The previous commit introduced a bug where it was not intuitive
for the user to scroll again.
For the current narrow, new messages were fetched again only when
scrolled to the bottom as usually there are many messages displayed.
However when the edge case mentioned in the previous commit
occured, it was not very obvious that a scroll should be done
or we could already be at the bottom and could not scroll again
to trigger a fetch.
`message_viewport.at_bottom` has a relevant comment explaining
this behaviour.
The previous commit handled the rare race condition. However,
there is a possibility that the rare race condition might occur
again while we are handling the previous condition.
This commit resolves these 2 problems by performing a re-fetch
while also resetting the `expected_max_message_id` and this
approach has two benefits:
1. The reset prevents an infinite loop, if somehow the expected
max message's id gets corrupted resulting in a situation
where the server can never send an id greater than that even
after fetching.
2. Even though we stop after just one re-fetch the race condition
might recursively occur while we handle the previous race
condition. And even though the reset prevents multiple re-fetches,
we don't have the missing message problem.
This is because we treat the next race condition as a new race
condition instead of it being a continuation of the previous.
The `expected_max_message_id` gets updated again, on receiving
a new message. Thus it can again enter the `fetch_status` block
as the reset value is updated again.
If a user sends a message while the latest batch of
messages are being fetched, the new message recieved
from `server_events` gets displayed temporarily out of
order (just after the the current batch of messages)
for the current narrow.
We could just discard the new message events if we havent
recieved the last message i.e. when `found_newest` = False,
since we would recieve them on furthur fetching of that
narrow.
But this would create another bug where the new messages
sent while fetching the last batch of messages would not
get rendered. Because, `found_newest` = True and we would
no longer fetch messages for that narrow, thus the new
messages would not get fetched and are also discarded from
the events codepath.
Thus to resolve both these bugs we use the following approach:
* We do not add the new batch of messages for the current narrow
while `has_found_newest` = False.
* We store the latest message id which should be displayed at the
bottom of the narrow in `fetch_status`.
* Ideally `expected_max_message_id`'s value should be equal to the
last item's id in `MessageListData`.
* So the messages received while `has_found_newest` = False,
will be fetched later and also the `expected_max_message_id`
value gets updated.
* And after fetching the last batch where `has_found_newest` = True,
we would again fetch messages if the `expected_max_message_id` is
greater than the last message's id found on fetching by refusing to
update the server provided `has_found_newest` = True in `fetch_status`.
Another benefit of not discarding the events is that the
message gets processed not rendered i.e. we still get desktop
notifications and unread count updates.
Fixes#14017
This commit removes is_old_stream property from the stream objects
returned by the API. This property was unnecessary and is essentially
equivalent to 'stream_weekly_traffic != null'.
We compute sub.is_old_stream in stream_data.update_calculated_fields
in frontend code and it is used to check whether we have a non-null
stream_weekly_traffic or not.
Fixes#15181.
We refactor these 2 notices to match with the loading indicators,
thus they have been moved to `message_scroll.js`.
After a successful message fetch, we have logic to decide whether
we want to display the notices and also whether we want to hide
the loading indicators (which are already displayed).
We also conservatively hide the notices similar to the indicators
every time we narrow.
The only exception is that we show the history limit notice on
deactivating the narrow (visiting `home_msg_list`).
This commit makes the `loading_older_messages_indicator` similar
to the `loading_newer_messages_indicator`.
Now all the decisions about whether to show a loading indicator
will be made from the `fetch_status` API. We still hide the
indicators everytime the view is changed, as explained in the
previous commit.
As explained in 67053ff479,
multiple message fetches may be taking place at the same time.
So some other narrows / the home message list's indicator might
get shown for the current narrow.
This commit moves the updation of the indicators display logic
to the `fetch_status` API.
Now the `loading_newer_messages_indicator` gets displayed along
with the `loading_newer` = true updation for that narrow's message
list, i.e. just before we send the API request. But only if the
message list we are fetching matches with our current message list.
The same indicator is hidden similarly, along with the
`loading_older` = false updation for that narrow's message list,
i.e. just after the success response is recieved. But only if
the message list whose data we recieved matches with our current
message list.
Also the indicators are hidden everytime we activate narrow
or deactivate narrow (`home_msg_list`). And on entering
`narrow.activate` we fetch for it's messages so they get
displayed again, if need be.
This is the reason `message_scroll.hide_indicators();` was
moved to a location above `fetch_messages`.
Fixes#15374.
This is the exact same bug as observed in
02ab48a61e.
The bug is in the way we invoke `Filter.parse`.
`Filter.parse` returns a list of operators which
can contain only one 'search' term at max.
