Previously, they were only being deduplicated based on the transaction,
which meant that an undo was seen as a duplicate (but then if you undid
again, that wasn't a duplicate).
Change-Id: If432ea28e6c206a2ad5562e529e2d3ed808c20e4
Parsoid switched from <!doctype> to <!DOCTYPE>, which exposed the fact
that our dirty regex to detect whether we're dealing with a full
document or a fragment was case-sensitive. Made it case-insensitive.
Change-Id: Ia8a38488e06ca7d7a6fb9a9699b5d9b5c5eb03f2
Clearing by type in SurfaceFragment didn't actually work. Instead,
it followed a code path intended for setting and created an annotation
of that type with no data, then tried to clear that. What we really
want to do there is clear anything with that type.
This fixes the bug where unbolding of text that was already bold in
the article didn't work.
Bug: 47680
Change-Id: I77f00e63c8732420063b0453fede7f453083c913
In most places we call .contains we already know the index, so we
can avoid store lookups by using .containsIndex.
Change-Id: I45a9a421473f9bec479ab8ccceceb162b7004c3a
It is going to be used at least for figure tags for which Parsoid gives as a lot of CSS class names that are useless for rendering purpose
Change-Id: I4b1e8084a6b7ab5294e0c3cf153fc6cffb3e8dac
This is minimise the amount of data we need to serialise when
sending this over the wire.
The minimal IVStore data is added to the MW bug report, and
editedData fixed to only return the data array, not the full
LinearData object.
Documentation in AnnotationSet has finally been updated to
refelect the fact that it only stores Annotations
(was previous the generic OrderedHashSet).
getAnnotationFromOffset has been split out into a function
that just returns this indexes so that in cases where we
don't need the values we don't do an unneccesary store lookup.
Bug: 47318
Change-Id: I4819cf06d1bd0ae4f8b896052e278ca75c9551bf
Instead of calling $.append for every single char - buffer and call $.append only when really needed.
Change-Id: I53acfa795ea5dc6a8ca39ce11017daa85c9151d2
The fixUpStack is actually redundant code and closingStack
and openingStack handle all our cases. It was causing the
insertion to try to correct balance itself in the middle
of inserting two paragraphs, causing the creation on an
empty paragraph between them.
Added a test case for the fix and other cases to make
sure removing fixUpStack hasn't caused problems.
Bug: 46800
Change-Id: I35e54165709ac56e8116359a7c3b487eecf08ff7
Reuse the existing internal link annotation builder instead of using a
constructor directly, and incorrectly (it's meant to be passed an
element, not an attributes object)
Change-Id: I4cda6a9c3442cb10ebbc0844630fedba403adc91
* Only place them in a high z-index while resizing so they don't render
above dialogs and menus
* Add resize transition
ve.ce.ImageNode.js
* Switch from element attributes to CSS for setting dimensions
ve.ce.Node.css
* Add resizing class for resizable nodes for z-index
* Add transitioning class for resizable nodes for transitions
* Switch from border to inset box-shadow to not affect handle position
calculation
ve.ce.ResizableNode.js
* Add/remove resizing class while resizing
* Switch from using $image to $resiable to make the class useful for
non-image node
* Enable transition and set new dimensions before transaction processing
which will cause re-rendering)
* Delay transaction processing for resize until after transition is
complete
* Add hiding of context menu on resize start
ve.ce.Surface.js
* Add getSurface method so we can get to the context menu
Change-Id: I4667e394d0af4a80b651c2a0f6d11d30e196bf60
Redone using document.implementation.createHTMLDocument instead of the
iframe trick. It's supported by all browsers we target, including IE9.
This also makes VE work on Opera using a nasty hack.
* Previously, for reasons I'm not even trying to understand, Opera
would sometimes return an empty generic object from
ve.createDocumentFromHTML() - but only if you weren't debugging it
(Dragonfly was disabled). I have no idea what is it about the iframe
hack that makes it not like it, but fact is, it doesn't work.
