* Support webhook message deletions (discord)
Messages sent via webhook can now be deleted. It seems it can do this
without any special permissions.
This copies discordgo.WebhookExecute and makes it support the returning
of discordgo.Message.
A pull request has been sent upstream, so we should use that if
@bwmariin accepts the pull request:
https://github.com/bwmarrin/discordgo/pull/663
Changes in behaviour (webhook mode only):
- Previously messages *edited* on other platforms would just be
retransmitted as a brand new message to Discord.
- Message *edits* will now be ignored.
- Debug: message edits will now print out a "permission error".
In the future it may be good to send an "message edited" react to those
webhook messages, so at least people know that the message was edited on
other platforms. (Even though it can't actually show the new message.)
Alternatively, message edits could just send a brand new message with a
link back to the old one. This is a little ugly but it would ensure that
Discord users are able to see the edited message. These "message edit
notifications" would be sent from the bot user (not from a webhook), so
we could edit the "edit notification" if subsequent edits to the
original message are made.
Currently fully support mattermost,slack and discord.
Message deleted on the bridge or received from other bridges will be
deleted.
Partially support for Gitter.
Gitter bridge will delete messages received from other bridges.
But if you delete a message on gitter, this deletion will not be sent to
other bridges (this is a gitter API limitation, it doesn't propogate edits
or deletes via the API)
Our Message type has an extra ID field which contains the message ID of the specific bridge.
The Send() function has been modified to return a msg ID (after the message to that specific
bridge has been created).
There is a lru cache of 5000 entries (message IDs). All in memory, so editing messages
will only work for messages the bot has seen.
Currently we go out from the idea that every message ID is unique, so we don't keep
the ID separate for each bridge. (we do for each gateway though)
If there's a new message from a bridge, we put that message ID in the LRU cache as key
and the []*BrMsgID as value (this slice contains the message ID's of each bridge that
received the new message)
If there's a new message and this message ID already exists in the cache, it must be
an updated message. The value from the cache gets checked for each bridge and if there
is a message ID for this bridge, the ID will be added to the Message{} sent to that
bridge. If the bridge sees that the ID isn't empty, it'll know it has to update the
message with that specific ID instead of creating a new message.
