The SurfaceObserver lock guarded against setTimeout calls, but the lock
was only ever used synchronously.
Remove 'lock' and 'unlock' events. Instead, re-sync the SurfaceObserver
by listening to the 'documentUpdate' and 'select' events.
Signed-off-by: Roan Kattouw <roan.kattouw@gmail.com>
Change-Id: Ib15c39f3d25677da70625581b3b2765ae66994b4
...or really changeInternal(), so we can avoid adding undo transactions
to the undo stack.
Also get rid of the pattern where undo() and redo() return a selection
which the caller then has to restore, and instead just restore the
selection.
Bug: 53224
Change-Id: If5a3b4d4162e9f0713ee9cd26e79a66efe52770f
* Replace surface 'transact' event with 'documentUpdate' event
* Have surface listen for all document transactions and update selection
as appropriate (as well as emitting 'documentUpdate')
* Implement change() in terms of setSelection()
** Queue 'contextChange' events so contextChange is only emitted once
** Use this.transacting flag to prevent setSelection() (which is called
because the model emits transact events) from doing too much
** Behavioral change: lock/unlock now emitted separately for
transaction and selection changes
Change-Id: I44425d260ec70758f99d13f99e41f5c993d260c2
ve.dm.Surface.js:
* Stop emitting 'change' and remove its event documentation
ve.ce.Surface.js:
* Listen to 'select' instead of 'change'
* Perform a CE surface update after model-based keydown handling
ve.dm.Surface.test.js:
* Stop asserting that 'change' is emitted
Change-Id: I8f16289493e835d890709c6dfe093d04c18522b6
It was previously emitted before the selection was updated and with the
old selection as a parameter. Instead, emit it afterwards, and make sure
it's emitted even if the selection changes because it was translated
for a transaction.
Also correct its event documentation, which seems to have been copied
from a UI class somewhere.
Change-Id: I521eff0095959572587c0ecffd24dbf322e12d82
Instead of using @emits in both, use our custom @fires in
production (JSDuck 4), and in the future it'll just naturally
use the native one.
This way we can also index oojs without issues, which seems to
have started using @fires already.
Change-Id: I7c3b56dd112626d57fa87ab995d205fb782a0149
The default is to store the entire generated DOM node, but in
general classes can store anything that can be held by the store.
Change-Id: Ia761079fadfb5a6cfa2f00e5b5e23d6c6d3468ac
* Moved isNodeFocusable to ve.ce.NodeFactory
* Added isFocusable static property to ve.ce.Node
* Set isFocusable to true on ve.ce.FocusableNode
Change-Id: I3cf666280abdfce55bf9b0710827bb25c40bfd51
Fix things that 4aa86d0f8 broke:
* Update surface parameter to windowSet in all ve.ui.SurfaceDialog subclasses
* Do the same for ve.ui.SurfaceInspector subclasses
* Fix @extends documentation for SurfaceDialog
* Fix documentation for ve.ui.SurfaceInspector, copypasta from SurfaceDialog
Bonus:
* Add .getMetaList() getter to dm.Surface
Change-Id: I843e99e45e9b013cb9cb559f050384d39bbbddf2
Previously we returned ElementLinearData from the converter, then
stripped out the MetaLinearData. This meant that before processing
the ElementLinearData from the converter actually contained metadata
which is confusing.
The new document constructor stores the converter results in a
FlatLinearData object and simultaneously populates element and meta
data stores.
Also in this commit I have moved various methods from ElementLinearData
to FlatLinearData, from which ElementLinearData inherits.
Change-Id: I64561bde2c31d8f703c13ac7b0a0c5f7ade9f3d4
Objective:
* Make ve.Factory behave like ve.NamedClassFactory
* Remove the only remaining use of ve.Factory (actions)
* Remove ve.NamedClassFactory
Change-Id: Ie302ef5ea31081de7ab0db6091058a59946aef4c
InternalList.clone() assumed that all properties are automatically rebuilt
when a new document is built, but that's not true for .nextUniqueNumber
(or for .itemHtmlQueue for that matter). This meant that, in practice,
.nextUniqueNumber was being reset to 0 after auto/N numbers for existing
references had been assigned, but before assigning numbers to newly
created references. This caused all sorts of naming collision fun.
