Also adds browser tests for the behaviour of the mark all as read button
to ensure it only clears message notifications.
Dependency: Ifb7b1b7b7feb4a5af65c79bb16b91a5a9c70166c
Change-Id: I46e1de229e32d705e67cebde678ecea3f3633906
This removes the need from the super specific selectors.
It also allows us to create multiple notification elements
on a screen.
Change-Id: Ife5291a315915779689943ffc18ac73d77351d0e
IE8+ supports the pseudo element. The world will not explode if
IE6 and 7 do not have the chevron.
This simplifies code by dropping the use of the pokey element
Note: I know we can drop using the image too but I decided not to do
that in this patch.
Change-Id: Ia9ac434461e63bc2cf8554060126995ac65a3531
Danny noticed a bug where if both tabs have unread notifications,
then when opening the overlay and clicking on the alerts tab, the user
would be reverted back to the messages tab.
Test stops this from happening again.
Change-Id: I6bbbbf61251957336de8856201412fa3569ab22d
Don't mark messages as read until they have been acted on.
Show a mark as read button that marks entire list as read.
Change-Id: I4450a66cffd11c67b9a4ba9aac0fe958dc760e15
Note no design was defined so have taken this to mean use
mediawiki ui for consistency purposes.
* Use mw-ui-active and mw-ui-quiet for tabs
* Update tests
Change-Id: If7a51b2286cdfe6e839dacc2476c9a578bc7f1df
Shift to new API to support 2 tab view
When a new has no messages they will see the old style overlay with
Notifications heading. I have added tests to assure this is the case!
Later patches will:
1) Add the mark as read button only in message view
2) Note currently the tabs do not refresh when notifications is clear.
We need some kind of EventEmitter to make this sort of thing easier.
Change-Id: I62b590e81cd3fe867c4c13959cb43466aacfe2d5
Make various properties of the overlay properties of
EchoOverlay to reduce need for function parameters
Change-Id: I70ec0c650d58322aad93d71292a836d7cf7e7474
Not as scary as it looks. The change to _buildOverlay simply changes
indenting, and cleans up to reflect new parameter. Tests still pass ;-)
Kill noticationLimit concept
Only expose what's necessary on mw.echo.overlay
Change-Id: I2671a41af8188c14ad4c910396afa0d9000b6051
No usecase has been provided, and additionally the hooks
are not documented. Bartosz also points out that
mw.hook calls are asynchronous and memorized.
This reverts commit 9d3561afaf.
Change-Id: If735b46996fab3def835a54223412ef6d3105395
Adds two new hooks to allow extensions to tie into the notification
loading process. Between these two hooks and the new isInitialized()
method any extension can run code whenever a notification is displayed.
Change-Id: If351835be5f65ca098e2d235ea8c8e4dc40ae2b4
General code cleanup as reported by the PHPStorm static code
analysis. I hope it's not a problem that I made a lot of very
different (but all very tiny) changes in a single patch. If you
want to merge this but you think it's better to split it into
several patches first, please tell me.
Change-Id: I2e2c4bb47f8d20e038d28e236e2ff813b30504af
Tested with latest Firefox and Chrome as well as Opera 12.
See bug 57327 for full description, please.
Bug: 57327
Change-Id: Ibf6e7d590754402f4cfafa3a3db55bb0c6aa525e
The badge is always close to the top of the window (if it's visible), so
this is not necessary. Also, this behavior is definitely not desired
for the upcoming fixed header beta feature (the badge will always be
visible as a part of the fixed header).
Change-Id: I66e8a50b1139f7bcd005cfb3c3576578efe6a653
Provide users a better user experience by catching and logging
errors with Echo notifications.
Bug: 60906
Change-Id: Iee93a05c6eed468af8bbfa60249df0819c49c45b
Changed the font size of notifications which was earlier too small
and also the accessibility was also too low
Bug: 60239
Change-Id: I20be3cd2277e865ee694070f119999b7b170547c
If the echo-overlay-title-overflow text is too long, it's wrapped into multiple lines now.
Bug: 55919
Change-Id: Ib1d252ab26be7d73cbf71c6fb19d84b80d8d30c8
Apparently three wrongs make a right: hiding the overlay still worked.
* '#pt-notifications a' click event handler was not correctly checking
for events comig from inside the overlay due to a missing dot before
class name
* But if it were correctly checking, it would not be hiding the
overlay pokey (just its body) due to a copy-paste error
* Instead the clicks were handled by the 'body' click event handler,
which did not correctly check for events coming from
'#pt-notifications a'
I made things right again. The user-facing behavior is the same.
Change-Id: I02aba0e25ba4d81b234b327f120e0e2ff13d117c
Otherwise a new copy will be created every time the user opens and closes
the popup, and jQuery will hold on to each one forever (because it has
events bound), leaking a minuscule amount of memory.
Change-Id: I70c713be839f826fc27d07b04260c166f9052020
Core styles for Monobook include high-specificity background: transparent;
rule for #p-personal li a, we need to match it to set our background.
Also change hover behavior: switch to a deeper orange instead of
default white, similarly to how the badge already behaves.
Partially reverts I682182fe.
Bug: 56214
Change-Id: I9f343264c395ecf09c1e34e03d208ec2119fb622
Follows-up 383a818.
* There is no need for the additional element ("a") or
descendant ("#pt-mytalk") selector.
