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
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
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
The first run of getDomFromData() would preserve whitespace just fine,
but it blanked out the .veInternal.whitespace[1] element in certain
cases, contaminating the linear model and making the whitespace data
inconsistent. Subsequent runs of getDomFromData() would then refuse to
serialize that whitespace because the information about it was
inconsistent.
In getDomFromData(), we sometimes unset .veInternal.whitespace[1] (i.e.
set it to undefined) to prevent double processing. Because we're
potentially going to modify .veInternal, don't assign it by reference,
but copy the object.
Added tests asserting that the linear model is unchanged after calling
getDomFromData(), because that function should never modify
linear model data. This test failed in 4 cases (all whitespace-related)
before I added the copyObject() call.
Bug: 43543
Change-Id: Ic4c93510518163894201a693ab50331413715967
ve.dm.Transaction
* Replace operations are now built directly from the
linear model and automatically determine what metadata
replace information they need to include:
** retainMetadata
** replaceMetadata
** insertMetadata
ve.dm.Document
* Metadata array created empty and padded out after data parsing
as we are no longer using Document.spliceData to build it (a new
test checks for correct metadata length)
* spliceData replaced with getMetadataReplace, which instead returns
transactional steps of spliceData (retain, replace, insert)
ve.dm.MetaLinearData
* Add function for merging metadata items together. Only used
once in the code (Document.getMetadataReplace) but useful
for generating test data.
ve.dm.MetaList
* Replace operations with metadata need to calculate new offset
and indices directly, but can't be applied immediately lest they
put a metaItem out of place and affect findItem.
ve.dm.MetaItem
* Add methods to support queued moves as required by MetaList
Test files
* Updated to match new pushReplace API
* Remove any instances of Document.spliceData
* Extra check on sparse metadata array length
* Rewrite spliceData tests as getMetadataReplace tests
* Count expected cases in Transaction(Processor) tests
Bug: 46954
Change-Id: I4edad1c2dd37c723bff2792bab7d694ef17a86dc
Inez reported that unitalicizing from the toolbar was broken, because
the toolbar was somehow generating annotations that had .attributes={}
as opposed to .attributes=undefined. Turned out the cause was in the
default value for element in the ve.dm.Model constructor.
Change-Id: I64ea9ef56cd15d1131c1aa23484d7420c95a8225
* Create MWTemplateBlockNode & MWTemplateInlineNode (in ce and dm)
* Move Alien's 'isInline' code to ve.dm.Node.static.isHybridInline
* Move definition of ce.AlienBlock/InlineNode inside ce.Aline.js file
to match dm.AlienBlock/InlineNode and MWTemplate
* Duplicate AlienBlock/Inline styles for templates
* Create test case for inline templates
* Count test cases in ve.dm.Converter.test.js automatically
Change-Id: Id9bc7f049ea974dd5e7f8b7a66080939e0948bbd
Use ve.BranchNode.getNodeFromOffset instead of
ve.dm.Document.getNodeFromOffset so correct nodes will be retrieved to
emit update events to.
Change-Id: Iaf559f0424584a3dde065e548e403c4a53207312
Change 57076 implemented the functionality required to skip
words in UnicodeJS. This change simply removes all the existing
code we had and replaces it with this implementation.
Bug: 46794
Change-Id: I6b2700d65476c4d34ba4a01a88382d7af8e736fb
This node stores the rendered in the index-value store, hashed on
a custom hash of the dm (type + mw) which makes it unique it its
parameters.
Bug: 46571
Change-Id: I0ab4c9f7bca207121d5b42e83c821771b6139da8
For data->DOM, this is easy: .toDataElements() can optionally return an
array instead of an object, and that will be treated as the data to
insert. If this happens, the converter won't descend. The node handler
can recursively invoke the converter if it needs to (although I suspect
the current implementation is broken when converting block content in an
inline context).