* Calling .open(), .write() or .close() on the document returned by
document.implementation.createHTMLDocument acts as if it was
window.document - that is, the entire contents of the web page are
replaced with new ones. That's probably a one-word bug somewhere
deep in Opera's innards; I reported it (it got the identifier
DSK-384486). Until it gets fixed, we work around it by using
document.documentElement.innerHTML, which works reliably.
Change-Id: I90ea547c735edaba9f7ecb8f685351ac6499c53e
This involves setting some i18n messages for the target languages based on the
translations already provided - I hope this doesn't break anything for TWN but
the need for this only just became apparent; apologies!
Longer-term we will need to come up with a better way of doing this, if we are
keeping the in-VisualEditor feedback link around.
Change-Id: Id6ed80cdcd4314e84e75fb718421767162d73ef3
Parsoid is sending us some unescaped HTML in the data-parsoid
attribute. When we try to rebuild ref nodes (inline aliens)
this confuses Firefox which tries to sanitise the HTML by converting
<ref/> to <ref></span>.
As a temporary fix we can manually escape <>'s inside the
data-parsoid attribute.
Also in this commit the new MWReference nodes have been moved
to experimental as they are incomplete.
Bug: 47417
Change-Id: Ib6a0cfb880e769f28b42c9fa63ddc1abc75c399d
ve.ce.Node.css
* Added prefixes for use of box-sizing
ve.ui.MWLinkInspector.js
* Whitespace
ve.ui.Inspector.css
* Corrected input width, always 100% wide now by using box-sizing
ve.ui.DialogButtonTool.js, ve.ui.Context.js
* Updated use of getViewsForNode
ve.ui.ViewRegistry.js
* Added inheritance-based prioritization for matching views with annotations and nodes
Bug: 47413
Change-Id: I286a28002c1691e58bbd7de04ed08cceb8b3bb07
Using left and right arrow key to move to and over an image will
select the entire node.
Bug: 37870
Bug: 38129
Change-Id: I70deadd2c2707149ea33e3b8ee42fb0d8508aacc
This is showing a separate need for refactoring. We call
"this.serialize( ..., callback )" but if it failed callback
is never called and an event is emitter for the error.
That makes it rather disconnected from each other.
In this case we're lucky that all calls to serialize are similar
in nature and need the same kind of error callback but other
wise this would be pretty messed up. It obviously needs to be
untangled and get rid of this akward eventemitter dance.
This doesn't fix bug 47581, but it does fix the "Infinite loader
with no error" problem as a result of it by handling the error
in a more intuitive way.
Bug: 47581
Change-Id: Icdf64a792c13a326f494e051be47f2946928d142
iframe.contentDocument doesn't seem to have a key called 'document' at
all; I assume a different nesting was intended.
Change-Id: Ia37e3719d5247408bac2dfad1717d9193fb84c06
To help the selective serialiser we can return the original
HTML for generated content if it is unmodified.
As the output of toDomElements now depends on changes
to the dataElement we now have a 'modify' function in
the some test cases.
We also now have 'storeItems' to assert that the index-value
store is correctly populated and for loading values back
into the store for toDomElements tests.
Also make 'mw' an attribute and remove 'about' property.
Bug: 47394
Change-Id: I2bbb5d2d6a90c4eb87fa129671112c92a9b931e7
Turns out, the context property of a jQuery selection isn't always
there.
For example:
$( 'body' ).context === document
$( '<div>' ).context === undefined
Even if you later attach that div, so long as you have the old
selection around, the cached `context` property won't magically be
updated. This makes sense (although it's poorly documented in the
jQuery API) but causes issues for us, and pretty much makes the
context property useless.
Instead, we can just use the standard `ownerDocument` property,
available on all DOM elements.
This change also add support for passing in a DOM element directly, in
addition to the existing support of passing jQuery or Document objects
in.
Change-Id: Ib8a31b74f2a4f455b1318be9f5c7805a2a193c79
The 'add' tabLayout path is pretty basic and up to date. The
older (now active again) tabLayout 'replace' was fairly outdated
and unmaintained.
Fixes:
* Attributes copied from the original (except for the 'id'
attribute) were not actually beinged copied over because they
don't exist on the ca-edit list item, but on the anchor link
inside that list item.
