We want to scroll the left sidebar to the top as soon as the user
zooms in on a stream, and we don't want to wait for the server,
otherwise we'll get jumpiness.
This commit is a bit complicated, because we do full redraws of
the topic list frequently, and we don't want to randomly obliterate
our "No more topics found" message, so we need to keep a bit of
extra state around.
We now use a template to render the "more topics" link.
We also remove an unnecessary conditional and an unnecessary
attribute.
Finally, our unit tests are a bit more granular now.
This will make testing a bit easier (we can stub stuff before
building the widget), and it will eventually give us more control
on redrawing the topic list.
We were parameterizing max_topics, but it made the calling sequences
unnecessarily complicated. We don't ever override the value, not
even in tests, so now we just set in build_list().
This puts build_list on the widget object, which will make it a
bit easier to unit test, and it's more consistent with the rest of
the function. This also reduces the scope of the `my_stream_name`
variable and moves the initialization of `self.topic_items` into
build_list.
If a user re-narrows to another stream before our server gives
us more topic history, or they zoom out, we can avoid drawing
the topic list. Note that our data structures will still be
updated, although the only time that really matters is for
the corner case of a low-traffic stream. For a low traffic
stream that only had 3 or 4 topics in the original message
fetch, but has longer history, the next time you open the
stream in the sidebar, even when you're zoomed out, you will
see more topics.
Despite a few warts, we are going forward with getting topic
history from the server when you click "more topics." This
commit simplifies the code by removing the feature flag
checks.
Our old optimizations to prevent re-rendering of locally echoed
messages created a lot of code complexity. This commit is an
experiment to simplify the code, which it clearly does. The
danger of re-rendering messages is flicker, but our message
view has changed since the original local echo code was written.
It's kind of confusing to have a filter function that has massive
side effects. Now we just have a simple loop where we triage
some messages into non_echo_messages and do an early-exit in the
loop function. This change also introduces the more explicit
variable name of `non_echo_messages`; before we were shadowing
`messages`.
This fixes the characters like “g” and “l” from overflowing the input
bounds while maintaining the previous height of the input box itself.
Fixes: #6665.
This removes the `box-sizing` attribute on the spectrum input that was
causing the characters to overflow the bounds of the padding, which
would cut off things like the bottom of a “g” or the top of an “f”.
Fixes: #6361.
It's unlikely to be of any real consequence, but this code bugged me
in that it makes a whole set before throwing it away to make nearly
the same set.
Sadly Python's comprehensions lack a way to write these cleanly as one
comprehension; but with no extra code complexity we can make the
temporary a genexp, which does the job.
We need a migration to clear the tutorial_status for existing users,
so that we don't show hotspots to anyone who signed up for Zulip in
the month or so since we deleted the old tutorial.
The `disable_existing_loggers` option to the `logging.config` module
turns on a rather complicated behavior of disabling some, but not all,
loggers that might have been already configured when the call to
`logging.config.dictConfig` or `logging.config.fileConfig` is made:
> This behaviour is to disable any existing loggers unless they or
> their ancestors are explicitly named in the logging configuration.
(https://docs.python.org/3/library/logging.config)
Turns out the only reason this is there is as a compatibility hack to
match the behavior of Python 2.4 and below. See the thread where the
new behavior was introduced: https://bugs.python.org/issue3136
Just as the author of the new behavior explains in that thread from
2008, the legacy behavior forces all logging configuration to be
awkwardly centralized in one place. That makes the code harder to
read, and it perennially causes confusion when a perfectly
normal-looking `logging.getLogger` call at the top level of one module
mysteriously has no effect, while that in another module works fine,
under the influence of the details of what gets imported when.
So, switch to the shiny new behavior of Python 2.5. Here LOGGING is a
Django setting which just becomes an argument to logging.config.dictConfig.
This may cause a few of the logfiles in ZULIP_PATHS to become active
that have been dormant for a long time.
This is a start and fixes the most glaring problems from not updating
this documentation; I'd like Harshit to do a helpful pass on updating
this to cover some of the more subtle details about how our emoji
picker works, emoji aliases, etc.
This fixes a bug where the internal_prep_message code path would
incorrectly ignore the `realm` that was passed into it. As a result,
attempts to send messages using the system bots with this code path
would crash.
As a sidenote, we really need to make our test system consistent with
production in terms of whether the user's realm is the same as the
system realm.
We don't access any attributes of the sender other than the realm, and
as it turns out, we in some cases want to use a different realm than
the sender's.
Steve Howell
Greg Price
neiljp (Neil Pilgrim)
Tim Abbott
Brock Whittaker
Rishi Gupta
hollywoodno