This focuses on some scenarios that are
a) complex enough to be worth a test,
b) but simple enough so I don't need to spend hours on comming up
with a test setup. ;-)
This patch also simplifies the ARIA related code in
MWTransclusionOutlinePartWidget a bit.
* Check 1 of the 3 ARIA configs only. Only having one is already
helpful and should not be skipped.
* No need for the large conditional. setAriaDescribedBy() works fine
with undefined.
Bug: T291157
Change-Id: I142782ec9b96147de64497f4f6a373eae05b9c8e
Will be removed when parameters are added. Needs different margins
depending on beeing shown in the single transclusion mode without
header or on multiple transclusions.
Bug: T300710
Change-Id: Ieb95d7276aa4d4b0fcbb74f87ab734e4a393dc21
Before, the fallback algorithm was somewhat adaptive, trying harder and
harder the fewer CirrusSearch results have been found. This updated
algorithm guarantees that 10 results are shown.
Warning: You might see only 9 results. The reason is a bogus, unrelated
behavior in the mw.widgets.TitleWidget in core that's used as a base
class here. There is a "showMissing" option that's apparently enabled
by default, and prepends a non-existing title from the main namespace,
ignoring the "namespace" option. This extra result from the wrong
namespace is later dropped by the very same widget.
This code here sees 10 results before the bogus one is dropped.
Disabling "showMissing" causes other issues. We would need a series (?)
of custom hacks to work around all this, but this seems inappropriate.
Let's live with 9 for the moment.
Bug: T303524
Change-Id: I2c577c9ef2752b6c6cd360f4023e151e9272fcd5
The main advantage of this change is that it drops the assumption
that the index starts with 1. This is not necessarily the case when
we prepend extra search results. Dropping this assumption here allows
to simplify such code.
* The incoming list of pages is guaranteed to be an array.
* There is no point (any more) that could cause the array to become
sparse.
* Note we still make a copy of the `origPages` array at some point,
e.g. on `.filter()`.
Bug: T303524
Change-Id: Ifbd92bb052155c613d2ca21ab6d54a0b3ef28c0c
This option is not only buggy (it just doesn't work when the namespace
option is set the same time), it is not useful in this context even
if it would work. It doesn't make much sense to suggest non-existing
templates in the context of the template dialog. If adding a
non-existing template really is what the user wants, they can still do
this by simply typing the name of the template and submitting the form.
We never need this to show up in the suggester below the input field.
The main advantage of this change is that is saves 1 useless API
request that's potentially done every time a key is pressed.
Bug: T303524
Change-Id: I903340a06d6e6490bb58f628f41903aa044ccb21
This separates the two steps:
1. See if items in the list of `origPages` miss their `.index`
property, and add it if possible.
2. Later code doesn't need to care about redirects any more.
Note that `origPages` is not used for anything else. And even if,
it's not wrong to have the index for each search result on both the
redirect and the redirect target.
Change-Id: I12135f0430c944b4e33c49ece7779d7c3bb6c211
This is not a pageid, but a simple numeric index in an array. Luckily
it doesn't make a practical difference for this particluar way of
iterating something.
Change-Id: I7ec9ace00d4fba7adde17670058a0365b30f5617
The result is guaranteed to be in formatversion=2, where the list of
pages is an array, not an object.
Change-Id: Ic73a68c3e249a70108a6a19a89f4ff6c475794ed
This code is optimized for the 2 most relevant use cases:
1. When Cirrus finds 10 results, we still want to search for the top 1
prefix match. This is critical for templates like !!. This will appear
at the top. unshiftPages() makes sure the limit of 10 is enforced.
2. When Cirrus fails to find anything, we search for 10 prefix matches
and use these instead.
The code can also handle everything in between. For example, when
Cirrus finds 5 results, we search for 5 more prefix matches and add
them when Cirrus missed them. The total number in the end might be 5 to
10 depending on the number of duplicates. This is intentional. Why?
Let's say we always search for 10 prefix matches and add them to the
top when Cirrus missed them. This might remove _all_ Cirrus results.