* Clean up messages from the module registry that were unused.
Keys 'accesskey-ca-edit' and 'tooltip-ca-edit' were also inexistant.
* Add message keys for tooltip and accesskey of editsource tab.
Depends on I0bde1a228983c58b in mediawiki/core.
Bug: 47396
Change-Id: If598552fac639da645a8b1273c5fc6028695fcc1
Also add detection for whether the browser is actually broken (most are,
but some, like Opera, aren't), treat <textarea> and <listing> in addition
to <pre>, and fix a bug where the function would crash if the <pre> was
empty (because .firstChild was undefined/null).
Change-Id: I541b57e9fd5c9c42d19d0a59f6e29fb43d35c9b6
addSet:
* Instead of indexing items in the store, just union the indexStore arrays
removeSet/removeNotInSet:
* difference or intersect the indexStore arrays
filter:
* push indices into the result set instead of values
simpleArrayUnion/Intersect/Difference have been created as utilities
in ve. They are prefixed 'simple' because they use object keys to
do fast in-array comparisons. This means they are limited to string
values or values which will compare as strings (e.g. numbers).
Change-Id: I079cbdfece4f6d80ec0afd61959913f13217fcb3
By removing the transaction listeners from surface fragments we
no longer have to make sure they are always manually destroyed.
In order to retain the functionality of having fragments update
with transactions elsewhere we keep a pointer to a place in the
new complete history stack in the surface. The complete history
stack records all transactions, even undone ones.
Whenever getRange is called we replay all transactions in the
complete history (in the correct order) since the fragment was
last updated.
Also in this commit:
* Updated Format/IndentationAction to test undo(). This increases
coverage of surface fragment behaviour.
* .range is always accessed by .getRange now, although as an
optimisation we can use the noCopy mode when we a sure the
returned range will not be modified.
* Added undo test to .update (previously .onTransact)
Bug: 47343
Change-Id: I9e9818da1baa8319a3002f6d74fd1aad6732a8f5
Actually really resizing the image
Show bounding box on mouseover with 4 handles. Bounding box is resizable. Image resizes to match bounding box on mouse up.
Change-Id: I1f3dac64eb86dd1f258937e4915af101b3ac19d8
*.php
* Added links to new file
ve.ce.ImageNode.js
* Added relocatable node mixin
* Added $image reference to the actual img element, so if it's wrapped
in a sub class the functionality in the parent class doesn't break.
* Moved drag start event handling to relocatable node
* Removed drag end binding, not needed.
ve.ce.MWImageNode.js
* Moved addClass to initialization section of constructor.
* Copied 'view' data prop from image element to keep stuff working after
the wrapping.
ve.ce.Node.css
* Switched to default (arrow) cursor for images.
ve.ce.RelocatableNode.js
* New mixing for nodes that should be relocatable
* Added implementation for drag start, which tells the surface to allow
dragging this node.
ve.ce.Surface.js
* Added relocation support, which is used by relocatable nodes
* Split onDocumentDragDrop into onDocumentDragOver and onDocumentDrop
which now have implementations that support relocation of nodes
ve.ui.Context.js
* Added relocation tracking to prevent context being shown while
relocating
Change-Id: I8703adfb707af2c3224431afc3418356ac2c686c
Currently some issues, probably with loading nodes
after factories.
Toggled by global $wgVisualEditorEnableExperimentalCode.
Change-Id: Idab3dd68572c037289c6742d03fd327285110f67
Current set up leaves us with restoring when we're not and vice
versa. Not good. :-) (Partial fix of change 59968.)
Change-Id: Ia33a2f3318cf2e46b7469b2c773e91c5ee8fdefa
Don't call initialize inside the try-catch, it ends up sending exceptions thrown inside that method to /dev/null.
Change-Id: I8e0945f35c639ec156ee9a163b86fddfaed0ea7b
This was broken, especially in wrappers.
Changed the wrapping algorithm so that meta items are placed outside
wrappers if possible. On the left-hand side, this is already the case:
we don't open wrappers for meta items. On the right-hand side, this is
accomplished by buffering the meta items and only inserting them when
we encounter either real text (not whitespace) or the end of the wrapper.
