This is preparation for #27637, where it will be nice to be able
to be able to parse a url such as `/#streams/1/announce/general`
and use the `general` string to directly open a stream subscription
tab without converting it to `general_settings`.
date-fns-tz does not handle daylight saving time correctly, and can be
replaced with modern browser APIs.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
Subscripting an empty array does not throw an exception in JavaScript,
"default" is not a valid locale, and the browser APIs already fall
back on unrecognized locales anyway.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
Earlier, in the 'check_for_unsuspend' function, we didn't
handle the exception, if any, during callback execution,
resulting in 'watchdog_time' not always being updated.
This commit handles the exception using the try/catch block.
Fixes#27723.
At present, it's not obvious that clicking on the stream name in the
top navbar will take the user to stream settings. To make it more
apparent, we add a tooltip to the stream name, explicitly indicating its
functionality.
We also add a second line to the tooltip thar displays the number of
subscribers to the stream and remove the tooltip from the number of
subscribers indicator on the top navbar. These changes are in preparation
for removing the number of subscribers indicator from the top navbar.
Fixes#27360.
Co-authored-by: Sayam Samal <samal.sayam@gmail.com>
This commit adds a new option 'DMs, mentions, and followed topics'
to 'desktop_icon_count_display' setting.
The total unread count of DMs, mentions, and followed topics appears
in desktop sidebar and browser tab when this option is configured.
Some existing options are relabeled and renumbered. We finally have:
* All unread messages
* DMs, mentions, and followed topics
* DMs and mentions
* None
Fixes#27503.
Earlier, the send button's state was determined independently by the
message length and the upload status, and its containing `div`'s state
was separately determined on recipient change by whether the user had
permission to post in that stream, or if direct messages were allowed
in the organisation. The main problem was that as the message length
check was run on each character input, it often overrode the upload
status check.
Now for consistency, the code has been refactored to always disable the
send button by accordingly styling its containing `div`; and using the
observer pattern we update it anytime the message length, upload status,
or the recipient changes, considering **all** of them together.
(The Save button when editing a message still works like before -- only
dependent on upload status.)
Fixes: #21135.
Due to the way turndown pads every block element with 2 new lines, and
makes `br` double space by default, we would get 3 blank lines pasted
when there's just 1 line break between 2 paragraphs.
Now we set `br` to an empty string, and since turndown collapses
sequences of multiple new lines to `\n\n` (1 blank line), so any 2 block
elements will now always have 1 blank line between them, irrespective of
if and how many line breaks there are between them in the copied HTML.
This reverts commit 091e2f177b.
This version of python_to_js_linkifier fails for at least some real
linkifiers. We'll likely re-introduce this after a bit more debugging.
This commit adds new API endpoint to get stream email which is
used by the web-app as well to get the email when a user tries
to open the stream email modal.
The stream email is returned only to the users who have access
to it. Specifically for private streams only subscribed users
have access to its email. And for public streams, all non-guest
users and only subscribed guests have access to its email.
All users can access email of web-public streams.
This commit removes "email_address" field from Subscription objects
and we would instead a new endpoint in next commit to get email
address for stream with proper access check.
This change also fixes the bug where we would include email address
for the unsubscribed private stream as well when user did not have
permission to send message to the stream, and having email allowed
the unsubscribed user to send message to the stream.
Note that the unsubscribed user can still send message to the stream
if the user had noted down the email before being unsubscribed
and the stream token is not changed after unsubscribing the user.
This should now happen at the same time it did prior to this change,
without requiring the show_all_message_view wrapper to have any
business logic.
This fixes a potential scroll position bug in the event that
narrow.deactivate in fact calls itself recursively after a timeout.
We already do a very parallel construction in narrow.activate, so this
moves us towards being able to unify those code paths, while also just
being more readable by avoiding a small-but-important wrapper function
in hashchange.js.
I believe this fixes a bug where we were not saving scroll position in
browser history when navigating to "All messages" from another view.
Since at least 6ef0753a51, it's been the
case that narrow.activate already hides the inbox/recent views if
open, and the same is true for all messages.
Fixing the duplicate call is important in show_home_view, because
show_all_message_view relies on having an accurate value for whether the
recent/inbox views were already open in order to correctly update the
left sidebar.
Testing experimentally, removing the setTimeout seems to fix a visible
flicker when using Esc to navigate to "All messages" from the Inbox
view. That setTimeout has been moved around without real examination
since 5d79bb6a20 from early 2013; I
don't see any good reason why it would make be necessary only in the
"All messages" code path, and not when narrowing to any other view.