This shouldn't happen. This extra code is only to fill in glaring gaps,
not to replace Cirrus. 5 results are fine.
Bug: T303524
Change-Id: Ib0471795124c0c7001b6901edaf8e7b380e426b1
There is already some kind of "fallback" to prefixsearch. We always
check if the top-1 prefixsearch result is part of the result set.
Because of this the current worst-case scenario is that only this
1 result is shown.
This patch implements a full fallback to prefixsearch. But only when
there are 0 CirrusSearch results. Further tuning might be done in
later patches.
Bug: T304925
Change-Id: I1927eedad60c9b9ac2021481a85376c08ccf6fdb
.test() is the dedicated syntax for a boolean "does match? yes/no?"
check. .match() returns an array of matches, or null. This is just not
needed in these situations.
Change-Id: Ibb996ab843d1a6c7d7af98d6a112990665d543b2
We ended mixing two concepts in a single method:
* We need a method that allows us to create the parameter list widget
when it's needed, even if it's empty. This is relevant when a template
is entirely empty, and the first parameter is added. This wasn't
working. Instead the parameter list was created with all parameters.
* On initialization we either want all parameters to be shown, or only
used ones. But this code is only needed once, on initialization. I
ended inlining this code in this patch.
Bug: T300640
Bug: T304046
Change-Id: I6620a870e4420dcb8fecf522b3274458eeec891d
Due to the stickyness we need to scroll the header and the first
visible item from the parameter list to have the best result in all
cases.
We also only want to scroll when the user triggered hiding the
fields. Not when the template outline view is initialized.
Bug: T302965
Change-Id: I84d293888a7dbf13ec655c293c0fc3a79edca698
In a test case with 200 templates where all but a few parameters are
unused the loading time is cut in half.
Bug: T300974
Change-Id: Ice850cb9e5e95b9e3a19ff511b3a4f32117c7199
Same random finds while working on something else. I carefully
checked and made sure these methods are actually called without the
optional parameter.
Change-Id: Iab36fd130258322985b5d6e7f8e1f7b4ee235ba2
These are only needed when we need to access a specific `this` from
within another `function () {}` context. This is not the case in the
situations here.
This is split from Ibf25d7e to make it smaller and easier to argue
about.
Change-Id: Ide1476de91fc343aa992ad92a1321d3a38b06dd0
The `.` character class matches any character *except* line terminators,
but edit summaries can contain those. Use [^] to match truly everything
in the comment part. (In the section part, I assume `.` is okay.)
Bug: T302103
Change-Id: I29fcdd7489d118674bab5cfe5c0a15b8e4efac64
I bit more logic was needed to make sure, the state of the widget
resets when all unused fields are added during search.
Bug: T299811
Change-Id: I3006c233fda5490e323bc3a3e631bf0c1199bda3
Covers the base functionality to hide the parameters. Performance
optimizations could be done in a follow up.
Bug: T300640
Change-Id: Ia99b5da392273f1445e475a0720a656460612dcf
The behavior is now consistant with what would happen when the
buttons are triggered.
Instead of emitting a button click I directly wired the methods
that will by triggerd by the click. This might make it easier to
remove the old sidbar later.
To avoid movement when the buttons should be disabled, an
additional check was added to the onMove method. It's not identical
to the more complex check in the outlineControlsWidget, but should
be enough for our use case. The onDelete method already just does
nothing if nothing is selected.
Bug: T300971
Change-Id: I8a278c9657c91fd648944b5a8c1204c9fff75b7e
This option was added in 0.43.0. Now that the close button is handled,
the remaining functionality (store a flag in local storage, and fixing
link targets) doesn't really justify a separate class, especially as
it's currently only used once.
Change-Id: I0fd81cadccc077dbf957302f9f41409c5a1f4f20
The search field is of not much use when there is not really anything
to search. It wastes more space than the actual list of parameters.
Approved by UX, see T298259#7626538.