All strings with the 'search' operator present
in the query are combined to form this 'search'
term.
However on concatenating two filters we may get
two terms containing the 'search' operator. This
will lead to the search suggestions getting
generated based on only the last 'search' operator
term instead of all the terms having the 'search'
operator.
This is evident from the test change as suggestions
should be based on "s stream:of" but instead they
were based on just the latest query.
In commit 35c8dcb599 we introduced the
`_stream_params` object within filter.js but we didn't correctly
handle cases where `_stream param`s is undefined within `get_title()`,
`generate_url()` and `get_icon()`, which cause the navbar to if eg a
guest user tries to access a stream they weren't subscribed to.
This commit fixes this by:
* Adding the relevant checks
* Adding node tests that include non-existent streams.
* Adds the 'question-circle-o' icon for non-existent stream narrows.
A side note here is that "non-existent streams" fall under
"common narrows" as per our current definitions, which doesn't really
make sense but shouldn't bother us.
Fixes: #15387.
We simply pass the visible message ids to remove_and_rerender
which supports bulk delete operation.
This helps us avoid deleting messages in a loop which freezes the
UI for the duration of the loop.
Fixes#15285
This event will be used more now for guest users when moving
topic between streams (See #15277). So, instead of deleting
messages in the topic as part of different events which is
very slow and a bad UX, we now handle the messages to delete in
bulk which is a much better UX.
This commit adds the option of owner role in user role dropdown
and also takes care of the restrictions while adding/removing
owner status of the user.
This commit also handles the places where we dispaly role of
the user in UI.
The get_rendered_messages now returns a Map, that is formatted as:
{
'Verona > topic 1': ['message 1', 'message 2'],
'You and Cordelia Lear': ['private message 1']
}
We also add check_messages_sent (which was expected_messages in
casper). It now takes in a object, formatted similar to
get_rendered_messages above, instead of an array with topic and
messages.
We include all method needed from casper except
get_rendered_messages, and expected_messages. We will want to tweak
their API when we migrate them.
We also rename bunch of method here while migrating them.
The changes made here are as follows:
* We rename `show_history_limit_message` and `hide_history_limit_message`
to `show_history_limit_notice` and `hide_history_limit_notice`
respectively.
* We rename `hide_or_show_history_limit_message` to
`update_top_of_narrow_notices` as now this function is responsible
for updating the history limit notice as well the end of results
notice.
* We extract 2 functions responsible for hiding and showing the end
of results notice, similar to that of the history limit notice.
All instances of `$(".all-messages-search-caution").hide();` are
replaced with the call of `hide_end_of_results_notice` function.
The streams:all advertisement notice in search should only appear
after all results have been fetched to indicate we've gotten to the
beginning of the target feed.
The notice gets hidden at the start of `narrow.activate` and is
shown just after we've fetched an older batch of messages if the
"oldest" message has been found.
Previously it would get displayed after the first fetch which
takes place from `narrow.activate`. Thus we move this logic to
`notifications.hide_or_show_history_limit_message` which gets
called after a successful message fetch.
Since the home message view contains all the messages we are not
required to display this notice. However if it is already shown
we hide it as a part of `handle_post_narrow_deactivate_processes`.
To accomplish this we need to add `has_found_oldest` key to the
`fetch_status` API.
We also removed the `pre_scroll_cont` parameter as this was it's
only use case and is now redundant.
Apparently iamcal/emoji-data has a dedicated category for flag emojis.
And get_all_emoji_categories() in emoji_picker.js doesn't return the
Flags category, because we haven't declared that category in our emoji
data logic.
Note that the category looks quite sparse because it lacks country
flags, since we don't yet support emojis combined with a Zero Width
Joiner (ZWJ) (see #992 & #11767).
Fixes#15303.
This deduplication helps with readability.
Pass get_topic_key in recent_topic_row instead of
computing it in DOM.
Fix broken test_update_unread_count
after this change. This was a regression
which went unnoticed.
We now use the real implementation of
`stream_data` in our tests (as well as `people`),
while still just checking stubs for "heavier"
functions that we dispatch to.
Also, we use our regular blueslip helpers.
This is mostly a code move. There is a bit of
boilerplate at the top, and I just use
`assert.deepEqual` instead of `assert_same`.
I also use a little wrapper to provide
output like this:
Starting node tests...
running tests for dispatch_subs
test: add
test: peer add/remove
test: remove
test: update
test: add error handling
test: peer event error handling
One little piece of code that was obsolete
simply got deleted, not moved.
One of the goals here is to un-stub the
stream_data layer.