If we're interrupted by real text, we insert the meta items with the
unmodified whitespace. If we're interrupted by the end of the wrapper,
we insert the meta items outside of the wrapper with whitespace stripped.
Internally, this is done by stripping the whitespace into the whitespace[0]
of the meta item to its right. Then when we output the meta items, we
either decide to 'restore' the whitespace, or to 'fixup' by also setting
whitespace[3] on the element before the whitespace.
Change-Id: Ibeea2a9906c4aae9fe6d284613edd6ec853ca5e7
Objective:
Make it possible for inspectors to inspect nodes or annotations, rather
than only annotations. Meanwhile, also make it possible for dialogs to
edit an annotation.
Strategy:
Switch from using type patterns to associate inspectors with annotations
to using arrays of classes, similar to how dialogs already work.
Introduce a view registry which provides lookups for relationships
between models and views. This is more centralized and less repetitive
than implement matching functions for both annotations and nodes in both
the dialog and inspector factories.
Changes:
*.php
* Added links to new file
ve.AnnotationAction.js
* Removed unused parameter to filter annotations using a string or regexp
ve.dm.AnnotationSet.js
* Switched from property/value arguments to callbacks
ve.ui.*(Dialog|Inspector).js
* Replaced type patterns with class lists
* Added class to view registry
ve.ui.*Tool.js, ve.ui.Context.js
* Updated model/view relationship lookup
ve.ui.*Factory.js
* Removed overly-specific lookup functions
ve.ui.Inspector.js
* Removed typePattern property
* Updated model/view relationship lookup
ve.ui.ViewRegistry.js
* New class!
* Migrated node and annotation lookup functions from factories
Change-Id: Ic2bbcf072fdd87e5ce8a03fe1ae3e6d8d50e2593
* changes:
ve.init.mw.ViewPageTarget: Put the Edit source link in the visible tab area
ve.init.mw.ViewPageTarget: Switch tabLayout from 'add' to 'replace'
So. It turns out that the design of SurfaceFragment is a little -
shall we say - wonky.
One of the best things about ve.dm.SurfaceFragment is its magical
ability to retain the intention of its range, even as transactions
are being processed. This ability is granted by each fragment
listening to the surface's change event, and responding by using
translateRange for each transaction that gets processed. Surface
fragments also have these clever methods that allow you to get a
fragment based on another, which makes adjusting the range easy to do
inline without having to manually store multiple fragments or
modifying the original.
This sounded good, and we seemed to all be convinced it was well
designed. But if you add a console.log( 'hello' ); to the first line
of ve.dm.SurfaceFragment.prototype.onTransact, and then start using
the bold tool on various selections of text, you will find that there
may indeed be a flaw. What you will probably realize is that the
number of times that particular line of code is being called is
disturbingly large, and increases each time you do just about anything
in the editor. What's going on? How did we get here? Read on…
It turns out that fragments are immortal. We create them, they listen
to the surface's transact event, we are done with them, but the
surface keeps on emitting events to the now long forgotten about
fragments. They continue to build up over time, never go out of scope,
and bloat the hell out of our program.
The same ended up being true of toolbars - and each time the context
menu fired up a new one the old one was left in limbo, still
responding to events, still taking up memory, but not being visible to
the user.
All of this immortality was causing strange and difficult to track
down problems. This patch fixes this by introducing a destroy method.
This method unbinds events, allowing the object to finally fall out of
scope and die - and more importantly stop receiving notifications of
changes.
This is a hack, but Ed will no doubt get this situation sorted out
properly by making fragments lazy-evaluate their selections by only
storing an identifier of the most recent transaction they were based
on, see bug 47343.
Change-Id: I18bb986001a44732a7871b9d79dc3015eedfb168
*.php
* Added links to new file
ve.ce.ImageNode.js
* Added focusable node mixin
ve.ce.FocusableNode.js
* New class!