Bug: T298259
Change-Id: I01784a1c463d8b0b504897b20179719f91597d19
From the user's perspective nothing changes. A template without
parameters doesn't show anything. (Technically there was an empty
<div>, but it doesn't do anything.) The moment the first parameter
is added the required sub-widget is created and available from there
on.
This saves loading time and memory, especially when a multi-part
template contains many templates without parameters.
Bug: T298259
Change-Id: Ib1bd2cd0be4fece4acc92e6e5f63133a7986cf81
From the user's perspective this is the same as before: When a
template doesn't have any parameters, there is no search field. The
moment the first (undocumented) parameter is added the search field
appears.
This is just delayed now. The widgets are only created the moment
they are actually needed.
This saves loading time and memory, especially in a multi-part
transclusion with many zero-parameter templates.
This also makes it a lot easier to change the minimal number of
parameters from 1 to e.g. 4.
Includes reverting the flexible header composition done in
Ib050e30a50ef965c1524e977d3a600c3ff836774
Bug: T298259
Change-Id: Ied7541d8d5c0b478a439dd31ce072e634287f181
The search field will stay at the top of the window while scrolling a
single template. In multi-part transclusions, the header will also
be sticky.
Hides the template header in single-part transclusions.
Bug: T298618
Change-Id: Ib050e30a50ef965c1524e977d3a600c3ff836774
This is guaranteed via ve.init.mw.Target.getContentApi(). But the
ContentTranslation extension replaces this, and does not set a
formatversion. See e.g. SectionTranslationTarget.getContentApi().
Bug: T298599
Change-Id: I8768cae3153e9cbc29a8796ec21ef249f80471ed
Prevents accidentally treating plain text or user input
as HTML, which could be an XSS vulnerability.
Change-Id: Id4af48447a0907962a57340cb60aca08df9cc505
* Fix incorrect use of .append() instead of .text() (which was causing
some l10n messages to be treated as raw HTML)
* Avoid escaping and parsing HTML several times when plain text was
intended
* Remove some unused options and variables
Follow-up to 839b64d882.
Change-Id: I124257c73fe09713afefccdec8e90200e6ae433d
This sorting algorithm was introduced via Ic6bc348 (T274544). Note
there is no index parameter in the .onReplacePart() handler at this
point. When a part was moved, it was removed and simply appended
to the end. The additional sorting was needed to move it back to the
correct position.
This changed a few days later via Iafe29f1. There is now an index
parameter. The .onReplacePart() handler does the same as before, but
puts the part at the correct position right away. The additional
sorting is pointless since then.
The removed code alone is responsible for 1/3 of the total blocking
time when the template dialog opens.
Bug: T296335
Change-Id: I6c3fa70b532d34cd29d59c3b48ab81ebf608d548
Allows setting aria labels and descriptions on elements in a
convinient way. I did not use the the .mixin. convention here for
because there's already another mixin in that folder that's also
not having .mixin. as part of its name. And then there's also no
no need to open up that extra namespace here.
If we move this upstream at some point this can be changed though.
Bug: T291284
Change-Id: I1b3d40400d539f851f13719e16ced200968a7f92
When changing the source in the described-by attributes the screen
reader will read the text of the new source when the status changes.
Just changing the text within the elements holding the descriptions
does not work.
Bug: T291284
Change-Id: I31cc3061cf6c1f699babe41e99e0711f0eb03646
ARIA-selected only works on specific elements/roles. I tried several
combinations here, the most fitting seemed to be the option role, but
that role did not work very well in FF with NVDA. It also should only
be used as direct child to a listbox e.g. with several children.
The next role that's working with ARIA-selected that seemed fitting
is the gridcell. It's still a bit hacky but works well in IE and FF
with NVDA. I suspect that that's pretty good coverage already then.
