Query words are now separated using spaces instead of commas.
Tests have been updated accordingly to reflect this change.

The model (zerver/models/navigation_view.py) defines the "name"
property as nullable, while the schema doesn't.

Previously, the /digest developer tool rendered emails using standard CSS
classes. This differed from production behavior, where the email system
uses the css_inline library to convert these classes into inline styles.

This commit updates the digest_page view to use
zerver.lib.send_email.get_inliner_instance(). It renders the email content
separately before inlining to avoid processing the full page wrapper,
which ensures the page title and head elements remain unaffected.

Fixes zulip#36800

This currently skips PRs which don't backport cleanly.

Renames `all_messages` function in `MessageListData` to
`all_messages_after_mute_filtering` to make it clear that the function
returns non-muted messages.

This is required so that the base message can be changed based
on different requirements.

Fixes zulip#34525.

If the div in question is occupying the whole row, the tooltip message
looks weirdly placed, with the tooltip not centered around the text but
instead of the div which can have plenty of empty space.

While this problem does not occur on any instances where the
tooltip_message is currently, it would be good to change the width of
settings-checkbox-wrapper to fit-content so that we don't have the empty
space problem.

We also remove the fit-content rule enforced via control-label-disabled
since it's now redundant.

In organisations with owner_full_content_access set as True, there is no
point in letting the individual users toggle `Let administrators export
my private data`.

The INTEGRATIONS dict has recently been split into multiple lists.

This ignores the resolved topic prefix while sorting
to match the existing composebox typeahead behavior.

Signed-off-by: apoorvapendse <[email protected]>

We search from all the subscribed channels if
there is no channel operator.

While listing the suggestions, priority is given to
topics that are present in the currently
narrowed channel.

Fixes: zulip#34557.

Co-authored-by: Aman Vishwakarma <[email protected]>
Signed-off-by: apoorvapendse <[email protected]>

Previously, find_first_unread_anchor used to call
get_base_query_for_search unconditionally since we wanted to set
need_user_message to True and assumed that since we are just fetching
the query, there should be no extra calls to doing that.

But if need_user_message is True, we call
get_recursive_membership_groups, which is one query. If
need_user_message was already set to True, this is a wasteful DB call.
So this commit changes find_first_unread_anchor to only rebuild the
query in case of need_user_message being false.

The ideal thing would be to just have the CTE in the query being
returned by get_base_query_for_search. But
get_recursive_membership_groups uses a Django CTE that cannot be
reused in get_base_query_for_search since it uses SQLAlchemy. We can
replicate the logic being done by get_recursive_membership_groups for
our case, but it is not worth it to maintain another copy of
get_recursive_membership_groups just to save 1 query count.

We moved the channel search out of the scrollable area, so we
don't need to account for it.

Since we were trying to make the expanded channel visible instead
of just the channel name, this results in the narrowed channel
being moved out of view.

Fixed by scrolling to the channel name.

We were always scrolling to focus channel even if topic was
selected.

Fixed by checking if a topic is selected and scrolling to
topic in that case.

This commit updates the code in the webapp to use the
`can_create_topic_group` setting to validate the
`history_public_to_subscribers` setting when creating
a new stream or updating a stream's configuration.

The previous implementation was too destructive in a hard to control
way, by just allowing CASCADE on foreign keys to delete all related
objects, while doing `.delete()` on the `UserProfile`.
This caused bugs such as zulip#34803.

Instead of doing that, and needing a hacky dance of re-assigning the
objects we wanted to preserve to a temporary dummy `UserProfile`, we
can avoid `.delete()`-ing the original `UserProfile` at all.
Instead, we can simply overwrite the relevant fields, and `.delete()`
the specific related objects that we want in a controlled way.

Additionally, this allows us to unify the `do_delete_user` and
`do_delete_user_preserving_messages` functions to share the same core
logic.

Fixes: zulip#35958

We will be introducing additional commits to flag browsers we won't
provide support for, but are exactly in the same category as banned
browsers like Internet Explorer.

Introduced in ad45bcc.

Remove old Twitter quotes from /for/business.

This fixes several code quality issues:

- Asking for the peer_data subscriptions for our own user ID isn't a
  good way to get our own subscriptions, we have non-async functions
  that get this immediately.
- We don't need to fetch subscription objects for the other channels.
- More variables/loops than were necessary.
- Fixes the confusing string type of the channel_id variable.

Only the first issue was a functional one, in that it made the new
code conflict with the peer_data changes in zulip#35864 completely
unnecessarily.