We extract stream_edit.rerender to make
the live-update code easier to follow.
The function should eventually be inlined,
but I want to clean up some other stuff first.
These are basically shims for some deeper refactorings.
I basically just try to make the code express the
problems more clearly:
- use stream_name instead of sub
- make early-exit more explicit
- make it clear that add_subscriber needlessly
requires a name
- make it clear we have an unnecessary loop
I also fixed some phony data in the test.
We are trying to phase out the trigger-event way
of telling modules to do something.
In this case we not only remove the indirection
of the event handler, but we also get to remove
`compose_fade` from the `ui_init` startup sequence.
This also has us update `compose_fade` outside
the loop, although that's only a theoretical
improvement, since I don't think `peer_add` events
every actually include multiple streams.
To make the dispatch tests a little flatter, I
added a one-line change to zjsunit to add
`make_stub` to `global`.
To manually test:
* have Aaron reply to Denmark (keep compose box open)
* have Iago add Hamlet to Denmark
* have Hamlet unsubscribe
The discard button stub in createSaveButtons is set same
as that of save button. This isn't desired because it hides
the save button on running `change_save_button_sate` instead
of hiding discard button. It was previously working as we weren't
testing visibility of save button anywhere.
Fixes#2665.
Regenerated by tabbott with `lint --fix` after a rebase and change in
parameters.
Note from tabbott: In a few cases, this converts technical debt in the
form of unsorted imports into different technical debt in the form of
our largest files having very long, ugly import sequences at the
start. I expect this change will increase pressure for us to split
those files, which isn't a bad thing.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
Previously, the navbar failed at managing the searchbox text state in
cases where, eg, the user performs navigation by browser history.
This commit resolves the issue by ensuring that the searchbox text is
only (and always) set when the searchbox is made visible, and as such
there is no "state" to manage and we will always display the correct
text.
It also adds a test in `search_legacy.js` to make sure that the search
text is placed as intended.
Fixes: #14771.
This combines `upload_realm_logo` and `upload_realm_icon` into single
function called `upload_realm_logo_or_icon`. The functions wer near
duplicates.
Additional refactoring should be able to deduplicate the logic further.
Part of #14730.
We store the relevant data to hide/show a topic in the row itself,
and use jquery to hide/show it on filter change.
This also fixes search breaking the set filters.
This fixes the bug that message was undefined since we used to store
locally echoed message id and were not updating it after new message
id for the same message was received from the server.
We reuse the existing logic for displaying and updating stream color
from the stream left sidebar.
Tests fixtures were extracted and updated for this commit.
The approach that supposed to use this data was not implemented
and hence this data will no longer be used.
If this feature is implemented in future,
this data will still not be used since we would depend upon
starred_messages.js library to provide us the required information.
* Add action to mute topics.
* We don't need to store muted data per topic as previously planned.
* Moved launch topic test to the top so that they run on non-modified
data.
* Show an empty overlay of recent topics.
* Register click event to open recent topics.
* Launch recent topics on "t" keypress.
This is based on the draft overlay.
This was previously hardcoded with agreement between the Zulip backend
and frontend as 86400 seconds (1 day). Now, it's still hardcoded in
the backend, but arranged in a way where we could add a setting
without any changes to the mobile and terminal apps to update logic.
Fixes#15278.
Change user avatar spinner implementation to match
realm icon spinner implementation and have common css class
since similar implementation between similar widgets may help
in future deduplication.
In the past, the anchor message has always been the same as the
pointer, but we're about to change that as part of removing the
pointer entirely.
Using the anchor is logically what we meant, anyway, since we always
want to select a message that's actually within the range we just
fetched.
This was implemented in 2012 to avoid showing a loading indicator for
fetching messages for users with no message history. However, the
Zulip onboarding UI always creates some message history, and fetching
history is fast, so this is likely clutter more than a useful
optimization.
Generated by pyupgrade --py36-plus --keep-percent-format, but with the
NamedTuple changes reverted (see commit
ba7906a3c6, #15132).
Signed-off-by: Anders Kaseorg <anders@zulip.com>
The activity.process_loaded_messages code path was called when these
tests were originally written in f8e0137. We stopped calling that
code path in 43e5b2d (#15118). This assert test code is no longer
relevant; tested by adding console.log in the function. I came
across this when working on removing activity from the window.
Waiting till DOMContentLoaded event is triggered helps avoid flakes since
puppeteer is very fast and starts doing another task before
everything on the page is loaded. Adding this to log_in function
as almost all tests depend on this which leads to flaky tests if
the other parts of tests just start without even the page being
loaded.