* Adds isFocused and setFocused methods
* When a node is focused or blurred, 'focus' and 'blur' events are emitted
* While a node is focused, it will have the 've-ce-node-focused' class added to it's this.$
ve.ce.Surface.js
* Add detection of node focusing and setting focus and blur on nodes on change
Change-Id: I3f1ad6309571f2bfe568550e2e8f1bd5a0302085
Before, it took an array of objects and translated those to indexes
using the store. Literally every caller outside of the test suite got
an array of indexes from the linear model, translated those to objects,
then passed them into the AnnotationSet constructor which translated
them right back to indexes.
The previous behavior was kind of ridiculous on its face, but the
reason we found it is because Inez was investigating the performance
degradation when bolding a line and found that half of it was due
to the hundreds of ve.getHash() calls caused by this behavior.
Change-Id: I38df8ae9f6392849dacf477ea2f804283c964417
Subbu said that cloning of attributes like data-parsoid or typeof would
cause problems for Parsoid.
Also remove the attributes object if it becomes empty, and do the same
for the internal object.
Bug: 47297
Change-Id: I428becf95c70d0ed8af5b0c408e3966dc47fd8c3
Gabriel noticed whitespace diffs in the problem reports. These were
caused by browsers' .innerHTML being broken on <pre>s. We compensate
for this in the converter, but not when generating originalHtml.
However, a variable called originalHtml really shouldn't be generated,
it should just literally be the original, unprocessed HTML. This
does mean it includes the doctype, <html> and <head> which aren't
included in editedHtml (that one's just the contents of the <body>), but
that's much easier to deal with on Parsoid's end than random newline
diffs all over the place.
Change-Id: I8e66cb79887f49f84114ab6b4d0e0d24aea744b6
Context-sensitive nodes are HTML elements like <caption> that can only
appear in certain contexts (<table> in this case). This means that
serializing them by throwing them in a <div> and calling .innerHTML
fails, because the browser knows a <caption> can't be in a <div> and
unwraps it. jQuery's .html() function is clever though and knows to wrap
<caption> in a <table> (and has similar rules for other elements).
So use jQuery's .html() rather than manual .innerHTML stuff.
Change-Id: Id7d3eff968b3a2ba345680772f7cc32e3dcdb529
In HTML like <td>Foo <b>Bar</b></td>, the space would be stripped and
registered as trailing whitespace in the <td>, so it wouldn't be visible
in the editor and would be inserted after the </b> on the way out.
Thanks to Subbu for reporting this. This case was kind of ridiculous and
we're lucky the JRuby article contained it. To trigger the bug, you had
to have:
* a table cell
* containing unwrapped content
* consisting of
** some text
** whitespace
** open annotation (bold in my example, link in Subbu's case)
** text
** close annotation
** and nothing else
Change-Id: I2b83f02764b311a32a50956d4c8930a9394e91a4
This was broken ever since the introduction of IndexValueStore, because
the call to getDataFromDom() wasn't updated, so it crashed with a "doc
is undefined" error. Fixing part of this by passing in a new IVStore.
The data that is transmitted over the wire still has indices with no way
to find out what the corresponding annotations are. It needs to be fully
expanded but there's no way to do this in DM quite yet.
Bug: 47319
Change-Id: I761523d22e51ac560e37ae991d01a6b84224ca40
This gets us some new functionality, mainly the ability to get a position
from a point on screen.
Changes:
* Updated core and position
* Removed unused modules
Change-Id: I2145fe549c975fdbbcd7dfcf144afb26509d0050
This helps with some differences between browsers, specifically Firefox,
where e.which would be "1" even if the mouse button was not pressed.
Change-Id: Ia88449c2bc84073d903dc702bb586127618d86f2
Apparently Parsoid has always required this, and we've never sent it,
yet somehow the code in production works. This may well be the cause
of some of the selser issues we saw after the deployment attempt in
January.
Made oldid a required parameter in the API module, and default it to 0.
When we get 0, we translate that to the empty string for Parsoid's
benefit. We also need to explicitly get wgCurRevisionId in
ViewPageTarget, and that's also 0 on new pages.
Change-Id: I3a55025246014cd74e15d6d5b6c4ede7b823e5df