Bug: 54712
Change-Id: I1d087a5f3c23979d7d488e3ab32eb064ebc23e94
Document slice only ever contained linear data, with extra functionality
to preserve the range. It pre-dated LinearData, but now we should
refactor it to reflect its purpose.
Change-Id: Ifc908f7526c83a43a51372c8d2494d7260e7facd
We already getSlice which returns a ve.dm.DocumentSlice, so using
the word slice in this method is very confusing. What we are actually
doing is creating a ve.dm.Document from a range. Also remove argument
overloading as it's not particularly helpful and would make the new
name a lie.
Change-Id: I93da3419510410b170396e6765fbe2a87f9795be
The way we implemented undoing transactions was horrible. We'd process
the original transaction, but with a reversed=true flag. That meant we
had to keep track of the 'reversed' flag everywhere, and use ternaries
like insert = reversed ? op.remove : op.insert; all over the place to
access transaction operations. Redo then worked by reapplying the
transaction. We would verify that this was OK by tracking whether the
transaction was in an applied state or an undone state.
This commit makes it so every transaction can only be applied once. To
undo, you obtain a mirror image of the transaction with tx.reverse(),
then apply that. To redo, you clone the original transaction with
tx.clone() and apply that. All the code that had to use ternaries to
check whether the transaction was being applied in reverse or not is
gone now, because you can only apply a given transaction forwards,
never in reverse.
Bonus:
* Make ve.dm.Document's .completeHistory a simple array of
transactions, rather than transaction/boolean pairs
* In the protection of double application test, clone the example
document properly; it modified ve.dm.example.data, which was "fine"
because it ran .commit() and .rollback() the same number of times
Change-Id: I3050c5430be4a12510f22e20853560b92acebb67
Logic was failing because we were passing the index of the annotation
within the AnnotationSet, instead of the index within the Store, to
containsIndex().
Bug: 54332
Change-Id: Ibfd9abe6e4b44d9db744e0c5019418eee12f84a4
Replaces newFromNodeReplacement(). newFromNodeReplacement was very
simplistic and didn't support metadata or internal list items, so
if you had comments or references inside of the data you were editing
(reference contents or an image caption), they'd get mangled.
With this, you can do:
newDoc = doc.getDocumentSlice( node );
// Edit newDoc
tx = ve.dm.Transaction.newFromDocumentReplace( doc, node, newDoc );
surface.change( newDoc );
and that takes care of metadata, internal list items, and things like
references that reference internal list items.
ve.dm.Document.js:
* In getDocumentSlice(), store a reference to the original document
and the number of items in its InternalList at the time of slicing
in the created slice. This is used for reconciliation when the
modified slice is injected back into the parent document with
newFromDocumentReplace().
ve.dm.InternalList.js:
* Add a method for merging in another InternalList. This provides a
mapping from old to new InternalList indexes so the linear model data
being injected by newFromDocumentReplace() can have its InternalList
indexes remapped.
ve.dm.Transaction.js:
* Replace newFromNodeReplacement() with newFromDocumentReplace()
ve.ui.MWMediaEditDialog.js, ve.ui.MWReferenceDialog.js:
* Use getDocumentSlice/newFromDocumentReplace for editing captions/refs
* Change insertion code path to insert an empty internalItem/caption, then
newFromDocumentReplace into that
* Add empty internalList to new mini-documents
ve/test/dm/ve.dm.Transaction.test.js:
* Replace newFromNodeReplacement tests with newFromDocumentReplace tests
ve-mw/test/dm/ve.dm.Transaction.test.js (new):
* Add tests for newFromDocumentReplace with mwReference nodes
ve.dm.mwExample.js:
* Add data for newFromDocumentReplace with mwReference tests
VisualEditor.hooks.php:
* Add new test file
Bug: 52102
Change-Id: I4aa980780114b391924f04df588e81c990c32983
Documentation says third argument is 'data' when in fact it is
'insertion' which is then re-used as the result of fixupInsertion :/
Change-Id: I17c959f858eddddc90a6fa839da0d32da69a784f
Fixing a couple of mishaps in the language annotation, including adding a
getComparableObject in the ve.dm.LanguageAnnotation class.