It isn't overrriding anything, only hardcodes details that make
it harder to maintain or extend this stylesheet. For example,
there is a gadget that makes user messages green instead of
orange, it now was required to hardcode the "#pt-mytalk a" part
of the selector eventhough those are subject to change.
Separation of concerns.
* Cache/reuse the jQuery object instead of executing the same
query to the document, again. It also avoids a potential bug
where the second query matches different elements (e.g. after
appending alertMessage, there could potentially be an additional
nested anchor link; there isn't now though, as the message is
plain text).
* Add comment about weird echoNewMsgAlertDisplayed variable.
Change-Id: I682182fe15a868969f25fa5bfe2412e2a6f3dddf
* In some languages like persian, the number 0 is represented as '.', we can't compare
'.' with either 0 or '0' to detect the no-notification status of the badge
* The markread API doesn't respect uselang param, it would return 0 instead of . in a url with uselang=fa
Note: we need to provide raw and formatted count in the API since client side javascript
doesn't provide fancy function like $wgLang->formatNum()
bug: 54575
Change-Id: I0a49828253ec346ed27c5b9a976f8bdff4e1fa90
Removing unused functions and declaring correct dependencies.
Targetting to desktop and mobile so it can be used by both.
Also removing dismiss-related code from the formatter.
Change-Id: Icccce64cfb3c564ab468a93ccdba9c5a61687fd5
Also making sure that footer has some amount of separation from
the notification title even if there is no payload.
Change-Id: I85a1a7989539044a0b0b53e76e70ddee9bb7165c
The space after colon on this line is actually a non-breaking space
( ), which is not allowed in this place according to CSS syntax.
Therefore all browsers ignored this rule (logging an error like
"Invalid CSS property value") and the shadow was never shown.
We could fix the rule instead, but I'm pretty sure users are used to
this by now, and in my opinion it looks better without shadow.
Bug: 53490
Change-Id: I1fd508d2059bec5cd79a6dcdce8dc9be6e6d4229
We want the notifications in the flyout to behave just like links,
including standard middle-click and Ctrl-click behavior. The simplest
way to do that would be to actually make them links - but the area can
contain a few other links, so we can't do that and have to resort to
ugly hacks.
Or do we?
Turns out that while browsers won't accept HTML containing nested <a>
tags[1], such a structure is valid XHTML, and it's possible to create
such structure in HTML mode using DOM manipulation. It works like one
would expect: the entire thing is clickable, but inner <a> tags' hrefs
override outer ones.
Firefox even had a request to make that work[2] which was happily
fulfilled.
Tested the basic case [see below] on Firefox 22, Opera 12, Opera 15
(which uses the Blink engine like Chrome), IE 8 and IE 6 and it works
the same on all of them. Tested the XHTML variant [see below] on all
of the above except for the IEs which don't grok XHTML and it exhibits
the same behavior.
[1] Simple test: $('<div>1<a>2<a>3</a>4</a>5</div>').html() is
"1<a>2</a><a>3</a>45", not actually "1<a>2<a>3</a>4</a>5" like one
might expect.
[2] https://bugzilla.mozilla.org/show_bug.cgi?id=331959
----
The test cases used are below. When trying out the XHTML one make sure
that the browser uses application/xhtml+xml MIME type; saving the file
with .xhtml extension should be enough.
XHTML:
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<body>
<div>1<a href="http://google.com/">2<a href="http://example.com/">3</a>4</a>5</div>
</body>
</html>
HTML:
<!DOCTYPE html>
<html>
<body>
<script type="text/javascript">
var div = document.createElement('div');
var a1 = document.createElement('a');
a1.href = "http://google.com/";
var a2 = document.createElement('a');
a2.href = "http://example.com/";
div.appendChild( document.createTextNode('1') );
div.appendChild( a1 );
a1.appendChild( document.createTextNode('2') );
a1.appendChild( a2 );
a2.appendChild( document.createTextNode('3') );
a1.appendChild( document.createTextNode('4') );
div.appendChild( document.createTextNode('5') );
document.body.appendChild(div);
</script>
</body>
</html>
----
Bug: 52319
Change-Id: I311eca70f025ce92129c828cd88f96686b7cff72
Out of the default MW skins these only seem to affect CologneBlue.
* Reset padding and list-style-image on ul#mw-echo-special-container
ourselves, do not rely on the skin doing it
* Use transparent background on .mw-echo-notification (and
semi-transparent black on hover) instead of solid white and
light grey to accomodate colored skin backgrounds
Change-Id: I2c178627e4dbe889c4958afc41e4969aaa45a717
wgUserNewMsgRevisionId is null with only one revision and that revision is
not viewed, this makes it not reliable for determining if there is new
message. We just switch to check against $user->getNewtalk() instead.
We can still use wgUserNewMsgRevisionId to generate diff links if desired
Change-Id: I4cf50a944aada03151bd17f3610bd59b3bfb2bf2
separate into multiple ul elements with headings dividing them
This removes need for unnecessary styling rules in mobile web
Change-Id: I8acaef726ea0cc639054cf36d6c3649c7ab25bee
* Add missing spaces after 'function'
* Remove underscores from identifiers
* Don't use typeof for undefined with local variables and object properties
* API ok/err is deprecated since MediaWiki 1.20 (< 1.22-alpha), use .done and .fail
Change-Id: I5ca3403ee263a0c3e6709618e48e896340952337