There's no good reason to have the caller of deactivate pass this
parameter in.
This effectively reverts a18b1662cb,
which did this as part of trying to avoid an import cycle, with a more
appropriate solution using the existing message_scroll_state module.
Importantly, it also means that we again wait for scrolls longer than
50ms to finish before opening All messages; I think this might fix a
regression.
This introduces a new INTERACTIVE_HOVER_DELAY of 425 milliseconds.
It's meant to be short enough that the tooltip's contents are
displayed without too much of a lag, but long enough that a quick
interaction--like +1'ing an existing emoji--happens without the
quick flash of the tooltip itself.
We refactor the code checking if wildcard mentions are allowed by
renaming the pre-existing function `wildcard_mention_allowed` to
`wildcard_mention_allowed_in_large_stream`, adding a new function
`is_recipient_large_stream`, then redefining `wildcard_mention_allowed`
to combine these two functions.
Fixes: #27248.
There is a discrepancy between price per license shown on billing
and upgrade page for annual billing frequency due to difference
in methods used to calculate the amount.
To fix it, we use the same method on both the pages.
I didn't use the helpers.format_money directly here since that would
create a visual delay in showing the price per license which will
not be nice considering that will be the only thing on this page
being updated that way.
The stripe fixture used is same as
free_trial_upgrade_by_card--Customer.retrieve.3.json
Since the server-side implementation no longer uses look-ahead
or (more importantly) look-behind, it is possible to exactly implement
in Javascript. This removes a common class which would prevent local
echo.
This requires reworking the topic linking algorithm, to march the
server's as well. The tests and behaviour are adjusted in so doing --
previously, the JS implementation would have linked `#foo` with a
`foo` regex on the linkifier, but the server implementation would not
have.
Earlier, we only passed in the stream name which isn't considered in
`compose_actions.start` when populating the recipient.
There was no visible bug since we first narrow to the correct stream and
then call `compose_actions.start`, but this was brittle since it would
not work if we open the compose box before narrowing.
We need to use the `navigate-link-on-enter` class for the scheduled
messages link in the popover, since it's a popover menu item with a
`href` attribute. The reason for using this class has been documented
in 246c1d7b4e.
While opening the send later popover via the keyboard, we want to focus
the first text option, which is "Schedule message" at present. When the
popover is opened via mouse click, we want to keep the original behavior
of not focusing on any option.
Note: This is similar to the behavior of the message actions popover,
and hence this commit follows a similar pattern to add the focus.
Fixes: #27701.
Currently we prioritize (even partial) realm emojis above all.
Including over perfect matches if the emoji is **not** a realm emoji.
The commit changes this behavior to prioritize perfect matches above all,
regardless of emoji category.
close https://github.com/zulip/zulip/issues/27545
Earlier, we suggested only the first matching wildcard mention,
irrespective of how many equivalent wildcard mentions match.
That logic leads to the exclusion of '@topic' suggestion whenever
a stream wildcard is suggested, which is not correct as
stream wildcards and topic wildcard are not equivalent.
This commit updates the logic to make the behavior of suggesting
only the first match valid for stream wildcard mentions only.
Earlier, we preserved the stream name across reloads, and to restore the
compose state, passed that to `compose_actions.start`, which since
c3fe96a, considers the stream id, and not its name.
This caused a bug where on server initiated reload, the stream input
would not retain its value and be unexpectedly cleared.
This is fixed by preserving the stream id across reloads, instead of its
name, and passing that to `compose_actions.start` instead.
This commit sets the 'message_type' to "stream" in the beginning
of the test to properly verify wildcard mentions typeahead for
stream messages.
Earlier, 'message_type' was set to "False", and the test was passing
falsely because in the 'broadcast_mentions' function, the 'else' block
gets executed for 'message_type' having a value of either
'stream' or 'False'.
```
if (compose_state.get_message_type() === "private") {
wildcard_string = $t({defaultMessage: "Notify recipients"});
} else {
wildcard_string = $t({defaultMessage: "Notify stream"});
wildcard_mention_array.push("stream");
}
```
We don't need to explicitly change the condition to:
```
else if (compose_state.get_message_type() === "stream")
```
because in practise, the 'message_type' is either set to 'stream'
or 'private' when the user types a message.
Only changing the 'message_type' in the test to simulate the
actual behaviour fulfils our purpose.
This should have been included in 5bddd8f.
For more clarity, the 3 related functions `is_search` in `filter.js`,
`message_list.js` and `message_list_data.js` are all renamed to
`is_keyword_search`.