Change-Id: Ia28900df55969f60576cfd716c70ffc1ba5f4c27
Conversions usually end in a close tag so this hasn't been a problem
before, but the mini-conversion we do in onCopy may not be balanced.
Change-Id: Ia2db29f116ca84ee886b4c5cafd4ac45cd926b79
Simply generating ':3' as the "unique" name for the 4th reference
doesn't work. Even if getUniqueListKey() had been used, that only
checks for conflicts with names that have already been encountered
(i.e. occur in <ref> tags that precede the current one), not for
conflicts with names that first occur further down in the document.
The solution is to generate names at serialization time, when we
have full knowledge of which names are in use. Internally, we use
'literal/<name>' for names that literally appeared in the source,
and 'auto/<number>' for unnamed references. Then at serialization
time, we translate 'auto/<number>' to 'literal/:<number>' if needed
(i.e. if the reference was reused).
ve.dm.MWReferenceNode.js:
* toDataElement()
** Prefix listKey with literal/ or auto/ as appropriate
* toDomElements()
** Map auto/ listKeys to unique names
** Don't try to unset the name if not present (was unsetting a property
that didn't exist anyway)
ve.dm.InternalList.js:
* Remove now-unused isUniqueListKey()
* Rewrite getUniqueListKey()
** Make prefix configurable
** Take previously generated unique keys into account
** Map the same old key (auto/N) to the same generated key (literal/:M)
* Add getNextUniqueNumber() as a source for auto/N numbers: previously
used the length of the itemHtmlQueue, but that only works during
conversion, not from the UI dialog
ve.ui.MWReferenceDialog.js:
* For new references or conflicting names, generate an auto/N key and
let toDomElements() deal with actually mapping that to name
ve.dm.InternalList.test.js:
* Rename listKeys to new style
* Split the test case into two groups so we can test multi-group cases
* Add tests for getUniqueListKey()
ve.dm.mwExample.js:
* Rename things to new style
* Modify the test case so it attempts to trigger bug 54341
Bug: 54341
Change-Id: I726fb83e6fb66ffec643d996768a854ec9474b3d
Correct the return value descriptions for the `newFromMetadataInsertion` and
`newFromMetadataRemoval` methods (which appear to have been cut-and-pasted
from another method).
Change-Id: I019b0ef9c75ff47fc0f2df8c95da9d090710a5a8
Code with a similar purpose was added in 568e0e5701 but got lost
when some things were moved from ve.Surface to ve.ce.Surface in
5012ed10.
Initializing the selection at (0,0) was known to cause problems before,
and since 789d0caf09 breaks editing of empty documents: typing in an
empty document begins in an inline slug, but SurfaceObserver doesn't
notice typing in an inline slug unless the ce.Surface pawns it, which
is OK because insertions in slugs are always pawned, but the pawning
logic believes the cursor to be at offset 0 where there is no slug
(it's at offset 1) and so it doesn't pawn.
Bonus: update tests and add descriptions for dm.Surface.change tests
Change-Id: Id72314d0fe650dacc7cdb842f5cea2f3bfba5145
Previous was checking by string match to 'preformatted' which didn't
match 'mwPreformatted'. Now using node factory to see if
hasSignificantWhitespace is set.
Bug: 54083
Change-Id: I1be2d4568df7848e08074d200fb374acd508f6fa
The previous implementation couldn't deal with transactions that
replaced both data and metadata at the same time (rather than replacing
data while moving metadata around), and extending its approach to
deal with that case would have made it much more complex.
So I rewrote the algorithm from scratch. The previous implementation
scheduled deferred moves for existing items, but immediately processed
insertions and removals. This is problematic for replacements and
maintaining the order in the binary search list. So instead, this new
implementation builds an array representing what the new item list
should be, then processes insertions, removals and moves in the correct
order to achieve that state.