One example this commit helps is for the test `02-site.js`
which is dependent on a function that runs some jquery on the site.
But because of the page not being loaded, we miss jquery and thus
the test fails. `02-site.js` and related code is added in the next
commit.
Co-authored-by: Priyank Patel <priyankp390@gmail.com>
When no credentials are sent to `log_in` function we want to
use the default generated test credentials. This saves us the
work of importing test_credentials everytime we run this function
in a different test which doesn't focus on what credentials are sent
to login.
If typeahead is used, this adds comma separated search queries
so that multiple search pills don't get combined as one and the
search behaviour remains same as search_pills_enabled = False case.
If typeahead is not used, this prevent the typing of a single comma
after the pill gets created.
Now node_tests/search.js and node_tests/search_legacy.js have
almost identical test besides the search pills related data
processing work.
This is possible since we no longer depend on the values stored
in the input pills and can narrow or search for term based on
the `#search_query` box value.
This was done in commit 02ab48a61e.
This commit removes the 'get_active_user_for_email' function
from people.js. We have removed the use of this function
in the previous commits, which changed the functions using
'get_active_user_for_email' to use user_ids instead of emails.
This commit changes the would_receive_message to use user_id
instead of emails.
This change is done because user_ids are immutable and using
user_ids is the correct way of uniquely identifying user.
The change in 'would_receive_message' also leads to change
in util.is_pm_recipient to use a string of user_ids instead
of emails.
We also know that user_ids passed to 'would_receive_message'
are active user_ids, since we get them from buddy_list.
So we don't need to check whether the user is active, which
was previously being checked by get_active_user_for_email.
This commit changes the needs_subscribe_warning function to
use user_id instead of emails.
This change is done because user_ids are immutable and using
user_ids is the correct way to uniquely identify a user.
We already know that user_ids being passed in this function are
active user_ids, since they come from typeaheads.
So, we only need to call 'people.get_by_user_id', to get the user
object from user_id and do not need to check the active status of
user, which was done previously using 'get_active_user_for_email'.
This commit changes the compose_invite_users template to use
data-user-id as property intead of data-useremail.
This is changed to maintain consistency with other parts of the
code where user_ids are used for referring to users.
This also helps in removing some of the checks for the case of
undefined emails.
We now send user_ids to the backend API for subscribing/unsubscribing
users to a stream instead of emails.
This change is done now because we have just migrated the backend API to
support sending user_ids in 2187c84, so it wasn't possible before.
This change is helpful because sending user_ids is more robust, as those
are an immutable reference to a user, rather than something that can
change with time.
The get_active_humans and get_non_active_humans functions used
to return a list of user objects. The get_active_humans is used
on settings_users.js and settings_bots.js, and in both places the
only attributes needed of the person object are the user_id and
full_name.
To make the function return smaller, instead of a list of active
humans, we are returning a list of active human ids, saving memory.
With the ids we can call the people API to get the full_name attribute.
This is done to avoid flakes where our code for focusing first input
interferes with puppeteer's typing. More details in comments.
Thanks Dinesh for testing this out earlier as part finding a solution
to this flake.
Co-authored-by: Dinesh <chdinesh1089@gmail.com>
This reimplements our Zoom video call integration to use an OAuth
application. In addition to providing a cleaner setup experience,
especially on zulipchat.com where the server administrators can have
done the app registration already, it also fixes the limitation of the
previous integration that it could only have one call active at a time
when set up with typical Zoom API keys.
Fixes#11672.
Co-authored-by: Marco Burstein <marco@marco.how>
Co-authored-by: Tim Abbott <tabbott@zulipchat.com>
Signed-off-by: Anders Kaseorg <andersk@mit.edu>
This commit fixes the bug for subscribing the user from mention
warning which was introduced in e52b544.
This is fixed by changing email to be passed as list to
'invite_user_to_stream'.
The reason for this change is that, this is where `Filter` and
actual tracking of what messages are contiguous lives. This
will be beneficial when we will to move to a model where we
cache `MessageListData` objects for a large number of views.
This commit changes the person dict in event sent by do_change_user_role
to send role instead of is_admin or is_guest.
This makes things much more straightforward for our upcoming primary
owners feature.
Restored old behavior accidentally removed in
1ae07b93d8 (diff-e353fab8bea58b8746ec68c83aa39b36L48)
The server only remembers the most recent presence status update per
device. Meaning that, for instance, if the user only uses one client and
that client's last status update was IDLE, then the server only knows
that, doesn't know anything about the user's last ACTIVE time. Thus the
"active_timestamp" the server will serve about this user to the webapp
will be "undefined".