Bug: T291284
Change-Id: I85c865b0ab12d3923e472e5f36b5c07b7c722180
The following keyboard combinations are now supported on MacOS:
1) CMD + DELETE
2) CMD + DEL
3) CMD + FN + DELETE
(1) for keyboards without a DEL key use DELETE/BACKSPACE
(2) for keyboards with a DEL key
(3) FN + DELETE is a built-in an alias for DEL
Bug: T294519
Change-Id: I896837954b2f0b36a25484080e57d929a5abf774
We're mainly interested in the layout of a button here. From a
semantic perspective this is just a header to an editable area
and not a button.
Bug: T291284
Change-Id: I683cca2e7d6549e652bd03ae1e46f4eff8c07d0a
This is a more radical change, compared to the previous patch.
I will post more detailled explanations as comments on Gerrit.
Change-Id: I6909b3f0b2c153b7ee9995441e995ffa793eab40
The event "focusTemplateParameterById" targets the right side of the
dialog. The input field for the parameter should be focussed. This
doesn't make any sense when the parameter was just removed (i.e.
unselected).
Change-Id: Ie75b1edaebe9d0444b98e66cb56a5c7774393bea
Some cleanup to improve readability and reduce the amount of code.
Relevant bits:
* One method name was wrong. It can actually return parameter ids,
not only top-level part ids.
* I got rid of some fail-safe checks that are never needed or moved
them to a more central place.
Change-Id: I08f2ad7bc7d3f985d6189dff170dda554f3d37c2
The idea of this piece of code is to make sure both sides of the
dialog show the same element. But it doesn't make sense to force the
*header* of a template into view when I clicked on a *parameter*.
Bug: T292718
Change-Id: I9945f8e54c856152f05bf717e43468ab5ab24d2f
I run into this in some local test. There are two reasons this code
can be reached:
* When a wiki doesn't have the TemplateData extension, the
additional API call from line #154 will fail. But the original
search query succeeded. We have the `originalResponse` and can
return it. This makes the code behave as if the additional
TemplateData API call was never done.
* But what if the original search query failed? We still end in
line #183 – as we should. But this time it can't return anything
but undefined. This will be considered a valid, successful API
response. But it isn't.
There might be a better way to clean up this chain of promises.
This is the smallest fix I found.
Change-Id: I02d3d053156da222ee424382007621f314777015
Instead of using an object mapping namespace ids to if they have
subpages enabled or not, pass an array of the namespaces where
subpages are enabled, reducing the size of the configuration that
gets loaded on all requests. Only requires a minor update to the
JavaScript that uses the value (check for array index instead of
object value).
Bug: T291729
Change-Id: Ia0ecac71721eceed52cc90f39ecc560bdf1b7f9b
I moved some code around and found that quite a lot of code wants to
know:
* Is the length of this transclusion exactly 1?
* I need that 1st part.
There is more that can potentially moved from the dialog to the model.
But I don't want to make this patch to big.
Bug: T292371
Change-Id: Ia94ed0450d04dd97c4c41f5bf7c266f9a534e821
This method does not only select a "part", i.e. a top-level item like
a template. It also selects sub-items like template parameters. The
new name reflects this better.
Change-Id: I51a8ddbd05b283248afba5a623cc52da7b2434f5
Note this patch is somewhat incomplete. The feature fully works and
I would like to see this patch merged as it is. But whenever you
press one of the keys the focus is stolen by some element on the
right side of the dialog. This makes it impossible to e.g. press
Ctrl+Shift+Down multiple times. The idea is to work on this in the
next patch.
Bug: T290262
Change-Id: Ic67f2a696c94f1e5c71134d681161221aecbfdf6
… obviously only to methods that are meant to be private, i.e.
only called from within the class (and possibly tests).
Change-Id: I581558078dc7210abac5f5724f71316ac45745e6
The .onUpdateOutlineControlButtons() method doesn't describe what it
actually does. This issue was introduced in I9c5478a. (Intentionally,
to not make the patch to complicated.) Let's continue to rename
things to be a) unique and b) honest about what they do.
This is an alternative to I8d98e61.
Bug: T289043
Change-Id: I4d52ffa6e9e5df2025a0c33031c1517bcb421279
I can't really tell what insight we get from the word "container".
Every widget is a "container" in some sense, isn't it?