The message shown on successful invitation included "User(s)"
previously. This commit updates it to include "User" or "Users"
according to the number of users invited.
Previously, the "All messages" typeahead would be shown whenever
the search bar had an empty input. This change removes that typeahead
completely in favor of using other filters such as -is:muted.
Fixes#27358
Co-authored-by: Heran Zhang & Wei-Cheng Sung
Turndown would strip `<style>` tags, but retain the content enclosed.
This resulted in unwanted content being pasted, like the comment nodes
inside the `<style>` tag when pasting cell/s copied from Google Sheets.
We usually default to pasting formatted markdown (unless the Shift key
is used to paste plainly instead), but when we're inside a code block,
we always want to paste plainly.
This is a prep commit for the following 2 commits, so we can opt out of
the formatted paste and automatic bulleting / numbering behaviors when
inside a code block.
When navigating to the next unread followed topic using
the Shift+N hotkey, we notify the user when there are no more
unread messages in followed topics.
Earlier, the hotkey simply did nothing in such a case.
We use feedback_widget to do so.
Fixes#27604.
It is common for users in other chat tools to start off a direct message
with @ mentioning the addressee. However, this results in potentially
unexpected behavior in Zulip, where the message is highlighted and shows
up in @ mentions in addition to DMs. We want to prevent this behaviour by
automatically changing the @ mentions to silent @ mentions for the user.
These changes only apply to 1:1 DMs and not group chats.
Fixes: #17998.
Rename the existing 'wildcard_mentioned' flag to
'stream_wildcard_mentioned'.
The 'wildcard_mentioned' flag is deprecated and exists for
backwards compatibility.
We have two separate flags for stream and topic wildcard mentions,
i.e., 'stream_wildcard_mentioned' and 'topic_wildcard_mentioned',
respectively.
* stream wildcard mentions: `@all`, `@everyone`, and `@stream`
* topic wildcard mentions: `@topic`
The `wildcard_mentioned` flag is included in the events and
API response if either `stream_wildcard_mentioned` or
`topic_wildcard_mentioned` is set.
This commit addresses the issue where scheduling a message for a
time less than 5 minutes in the future resulted in the confirmation
banner not displaying the scheduled time, which should have actually
been 5 minutes in the future, as enforced at the UI level by the
MINIMUM_SCHEDULED_MESSAGE_DELAY_SECONDS variable.
We remove the `is_send_later_timestamp_missing_or_expired` method along
with it's tests since `selected_send_later_timestamp` is already set to
the minimum of 5 minutes in the future through the `minDate` option of
the flatpickr.
Fixes: #26784.
Co-authored-by: Sayam Samal <samal.sayam@gmail.com>
Fixes#27489
For sticky message headers, we should not use the `first_message_in_group`
for narrowing to the topic since can be out of view for the user.
Presently in a session when an image or video title is updated,
the same change is not reflected in the lightbox view of
them, until a new session is made.
This is because the metadata of the media is stored in Map
for the session with the preview_src as keys, and hence,
when title is changed the preview_src being same, the title
change is not reflected inside lightbox view.
This is fixed by modifing the asset_map (cache map) keys
from preview_src to an array of preview_src, and the media
title. Hence, when title is modified now, the keys won't map
to the existing value, and hence new metadata would be
shown.
A new function get_asset_map_key_for_media() introduced
to get the corresponing metadata from this asset_map
by passing in the prview_src and media elements.
Fixes#21311
Bug reported here:
https://chat.zulip.org/#narrow/stream/464-kandra-js-errors/topic/Error.3A.20The.20partial.20.40partial-block.20could.20not.20be.20found/near/1673466
The reported issue was for a message edit request, and it was
happening because `channel.xhr_error_message` can return the
empty string sometimes, and `render_compose_banner` shows the
banner text if it exists and otherwise tries to render a
`@partial-block`. Unfortunately the empty string acts as falsy
in the template, leading to the partial block error.
This commit adds a default error message that was removed
in 96680e95fb.
It also adds a check to not show an error banner at all if
`readyState` is 0, which means the request was cancelled.
The popover didn't open since sticky_header is part of message feed
and we were hiding popovers if `elements_at_reference_position`
had `sticky_header`. Now, we just don't do that check if the
reference is part of sticky_header.
Align the `dropdown-widget` and `<select>` UIs.
Removed default `<select>` tag styles with `appearance: none` css.
Added custom chevron-down icon with `background-image` attribute
to the `<select>` tag.
The svg icon is used in a CSS variable with a data URI.
The CSS variable is used to set different colors for dark and light
theme.
Tweaked by tabbott to scope selectors appropriately and also delete a
10px font-size in one settings dropdown widget.
