This query doesn't add any value, because it'll be overwritten in
migration 0002 anyway. And because it isn't batched, it can take
several minutes to run on servers with hundreds of thousands to
millions of messages of history. During that time, it's in a
transaction, and thus one can't send messages, so it forces downtime.
This should help make it explicit whenever we add a new table to Zulip
that we need to correctly categorize it for whether it will be
included in the data export, or not.
This directly prevents a traceback when submessage events
arrive in the wrong order. This was probably a symptom
of not updating message.submessages for not-yet-widgeted
messages, which was fixed in an earlier commit, but we
want defensive code in case of races or other glitches, and
it's not the end of the world is somebody sees partial
survey results due to some corner case.
The "if" condition that was removed in this commit
is no longer needed, since the called code now
handles the cannot-apply-locally use case. (We
wanted the called functions to be defensive, so
they already were effectively handling the conditions
anyway, and recent commits has them returning
appropriate values and doing the right things.)
This commit makes it so that any query for
which we do a local filter leads to us
examining the full list of unread message
ids in our cache to find a potentially
unread message that passes the filter. This
can often allow us to more immediately
jump to a new narrow with an appropriately
selected message.
Fixes#9319
We want to be able to call get_first_unread_info() even
if we cannot apply a search locally. It was returning
the correct value before, but this change removes a
blueslip warning that will allow our callers to remove
some guard code in a subsequent commit.
This changes run-dev.py to ensure that we have in fact compiled
handlebars templates before running webpack, which is the right model.
Future work will likely include running the handlebars compiler from
webpack, and thus eliminating this extra process.
I only renamed references that I thought were absolutely necessary
and only if the resulting sentence structure wasn't awkward.
If the renaming resulted in awkward structure, I replaced the term
"webhook" with "integration" (but only in some very obvious cases).
Fixes#9500.
These styles aren't great, but they're hopefully
better than none at all and inspire a follow up
commit from somebody else (so it starts to look
like a "real" feature instead of feeling completely
proof-of-concept).
We want to update message.submessages for new events, even
though our couple of widgets (poll/tictactoe) that can process
"update" events currently just apply events as "deltas"
to their current data.
This does fix a subtle issue where you may get incoming
events for a message that the client knows about but which
it hasn't yet activated as a widget. Up until now,
we've rarely seen the bug that's fixed here, since it's
usually the case that as soon as we receive a message, we
widgetize it right away.
The user can now specify the value while creating a stream.
An admin can later change it via `Change stream permissions`
modal. Add is_announcement_only to subscription type text.
For some reason in my original version I was sending both
content and data to the client for submessage events,
where data === JSON.parse(content). There's no reason
to not just let the client parse it, since the client
already does it for data that comes on the original
message, and since we might eventually have non-JSON
payloads.
The server still continues to validate that the payload
is JSON, and the client will blueslip if the server
regressses and sends bad JSON for some reason.