This widget is just _the_ outline, I would argue.
Other suggestions?
Change-Id: I1fb27ee58c1a3dd790022504e978198dadf7ea02
This prevents multiple highlighting, which was possible to achieve by clicking
sequentially in the input fields of two parameters from different templates.
Change-Id: I404936f1569ab544b693a9bc6921381636ea8f40
The weird auto-scroll feature is described in T289043#7297679.
This also fixes T291381 different than I393a2b1. Only one of the
patches should be merged.
Bug: T289043
Bug: T291381
Change-Id: I70d87f12fd68001e880510fb6c38d7c419d64b15
This code was introduced in I8fafee6. I can't tell any more what
the "bug" mentioned in the commit message was. Let's get rid of the
duplicate code path, see if we run into regressions and deal with
them one by one. That's much easier to handle than keeping this
confusing code path around.
Note this "focusin" event handler was actually re-implementing
parts of the upstream BookletLayout, namely
OO.ui.BookletLayout.onStackLayoutFocus().
Bug: T289043
Bug: T291381
Change-Id: Ib386ae6efec08465122f0e8ee81cd6dc9a2d337a
The previous patch Id314ee8 was incomplete. The event changed.
The id in the event is not guaranteed to be a top-level partId any
more, but can be a template parameter's id.
Note: "Parameter id" and "pageName" is the same. The fact that
these ids match is how the left and the right side of the dialog
communicate.
Bug: T289043
Bug: T291151
Change-Id: I391f0f8edb96398fd33a2e0b01003013c52776da
The behavior of the enter key in the new template dialog sidebar
is somewhat inconsistent. When pressing enter on the name of a
template it sometimes just doesn't work, but focuses something
else.
I realized this is because the message "The … template doesn't
yet exist." does not have a link. There is nothing to focus in
this element. The code just gives up and the selection returns
to whatever was selected before.
It works when there is a link in the template header. But this
is not even that useful.
Let's try to always focus the first parameter instead. The user
can still press Shift + Tab to focus the link to the template
page.
Bug: T289043
Change-Id: Id314ee8ebf47d387df08c7fb432094b6d8f7a3d2
This patch improves the error handling for when a user tries to add
a parameter which is either an alias of a existing parameter, the
primary name of a existing aliased parameter, or a name/alias of an
existing parameter which is shown with an override label.
The error message was modified to always refer to the conflicting
parameter using the same name that is has in the sidebar.
Example: A parameter named "Parameter B" is already present in the
sidebar under its alias "B". When a user tries to add "Parameter B",
the new error message will inform the user that the parameter they
are trying to add already exists as "B".
Bug: T285869
Change-Id: I762b72b6cf14eb8ff5fcef63b4dcb70e297050de
There is no point in firing this event when noting changed.
This should reduce flickering and some of the issues described
in I97d77f4.
Change-Id: I7c387889a4a33dac5053cec11a0641d358020b56
As a reminder (not part of this patch): Pressing enter on the name of
a template should select it, and jump to the content area on the right.
Pressing space (that's what this patch is about) should select as well,
but not move the focus.
The best way to test the behavior is with a multi-part template.
Bug: T285323
Bug: T289043
Change-Id: I97d77f43b231696f92ba6758a6b8feac34e02e6d
I tried to review all of them. Some of the changes I did:
* Make sure the `config` parameter is not marked as optional
when it is not.
* Make sure default values are mentioned.
* List individual `@cfg` options when it makes sense.
Note I don't list all options a class could accept (e.g. via all
its parent classes and mixins). That's too much. Instead I checked
how a class is actually used and list only these options.
Even then I don't list everything, e.g. unspecific options
like "classes" that can be used pretty much everywhere.
Change-Id: Idf4fbe1dc3608ace277df9e385f2f140df3a2f50
We don't need to distinguish between these any more. Both are
"active", i.e. both focus the widget on the right side of the
dialog. Sometimes the "choose" event is fired to actually add
or remove a parameter. Sometimes it's fired, but the state of
the parameter doesn't change (for whatever reason, i.e.
because the parameter name was clicked instead of the
checkbox). There is nothing to do in this case, except for the
focus change.