For DOM->data, this is a bit more complex. The node sets
.static.handlesOwnChildren = true; , which triggers the converter to
pass a data slice rather than a single data element, and not to
descend. The node handler can invoke the converter to recursively
convert DOM subtrees to data.
ve.dm.Converter (data->DOM):
* Renamed createDataElement() to createDataElements()
** .toDataElement() may return element or array, handle this
* Renamed childDataElement to childDataElements, is now an array
* Actually alienate if .toDataElement() returns null
** Shockingly, this claimed to be supported before but wasn't
* Rather than pushing to data, concat to it
** Add closing if needed
* Don't descend if .toDataElement() returned an array of length >1, or
if the node has .handlesOwnChildren = true
ve.dm.Converter (DOM->data):
* Split getDomSubtreeFromData() and getDomFromData()
* When converting a node that handles its own children, pass in a data
slice and skip over that data
Change-Id: I196cb4c0895cbf0b428a189adb61b56565573ab3
A node could already implement a toDataElements() function that
returns a data element of another node type, but it couldn't return
an annotation or a meta item. This is fixed now, and any dm.Model
subclass can now morph into any other dm.Model subclass.
I didn't originally plan to do this today at all, but doing this now
makes my upcoming converter changes easier. Surprise feature!
Change-Id: Ief6ac302094df084221a5a97c32a522b929c2960
This will allow toDataElement() functions to just call this function
with a DOM element, rather than having to have all the recursion context
data to pass in.
Also expose this information using getters.
Change-Id: I89574c42385267e08704f018c0892d63014376a6
ve.ce.ImageNode.js
* Moved in generic stuff from MWImageNode
* Added drag end handler (empty, will be used soon)
ve.ce.MWImageNode.js
* Changed to inherit ImageNode
* Moved generic stuff out
ve.dm.ImageNode.js
* Added attribute extraction/preservation for src, width and height
ve.dm.MWImageNode.js
* Changed to inherit ImageNode
* Re-using ImageNode's attribute handling to extract/preserve attributes on both the image and wrapper level
Change-Id: Ied4e1ece24e6804220eac35330790f7084df55de
It's been passed in for a while, but nothing ever used it. As we know
some browsers don't like it when we create elements in the wrong
document, and this ensures we always use the correct document for
createElement().
Change-Id: Ia3d2fabe0516956105ad2b5625ed2f76c015c26e
ve.dm.Converter and ve.ce.ContentBranchNode were duplicating a fair bit
of logic for annotation rendering. Moved the annotation opening and
closing logic into ve.dm.Converter.openAndCloseAnnotations, and
implemented both annotation rendering code paths in terms of that
function with callbacks for caller-specific behavior.
Change-Id: I7cba7d2fda7002287b07949a1b8120ba80bfe854
This provides the functionality for keyboard word skipping
(i.e. pressing ctrl/alt + arrow key).
Bug: 46794
Change-Id: Ib0861fa075df805410717a148b8a6e166d947849
We weren't really using it exclusively for nodes any more, and the only
functionality in there was for using .static.name
Change-Id: Ie26928cd01faee95a10912201663b45f1f20fb19
Just like ve.dm.Model is a common base class for dm.Node, dm.Annotation
and dm.MetaItem.
ce.View abstracts the this.model and this.$ behavior, including liveness,
whitelisted HTML attribute rendering and adding a back reference in
.data(). The back reference has been renamed from .data( 'node' ) to
the more generic .data( 'view' ).
At this point this means ce.Annotation is just a shell around ce.View
(except where it defaults to a span rather than a div), but that could
change in the future.
Change-Id: I0eef5b80718e0b0fcd3f8bba096b452f0bb680d0
ve.dm.Model is now the common base class for these three. ve.dm.Node
inherited from ve.Node before, so it now uses it as a mixin instead.
This required changing ve.Node's usage of ve.EventEmitter from
inhertiance to a mixin as well, because inherited methods apparently
don't get mixed in correctly.
* Change annotation terminology from linmodAnnotation to element for
consistency with Node, MetaItem and Model
* Reimplement getClonedElement() in Node for .internal treatment
Change-Id: Ifd3922af23557c0b0f8984d36b31c8a1e2ec497e
This changes the annotation API to be the same as the node API, sans
a few boolean flags that don't apply. The APIs were different, but
there was really no good reason why, so this makes things simpler for
API users. It also means we'll be able to factor a bunch of things out
because they're now duplicated between nodes, meta items and annotations.
