMock class.
There is one caveat: `discord.py` very rarely defines regular methods
that return a coroutine object. Since the returned coroutine should
still be awaited, these regular methods should also be mocked with an
AsyncMock. However, since they are regular methods, `inspect` does
not detect them and they have to be added manually. (The only case of
this I've found so far is `Client.wait_for`.)
2. Properly set special attributes using `kwargs.get`
As we want attributes that point to other discord.py objects to use
our custom mocks (.e.g, `Message.author` should use `MockMember`),
the `__init__` method of our custom mocks make sure to correctly
instantiate these attributes.
However, the way we previously did that means we can't instantiate
the custom mock with a mock instance we provide, since this special
instantiation would overwrite the custom object we'd passed. I've
solved this by using `kwargs.get`, with a new mock as the default
value. This makes sure we only create a new mock if we didn't pass
a custom one:
```py
class MockMesseage:
def __init__(self, **kwargs):
self.author = kwargs.get('author', MockMember())
```
As you can see, we will only create a new MockMember if we did not
pass an `author` argument.
3. Factoring out duplicate lines
Since our `CustomMockMixin` is a parent to all of our custom mock
types, it makes sense to use it to factor out common code of all of
our custom mocks.
I've made the following changes:
- Set a default child mock type in the mixin.
- Create an `__init__` that takes care of the `inspect` of point 1
This means we won't have to repeat this in all of the child classes.
4. Three new Mock types: Emoji, PartialEmoji, and Reaction
I have added three more custom mocks:
- MockEmoji
- MockPartialEmoji
- MockReaction
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- https://docs.python.org/3/library/unittest.mock.html
We previously used an override of the `__new__` method to prevent our
custom mock types from instantiating their children with their own
type instead of a general mock type like `MagicMock` or `Mock`.
As it turns out, the Python documentation suggests another method of
doing this that does not involve overriding `__new__`. This commit
implements this new method to make sure we're using the idiomatic way
of handling this.
The suggested method is overriding the `_get_child_mock` method in
the subclass. To make our code DRY, I've created a mixin that should
come BEFORE the mock type we're subclassing in the MRO.
---
In addition, I have also added this new mixin to our `AsyncMock`
class to make sure that its `__call__` method returns a proper mock
object after it has been awaited. This makes sure that subsequent
attribute access on the returned object is mocked as expected.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit introduces some new Mock-types to the already existing
Mock-types for discord.py objects. The total list is now:
- MockGuild
- MockRole
- MockMember
- MockBot
- MockContext
- MockTextChannel
- MockMessage
In addition, I've added all coroutines in the documentation for these
discord.py objects as `AsyncMock` attributes to ease testing. Tests
ensure that the attributes set for the Mocks exist for the actual
discord.py objects as well.
|
|
|
After a discussion in the core developers channel, we have decided to
migrate from `pytest` to `unittest` as the testing framework. This
commit sets up the repository to use `unittest` and migrates the
first couple of tests files to the new framework.
What I have done to migrate to `unitest`:
- Removed all `pytest` test files, since they are incompatible.
- Removed `pytest`-related dependencies from the Pipfile.
- Added `coverage.py` to the Pipfile dev-packages and relocked.
- Added convenience scripts to Pipfile for running the test suite.
- Adjust to `azure-pipelines.yml` to use `coverage.py` and `unittest`.
- Migrated four test files from `pytest` to `unittest` format.
In addition, I've added five helper Mock subclasses in `helpers.py`
and created a `TestCase` subclass in `base.py` to add an assertion
that asserts that no log records were logged within the context of
the context manager. Obviously, these new utility functions and
classes are fully tested in their respective `test_` files.
Finally, I've started with an introductory guide for writing tests
for our bot in `README.md`.
|
Xithrius
Chris Lovering
Amrou Bellalouna
Sebastiaan Zeeff