Due to a quirk in InteractionResponse.send_message not returning a
message, all messages sent with an associated View would end up having
no message_id set. When multiple instances of a View are responded to
in a slash command context, this meant that the newest one would
override the storage of the older one. Ultimately leading to the first
view instance causing interaction failures.
Since fetching the original message is an unacceptable solution to the
problem due to incurred requests, the next best thing is to store an
intermediate interaction_id as a stop gap to differentiate between
the multiple instances. This change however, came with its own set of
complications.
Due to the interaction_id being an intermediate stop gap, the
underlying storage of the view store had to be changed to accommodate
the different way of accessing the data. Mainly, the interaction_id
key had to be quick to swap and remove to another key. This solution
attempts to change the interaction_id interim key with
a full fledged message_id key when it receives one if it's possible.
Note that the only way to obtain the interaction_id back from the
component interaction is to retrieve it from the MessageInteraction
data structure. This is because using the interaction_id of the button
press would be a different interaction ID than the one set as an
interim key. As a consequence, this stop gap only works for application
command based interactions. I am not aware of this bug manifesting in
component based interactions.
This patch also fixes a bug with ViewStore.remove_view not working due
to a bug being suppressed by a type: ignore comment. It also removes
the older __verify_integrity helper method since clean-up is already
done through View.stop() or View timeout.
Hopefully in the near future, the `/callback` endpoint will return
actual message data and this stop gap fix will no longer be necessary.
This results in poor ergonomics due to the lack of default generics
for the common case. For most users this ends up in a degraded
experience since the type will resolve to Unknown rather than at the
very least a Client.
Segments where readability was hampered were fixed by appropriate
format skipping directives. New code should hopefully be black
compatible. The moment they remove the -S option is probably the moment
I stop using black though.
The first issue involved copied users which would lead to user updates
causing faster evictions of the cache than was expected.
The second issue involved users that weren't bound to an internal
lifetime eviction policy. These users would not get evicted.
For example, a user without mutual guilds or being part of the internal
cache in general (messages, DMs) would never end up being evicted for
some strange reason. To handle this case, store_user would get a
counterpart named create_user which would create a user without
potentially storing them in the cache. That way only users with a
bound lifetime within the library would be stored.