It looks like the previous implementation didn't always work correctly,
which was masked because the test suite passed full=false to
assertItemsMatchMetadata(). This rewrite fixes this.
Also remove setMove/applyMove from MetaItem, because we don't need them
anymore and they're evil anyway; and add isAttached(), because the new
algorithm needs it.
Change-Id: I899d2b3c94c2cfa55823879bca95456750f64382
This was already possible in the operation structure by using
removeMetadata and insertMetadata, but no transaction builders
generated transactions like that.
With this change, pushReplace() can be instructed to overwrite the
metadata with arbitrary data rather than collapsing it by passing the
optional insertMeta parameter.
Change-Id: I0272fe38e053161c738591cf0a7b447a0827d4dc
Transactions that replaced metadata twice at the same offset
(with retainMeta-replaceMeta-retainMeta-replaceMeta) were broken
because the processor for replaceMetadata didn't advance the
metadata cursor.
Change-Id: I7ad24e7ffb4c39b40ec9c347db301f8e28f3692d
Add some test cases for documents with trailing metadata, and fix an
off-by-one error in the metadata-mutating transactions (since the
document metadata array is one larger than the document data array).
Change-Id: I8f049466e03ed55010dfcf0a35702536edfa7b0a
These are not considered errors in jsduck 4.x but are in 5.x.
Though we don't specifically want to support 5.x yet, these are
actually genuine errors that were already causing incorrect
documentation to be generated, even under 4.x (it just doesn't
warn for these under 4.x).
Not from 5.x warnings, but fixed while at it:
* Removed more redundant @method
* Added a missing @returns for ve.Range#flip
* Added newline at EOF
Change-Id: Idc22b46b082a813e8e1761403e1ebf908bf690c4
This version pushes a `replaceMetadata` operation after a `replace` to
fixup trailing metadata if there is no inserted region and the removed
region includes metadata. This avoids a corner case where the
size of the metadata arrays inserted/removed in `Transaction.pushReplace()`
do not match the size of the data arrays inserted/removed.
We remove the now-unused `Document.getMetadataReplace()` method.
We also adjust `MetaList.onTransact()` so that it continues to work
properly when the number of metadata entries in `replace.insert` is
not the same as the number of metadata entries in `replace.remove`.
Change-Id: I1d600405b855ca1cb569853bb885b0752df47173
The `Transaction.pushReplace` method has a corner case if the removed
region has metadata and the inserted region is empty. This works fine
unless there are two adjacent `pushReplace` operations, which can occur
in `Transaction.newFromUnwrap`. Fix this by having `pushReplace` look
at a preceding replace and correctly merge the two operations if
possible (in particular in the tricky case where the previous case has
a zero-length insertion). Pleasantly, this can be done without a lot of
special-casing code in `pushReplace` or `newFromUnwrap`.
Add test cases verifying the `newFromUnwrap` works correctly (both
in commit and in rollback) when there is metadata present.
Change-Id: I6cfec0d2b1823dad724422f018a3c73dc0c7f186
CODING.md
* Document the procedure for adding a new javascript class
ve.dm.MWExternalLinkAnnotation.js
ve.dm.MWInternalLinkAnnotation.js
ve.dm.ElementLinearData.js
ve.dm.LinearData.js
* Add whitespace line before preformatted code to fix a
rendering bug
Change-Id: I54443ea3d4799328655d279f379d4ddc176c50a0
ve.dm.MetaLinearData.static.merge( [ undefined, undefined, ... ] )
returned [ [] ], make it return [ undefined ] instead. This allows
us to lazily use the merge function to create metadata without having
to worry about denormalizing undefined to [] all over the place.
Change-Id: I41fe6472c2185a6315119a546a73765966d013d3
Avoid making a copy of the entire data array by using `doc.data` and
`doc.data.getLength()` instead of `doc.getData()` and
`doc.getData().length`.
Get rid of some unnecessary conditionals in `Transaction.newFromWrap`: the
`tx.pushReplace()` and `tx.pushRetain()` methods will already gracefully
no-op if the replaced/retained region is empty; we don't need to add
extra control flow to handle that case.
