| Commit message (Collapse) | Author | Lines |
|---|
|
Useful for those that run redis with docker-compose but not the bot.
The bot on the host won't have access to the Docker network in such
case so the port must be exposed.
|
|
|
|
This is a simple validation that only check the type of the collection.
It does not validate the types inside the collection because that has
proven to be quite complex.
|
|
Note that `Optional[x]` is just an alias for `Union[None, x]` so this
effectively supports `Optional` too.
This was especially troublesome because the redis password must be
unset/None in order to avoid authentication, but the test would complain
that `None` isn't a `str`. Setting to an empty string would pass the
test but then make redis authenticate and fail.
|
|
Changed a RuntimeError to a KeyError (thanks @MarkKoz), and also added
some tests to ensure that the right errors are raised whenever this
method is used incorrectly.
|
|
Without this option, all contributors would need to set up a Redis
server in order to run the bot. But with use_fakeredis set to True, this
is no longer necessary because it will just set up a fakeredis redis
pool instead of trying to contact an actual server.
This is more than good enough for most local testing purposes, since
data persistence across restarts isn't really relevant for them.
This also means we need to move fakeredis into our real dependency list
instead of having it as a dev dependency, so there's a minor change for
that as well.
I also made a small kaizen change to sort all the dependencies in the
Pipfile alphabetically.
|
|
Forgot to update the additional_spec_asyncs when changing the name of
this Bot attribute to be public.
|
|
Mostly trace and exception logging.
|
|
It's a public attribute, we're accessing it from RedisCache.
|
|
Sometimes, we just want to store a counter in the cache. In this case,
it is convenient to have a single method that will allow us to increment
or decrement this counter.
These methods allow you to decrement or increment floats and integers by
an specified amount. By default, it'll increment or decrement by 1.
Since this involves several API requests, we create an asyncio.Lock so
that we don't end up with race conditions.
|
|
There really was no compelling reason why this method should return an
AsyncIterator or than that `async for items in cache.items()` has nice
readability, but there were a few concerns. One is a concern about race
conditions raised by @SebastiaanZ, and @MarkKoz raised a concern that it
was misleading to have an AsyncIterator that only "pretended" to be
lazy.
To address these concerns, I've refactored it to return a regular
ItemsView instead.
I also improved the docstring, and fixed the relevant tests.
|
|
We provide suggestions for how to solve these problems now.
|
|
Thanks to @MarkKoz for suggesting most of these in their code review.
|
|
It's leaner to just move that code out of the class and up to the module
level as constants.
This commit also renames ValidRedisType to RedisType.
|
|
Co-authored-by: Mark <[email protected]>
|
|
Instead of relying on __str__ representation,
we'll use the __name__ dunder.
Co-authored-by: Mark <[email protected]>
|
|
|
|
It's not feasible to mock it because all the commands return futures
rather than being coroutines, so they cannot automatically be turned
into AsyncMocks. Furthermore, no code should ever use the redis session
directly besides RedisCache. Since the tests for RedisCache already use
fakeredis, there's no use in trying to mock redis in MockBot.
|
|
Thanks to @kwzrd for this idea, basically we're making a constant with
the typestring prefixes and iterating that in all our converters.
These converter functions will also now raise TypeErrors if we try
to convert something that isn't in this constants list.
I've also added a new test that tests this functionality.
|
|
Previously we would try to .delete keys that did not exist if a default
was provided when calling .pop. This is okay to do (because HDEL will
just ignore any attempts to delete non-existing calls), but it does add
an additional pointless API call to Redis, so I've added some validation
as a small optimization.
This also adds a few additional lines of documentation as requested by
@SebastiaanZ in their review.
|
|
|
|
This was incorrectly suggesting the user needed to
create an instance of RedisCache, when in fact it is
the parent that needs to be instantiated.
Co-authored-by: Sebastiaan Zeeff <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
This commit just alters existing code to work with the new interface,
and with async. All tests are passing successfully.
|
|
- All methods will now do a validation check
- Complete interface spec added to class:
- .update
- .clear
- .pop
- .to_dict
- .length
- .contains
- .delete
- .get
- .set
|
|
The .set and .get will accept ints, floats, and strings. These will be
converted into "typestrings", which is basically just a simple format
that's been invented for this object.
For example, an int looks like `b"i|2423"`. Note how it is still stored
as a bytestring (like everything in Redis), but because of this prefix
we are able to coerce it into the type we want on the way out of the db.
|
|
This will help catch anything that tries to get/set an attribute/method
which doesn't exist. It'll also catch missing/too many parameters being
passed to methods.
|
|
Because some of the redis pool/connection methods return futures rather
than being coroutines, the redis pool had to be mocked using the
CustomMockMixin so it could take advantage of `additional_spec_asyncs`
to use AsyncMocks for these future-returning methods.
|
|
The fix is to mock the loop and pass it to the Bot. It will then set
it as `self.loop` rather than trying to get an event loop from asyncio.
The `create_task` patch has been moved to this loop mock rather than
being done in MockBot to ensure that it applies to anything calling it
when instantiating the Bot.
|
|
I'm not sure how it even managed to work before. It was calling the
`post` coroutine (without specifying a URL) and then changing
`__aenter__`. Now, a separate mock is created for the context manager
and the `post` simply returns that mocked context manager.
|
|
The assertion wasn't using the assertion method. Furthermore, it was
testing a non-existent function `create_loop` rather than `create_task`.
|
|
|
|
Co-authored-by: Mark <[email protected]>
|
|
This global connection is the one we will be using in RedisCache to
power all our commands.
This also ensures that connection is closed when the bot starts its
shutdown process.
|
|
|
|
In production, we will need this password to make a connection to Redis.
|
|
Co-authored-by: Joseph Banks <[email protected]>
|
|
Co-authored-by: Joseph Banks <[email protected]>
|
|
|
|
publish can take place
|
|
In production, we will need this password to make a connection to Redis.
|
|
The `discord-py` package is no longer the official release, and so
making this change silences some warnings about deprecation.
|
|
|
|
|
|
As @mathsman5133 pointed out, it's better to use the `Command`-instance
we typically already have in the current context than to rely on parsing
the qualified name again.
The invocation is now done as: `await ctx.send_help(ctx.command)`
|
|
|
MarkKoz
Leon Sandøy
Leon Sandøy
ks129
S. Co1
Joseph Banks
decorator-factory
Sebastiaan Zeeff