The old behavior was that for the sake of the "Last active: x ago"
status in buddy list popover, the latest status timestamp was used,
whether IDLE or ACTIVE.
The change linked about changed that to only pay attention to
ACTIVE. Thus, if the server doesn't remember any ACTIVE statuses, webapp
would show "Last active: More than 2 weeks ago", which was incorrect.
We restore the old behavior and further improvements can be made on top
of this.
Previously, we had to fiddle with the generated HTML to update
individual values. Now, we can simply ask the widget to rerender
the row that we updated.
This is done by passing an html_selector function that returns
a selector for the rendered item.
If:
- we do not provide html_selector function
- item is not currently rendered
- new html is not a string.
then the render_item() call is a noop.
We remove the "GROUP PMs" section that used
to be in the lower right sidebar.
Most of this is straightforward code removal.
A couple quick notes:
- The message fetching code now just
calls `huddle_data.process_loaded_messages`,
which we still need for search suggestions.
We removed `activity.process_loaded_messages`.
- The `huddle_data.process_loaded_messages`
function no longer needs to return `need_resize`.
- In `resize.js` we now just calculate
`res.buddy_list_wrapper_max_height` directly
from `usable_height`.
We had a bunch of places where we
were calling `resize.resize_bottom_whitespace`
with no arguments, which has been a no-op
since the below commit that removed support
for our `autoscroll_forever` option:
fa44d2ea69
With the `autoscroll_forever` options things
like opening/closing the compose box could
alter how much bottom whitespace you'd want,
but we stopped supporting that feature in
2017.
Since then bottom_whitespace has just always
been 40% of the viewport size. So we only need
to change it on actual resize events.
It's worth noting that we still call
`resize_bottom_whitespace` indirectly in many
places, via `resize_page_components`, and
the latter actually causes
`resize_bottom_whitespace` to do real work,
but that work is redundant for most of those
codepaths, since they're not triggered by
changes to the viewport. So there are other
opportunities for cleanup.
Trigger realm icon upload by clicking on realm icon element itself
rather than having a big upload button and to match our user avatar UI.
Added new spinner over the icon element itself to show while
uploading realm icon.
Since we are doing a assert call checking a url without no page.waitForSelector
or similar calls, we add this to avoid flakes where the assert is called
when we are in process of navigation to the page. Also, we use Promise.all
here, and call page.waitForNavigation first, to avoid race conditions where
form is submitted quickly and navigation occurs and then we are stuck waiting
for a navigation. There were no flakes in the CI related to this issue, but
I expect someone with a slow hardware would probably stumble upto this in future.
We were missing the step for clicking the checkbox to show the subdomain
input field. There was a flake because of this issue in CI where the
"testsubdomain" input was getting typed into the name field. This fixes
that flake.
There were two reason there were flakes in login tests. One problem was
that the assert call for checking that the login page url was happening
too early in rare cases. To fix this issue we wait for email input field
element in the login page. The second issue was the sometimes the url was
/login/ instead of /accounts/login/. The time between the redirect were
super close here too (around ~10ms) so I am suprised this didn't occur flake
locally more often.
Server logs that show the redirect:
2020-05-26 18:55:27.065 INFO [zr] 127.0.0.1 GET 200 36ms (db: 5ms/2q) (+start: 14ms) /accounts/login/ (unauth@zulip via ?)
2020-05-26 18:55:27.074 INFO [zr] 127.0.0.1 GET 200 28ms (db: 3ms/2q) (+start: 23ms) /login/ (unauth@zulip via ?)
To take a screenshot on failure where we have our common error
handling code in common.run_test method we need to have access to
the Page instance. Currently, we create a new Page in the test method
we pass into run_test, so we pass in a new Page instance to that method
so the test and the run_test method have access to the same Page. And,
then we take a screenshot on failure. It will be saved as `failure-${num}`
in var/puppeteer directory.
This is mostly just moving code from the `activity.js`
tests, but I also now explicitly cover the "100%"
use case (i.e. all four folks in the huddle are present).
Computed indexes into these raw objects should be guarded with
Object.prototype.hasOwnProperty; make our accessors do this
automatically and use them consistently.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
This is helpful because if the user pastes multiple queries in the
searchbox and there are invalid search operators, then it is visible
through the typeahead.
Ryan Rehman
sahil839
YashRE42
Tim Abbott
Aman Agrawal
Dinesh
Priyank Patel
Gittenburg
Steve Howell
Anders Kaseorg
jagansivam28
Aman Agrawal
Tim Abbott
Clara Dantas
Anders Kaseorg
Mateusz Mandera
Alex Vandiver
Rohitt Vashishtha