Clarify offset math in `Transaction.newFromWrap`.
Fix a minor bug in `Transaction.newFromInsertion`: the final retain didn't
account for the length of the removed region, if `doc.fixupInsertion`
creates one. (This usually doesn't happen.)
Remove an inaccurate TODO in `Transaction.pushReplaceMetadata()` -- it is
no longer a straight copy/paste of `Transaction.pushReplace`, so a refactor
isn't really called for.
Change-Id: I7d86a2449978365d69d4a5ed43116c2e9945475d
Objectives:
* Got rid of mw prefixing in tools, inspectors and dialogs
* Simplify tool classes so they can be generically used as items in bars, lists and menus
* Add support for a catch-all toolbar group
* Simplify tool registration, leaning on tool classes' static name property
* Move default commands to command registry
* Move default triggers to trigger registry
* Get language tool working in standalone
Change-Id: Ic97a636f9a193374728629931b6702bee1b3416a
In ve.dm.Document.getMetadataReplace(), we used to only merge metadata
if the amount removed is larger than the amount inserted. But this
could end up putting metadata in odd positions, for example if you
have Foo[[Category:Bar]]Baz and you delete 'ooBa' and replace it with
{image}xxx{/image}, then the category ends up inside the image.
We should always merge metadata when a segment is deleted, so that it
appears outside any new structure added.
There's a weird corner case here when a segment is removed but no
insertion is made: the removed metadata then needs to get glommed onto
the next element. We extend the insert/replace metadata array
when this happens.
Bug: 53444
Bug: 53445
Change-Id: I51d55fb370b473273f9cf152fdd0f356377d4109
This avoids problems when unnamed references were copy-pasted.
Knowing that key is always non-null simplifies a lot of logic
elsewhere.
Bug: 53365
Change-Id: I3a23123ae732d9583814d38dd880a0cdf691fd5d
ce.AlienNode was mixing in GeneratedContentNode, but wasn't actually using
it properly. This commit fixes this so that AlienNode can benefit from
the <meta>/<link>/<style> stripping code in GeneratedContentNode.
Also added a getHashObject() function to dm.AlienNode to summarize the
domElements attribute because ve.getHash() chokes on it otherwise.
Change-Id: Ief3be94f9730297abe0e3c57506b81a8ff1d136d
Add a little robustness, guaranteeing that we don't end up with multiple
history tracking tasks running, leaking one, or try to clear a non-running
interval.
Change-Id: I41db2d6fefc7f45f150aa14ecefc648760ad6200
The selection property is never null; it is initialized to Range(0,0).
If it is set to null in purgeHistory(), the next call to
ve.dm.Surface.change() will crash.
Change-Id: Ia45c0ba26291e8ad09c445fdf2323710b5ab409f
Was previously broken as getSlice was using ve.Range#equals to
compare ranges which is direction-sensitive.
Bug: 51538
Change-Id: Ib58d1d8fd11b62388c111a5da66171d13a9db9c2
Currently ignores all non-element data, but element content (e.g.
images) can be annotated.
Added test cases and updated the test runner to only compare
store indexes for a more readable output.
Bug: 50127
Change-Id: I234586a28072811c8288aab56f6abaaa0da0c88d
This isn't a problem when using the UI buttons as they get disabled
but the command keys can still trigger these methods.
Also fix hasPastState to include check for small stack. This fixes
an existing bug where the undo button doesn't become active until
~1s after the first change is commited (i.e. after the small stack
is committed to the big stack).
Bug: 52113
Change-Id: Idbd34953c805620881a609409290256462af80a5
GeneratedContentNode didn't track concurrent updates at all, so a
race condition was possible: if the node was updated a second time
before the first update had been rendered, the second update might
render first and then be overwritten by the other one.
To prevent this, we track the promise associated with the current
render. If a new update is launched while a previous one is still
pending we attempt to abort the old one by calling .abort() on it,
and ignore any future resolution or rejection from it.