Change-Id: I3c7c0c81a075ccff76eda0a4fb2aa1ac7be3cec5
* The template model fires an "add" event. Listeners don't
automatically steal the focus any more.
* Instead there is a separate "focusTemplateParameterById" event
fired from all relevant places that add a parameter.
* The "remove" doesn't steal the focus any more.
Bug: T285323
Change-Id: I93f17727524bfbcf6f11647a6c2441781337c4cc
The 'classes' property is a OOUI interface. Personally, I like
this code style better.
However. It appears like the code style in this codebase is
somewhat mixed. It looks like the top-level .$element always
uses .addClass(), while other code uses the 'classes' property.
Should we unify this?
Change-Id: I9ecd75e22d00f06ffd707f766dc9e8d748ff9a37
In JavaScript .split() behaves different, compared to PHP. In
PHP the last element contains the rest of the string. In
JavaScript the rest of the string is discarded. The limit acts
as if the array is truncated. That's why we can reduce the
number in
'foo/bar'.split( '/', 1 )[ 0 ]
to 1, as we are only interested in the element "foo".
The same code in the other class is currently not covered by a
test. But changing it accordingly should be obviously fine now.
Change-Id: I20c27d480ddb1799df9eb1e5bc119b724e80653d
The original idea was to make the interface as narrow as
possible. However, it turns out it's better to model the
"templateParameterClick" event more closely after the "choose"
event.
This is split off to make reviewing the following patches
easier.
Change-Id: I271f576c6cd756cecfc6cb1fd64810f8da5c3575
The list of parameters should remove itself from the list of
possible tab navigation targets when it's empty.
Note there is no way to remove elements from the parameter
list. That's why we don't care about "remove" or "clear"
actions.
Change-Id: I8b1215117e0ddc94f787d173e9bea6f7567d9671
E.g. avoid calling the rather expensive method multiple times
in a row, if only 1 of the results is needed.
Change-Id: Iff1d2c0892367e927303f6f45d3231e04c045cab
I tried hard to come up with the best possible names. Some of the
criteria I used:
* Longer and more unique is better. This makes it much easier to
e.g. search for the event name.
* The term "part" should only be used for top-level parts. While
template parameters have a unique id, they are not a subclass
of …TransclusionPartModel and therefor not "parts".
* BookletLayout manages "pages" via "page names".
* The page names of top-level parts are identical with the part
id, see ve.ui.MWTemplateDialog.getPageFromPart.
* The page names of parameters are identical with the parameter
model id, see ve.ui.MWTemplateDialog.onAddParameter.
Some code knows parameter ids, but not what pages are. Other code
knows page names, but not what parameters are. The transition
currently happens in the …OutlineContainerWidget. We might want
to move this point up to the …TemplateDialog. But I would argue
this is good enough for now and can be changed later, if needed.
Bug: T285323
Change-Id: Iab2805b3203988db400b67c8d00e48905fdc53dc
I tried to memorize the cursor position in the list of
parameters. This way you could leave the list with tab, and
return to the original position with shift+tab. Unfortunately
this is not how the SelectWidget works. The moment the
highlight is gone it's gone. There is nothing that remembers
a position. We could introduce code to do this. But I feel
like this is a lot of effort for not much benefit. Also not
listed as a requirement anywhere, at the moment.
Bug: T285323
Change-Id: I8d44ba4539ec4b5535bc031accfeacd87e1886eb
This got lost in patch I20dbd2b.
Both events come from the same sidebar class. The difference
between the two is:
* selectPart is when the button representing a top-level part
is clicked.
* focusPart is when a parameter name is clicked while the
parameter is already checked.
Yes, this is confusing at the moment. Following patches will
rename, merge and split a lot of these events to be much more
self-explaining.
Bug: T285323
Change-Id: I0c6b53c93c712ff5e47c1beb5199d590cba7ab1a