Linear model annotations are now objects with 'type' and 'attributes'
properties (rather than 'name' and 'data'), for consistency with elements.
They now also contain html/0/* attributes for HTML attribute preservation,
which obsoletes the htmlTagName and htmlAttributes properties.
dm.Annotation subclasses take a reference to such an object and implement
conversion using .static.toDataElement and .static.toDomElements just
like nodes do. The custom .getHash() functions are no longer necessary
because of the way HTML attribute preservation was reimplemented.
CE rendering has been moved out of dm.Annotation (it never made sense to
have CE rendering functions in DM classes, this was bothering me) and into
separate ce.Annotation subclasses. These are very similar to CE nodes in
that they have a this.$ generated based on something in the DM; the main
difference is that nodes listen to events and update themselves, whereas
annotations are static and are simply destroyed and rebuilt when they
change. This change also adds whitelisted HTML attribute rendering for
annotations, as well as class="ve-ce-FooAnnotation" attributes.
Now that annotation classes produce real DOM nodes rather than weird
objects describing HTML tags, we can't generate HTML as a string in
ce.ContentBranchNode anymore. getRenderedContents() has been rewritten
to be much more similar to the way the converter renders annotations;
in fact, significant parts of it were copied from the converter, so that
should be factored out in the future. This change actually fixes an
annotation rendering discrepancy between ce.ContentBranchNode and
dm.Converter; see the diff of ve.ce.ContentBranchNode.test.js.
ve.ce.MWEntityNode.js:
* Remove stray property
ve.dm.MWExternalLinkAnnotation.js:
* Store 'rel' attribute
ve.dm.TextStyleAnnotation.js:
* Put all the conversion logic in the abstract base class
ve.dm.Converter.js:
* Also feed annotations through getDomElementsFromDataElement() and
createDataElement()
ve.dm.Node.js:
* Fix undocumented property
ve.ce.ContentBranchNode.test.js:
* Add descriptive messages for each test case
* Compare DOM trees, not HTML strings
* Compare without all the class="ve-ce-WhateverAnnotation" clutter
ve.ui.LinkInspector.js:
* Replace direct .getHash() calls (evil!) with ve.getHash()
Bug: 46464
Bug: 44808
Change-Id: I31991488579b8cce6d98ed8b29b486ba5ec38cdc
* In matchTypeRegExps(), skip string types
** Didn't break because we currently have mixes of strings and regexes
* Combine types from rel, typeof and property rather than picking one
** Added test case in ve.dm.example that resembles actual Parsoid output
* If the element has extension-specific types, not only restrict type
matching to extension-specific types, but also require that *all* types
present on the element be matched
Change-Id: Iacf3851a0ca9081d2c813b42435484a47cec6230
As described in the bug, ve.getHash performs JSON.stringify so to
customise a hash the object should just return an object to be
hashed, not the hash string itself.
Bug: 46895
Change-Id: If11071d4b04a01e25102ffb57240882f650ee10d
Created an IndexValueStore class which can store any object and return
an integer index to its hash map.
Linear data is now stored in ve.dm.LinearData instances. Two subclasses
for element and meta data contain methods specific to those data types
(ElementLinearData and MetaLinearData).
The static methods in ve.dm.Document that inspected data at a given
offset are now instance methods of ve.dm.ElementLinearData.
AnnotationSets (which are no longer OrderedHashSets) have been moved
to /dm and also have to be instantiated with a pointer the store.
Bug: 46320
Change-Id: I249a5d48726093d1cb3e36351893f4bff85f52e2
Using element.height was returning 0 if the attribute was empty
when in fact what we mean to store is null (i.e. auto height).
This takes care of the writing of attributes in CE as jQuery
ignores an attribute-set command if the value is null.
Also in this commit I've implemented a basic toDomElements
that outputs the original HTML (code copied from AlienNode).
This stops the code from throwing an exception but will
eventually need to be rewritten to rebuild the HTML from
the attributes stored in the DM.
Bug: 56336
Change-Id: I297a1d0a07e9ebf9d0110fb1cdf266f8415f25b7