Also allow rerenders based on non-model data by calling
.update( { config object } );
Change-Id: I8feefd9e8fb6c41d06b8b20131e3be5e37954e83
If you had an HTML element that was matched by two models, one with
a direct string match and one with a regex match, then the string
match would beat the regex match (which is correct) if they both
specified a tag name, but the regex match would win (which is wrong)
if they both didn't specify a tag name.
The fix is to only check for tagName === null if we're in tag-agnostic
mode (tag === '').
Change-Id: I9943611111e4c4ff498cdd95b7b3e72f95fb413b
This allows abstract classes to specify RDFa types based on a static
property overridden by a child class. The default implementation is to
just use .static.matchRdfaTypes.
Change-Id: Ic71fc552a6a1626d94f998e9517af971e8198e79
This is bit of a hack, as leading whitespace could be
significant if styled with white-space:pre.
Long term VE shouldn't be editing the user's HTML, and
should just highlight potential formatting issues.
We avoid the stripping in preformatted elements as we
expect they will have that styling.
Bug: 51462
Change-Id: I654d98e17dd604cb2a192831ff3f3597f95b9962
Not having a description yet is fine, but they should at least
be indexed as blocks so that they are searchable and listed
in the jsduck generated pages. jsduck defaults to @method + name
of prototype property. And it even guesses parameters sometimes.
Search: \n\n([a-zA-Z\.]+\.prototype\.[a-zA-Z]+)
Where: modules/ve,modules/ve-mw
Where-Not: modules/ve/test
Replace: \n\n/** */\n$1
Added @return in a few places where it was easy to add.
Change-Id: I830c94cc7dbc261bd7a077391f930cbfff165f9d
op !== operation in all iterations of this loop except the first,
and using information from one operation to apply another one
doesn't work very well.
The fact that undefined cast to a number is NaN rather than 0
is very unhelpful :(
Also fixed some commas that should be semicolons.
Bug: 52238
Change-Id: I4138c023c955f2866881084506e24bb8b6db5a4d
These have been pointing to the same method for a while now,
we can safely remove these obsolete aliases and just use it
as generic copy.
* Each file touched by my editor had its new line at EOF fixed
where absent
* Don't copy an otherwise unused empty object
(ve.dm.Converter)
* Use common ve#copy syntax instead to create a link
(ve.dm.Document, ve.dm.example)
* Remove redundant conditionals for isArray/copyArray/copyObject
(ve.dm.example)
Change-Id: If560e658dc1fb59bf01f702c97e3e82a50a8a255
Annotations' attributes might contain DOM elements, which cause infinite
recursion in ve.compare(). Annotation classes can protect against this
by overriding getHashObject() to summarize DOM nodes, but that doesn't
help if that's not respected everywhere.
Instead, compare the hash objects, those are safe. This does not appear
to be a problem in practice, currently, because the nowiki annotation
is experimental, oo.compare() now short-circuits if a === b, and because
of optimizations in openAndCloseAnnotations() which lead to the relevant
compareToForSerialization() code path being taken very rarely.
Bug: 51948
Change-Id: If1bcc3eee4fd14d107db1935d89dcc5516643b53
By using annotation indexes only we can avoid a lot of
ve.getHash calls. This reduces the number of getHash calls
on load of [[:en:Argentina]] from ~60,000 to ~2,000.
Bug: 52013
Change-Id: I0bc9aa8feea5f7e4e90a5fcd829de57cab803c15
We would dirty-diff "</span>\n<!-- comment -->\n<span>" to
"</span>\n\n<!-- comment --><span>", i.e. the second newline made
a bunny-hop to the left over the comment.
The actual bug turned out to involve a double bunny-hop, with
"</span> <!-- comment -->\n<span>" turning into
"</span>\n <!--comment --><span>", i.e. the newline bunny-hops
both the comment and the space.
This happened because outputWrappedMetaItems() didn't take
wrappedWhitespace into account when restoring meta items and
associated whitespace. I hacked a check for wrappedWhitespace into it,
but we should really just rewrite this pile of hacks into a unified
system for queuing and processing both whitespace and metadata.
Change-Id: I4375f4c07983ffec6877d0371aeaa9bf6e65fd6e