Fixes#26859.
This commit allow the users to modify notification settings for muted
streams through the 'notification' section located in the personal
settings.
Fixes#27272.
This commit changes .join(",") to Int1.ListFormat.
This produces more accurate translations.
Refactor compose_ui.test to update the testcase for Alice and Bob.
The previous test case output of "Alice,Bob" modify to "Alice and Bob"
to align with the behavior of Int1.ListFormat,
which joins lists using ',' & 'and'.
Previously when a user tries to click on the clipboard to copy the merge
base version info, the copied tooltip would appear in the wrong location
where the Zulip version is instead. This is caused by both clipboard
icon using the same Clipboard instance where the success event would only
display the tooltip at the Zulip version. These changes fixes this and
allows the clipboards to be independent of each other.
- Updates "Start a call", documents the "Jitsi server URL" setting,
and links to it in a (?) from the relevant admin UI.
- Fixes misplaced include block for opening the compose box.
Fixes#26907.
We introduce a state variable `shift_pressed` for the purpose of
detecting whether the Shift key is pressed when pasting content. Only if
it is not, do we proceed with the default formatted paste behavior.
In Chrome, plain paste works out of the box for `Ctrl+Shift+V`, but in
Firefox, we need to handle it ourselves.
We replace the call to `insert_syntax_and_focus` with a direct call to
`insert_and_scroll_into_view` when pasting content into the compose box,
which fixes the bug where leading and trailing spaces were added to any
pasted content.
Since focus on input elements on mobile opens keyboard which
changes window height, emoji popover can hide or scroll out of
view. To fix it, we don't focus on search input unless user
wants to when emoji popover is open.
If the popover reference is in an overlay, the below check:
```js
elements_at_reference_position.some(
(element) =>
element.id === "navbar-fixed-container" ||
element.id === "compose-content" ||
element.classList.contains("sticky_header"),
)
```
would make the popover not appear if the reference was above them.
To make things simpler for us to check, we only handle popover
references inside message list which we know for sure will
always be below the elements checked above in stacking context.
Hiding popovers for other view is done good enough by
`data-reference-hidden` check that we don't need to manually
handle them.
Since `#compose` occupies full width of the window, it is a
part of `elements_at_reference_position` if the reference is
under it.
Since `#compose-content` only occupies the part of window where
compose box is actually visible, we use it decide if the
popover should be hidden.
Before this change, we'd try to close the sidebar before
closing text fields. This means if the user sidebar was
open with a text search in progress, pressing Escape
would close the full sidebar instead of closing search.
This change deprioritizes closing the sidebar, in favor
of closing input fields like search.
This also helps with #26717, because the sidebar seems to
be marked as open even when it's not an overlay, which
means pressing Escape will first "close" it (no visual
changes), before a second Escape would trigger closing
the search field. Ideally we differentiate between
situations where we are showing the sidebar as an
overlay and when we aren't, but that might be a more
involved change.
Fixes part of #27491
We don't immediatley remove the input box, because there's
no existing UI for reopening the input box. (The input box
is always visible when the more topics view is visible.)
This commit makes the change of blurring and clearing the
input, instead of only blurring.
Fixes#27508
Show sidebar icon when sidebar popover is visible so that
it tells users for which row popover is visible and also helps
tippy track when the reference of the popover so that it
can apply [data-reference-hidden] property appropriately.
If the selection copied, is entirely within a single element, like text
within a heading (like `h1`) we discard this outermost element and only
convert the inner HTML to markdown.
However, we retain `pre` so a code block within it stays as is and does
not change into inline code, losing all its linebreaks.
Previously, when uploading of a file fails, the same file
couldn't be re-uploaded again without re opening the
compose box. This is because, the default behaviour of
uppy is that it won't allow re-upload of same files.
This could be fixed by removing the uploaded file when
the upload-error event is invoked for it.
Fixes#27495
ResizeObserver isn't supported for Safari iOS versions we support.
We support iOS Safari ≥ 12.2 but ResizeObserver support requires
iOS Safari >=13.4. So if they're on iOS Safari >= 12.2 and < 13.4,
it crashes.
Karl Stolley
evykassirer
Anders Kaseorg
Tim Abbott
Aman Agrawal
Prakhar Pratyush
retsambew
aryan
Vector73
N-Shar-ma
David
Sahil Batra
Alex Vandiver
Sayam Samal
kraktus
Prakhar Pratyush
Lalit
BrandonAWong
Joelute
Akarsh Jain
roanster007
ecxtacy
Pratik
Rahul