`rightListElement` is not used, so, we can put it into smaller scope.
`leftListElement` is used for storing original `leftBlockElement` when
`newListElementTagNameOfRightListElement` is some. Therefore, we can
stop using it with caching the original `leftBlockElement` before
maybe modified its value.
Differential Revision: https://phabricator.services.mozilla.com/D85531
Now, the new method,
`WhiteSpaceVisibilityKeeper::MergeFirstLineOfRightBlockElementIntoLeftBlockElement()`
is the only user of the method so that we can get rid of it since it does enough
simple thing.
Differential Revision: https://phabricator.services.mozilla.com/D85530
And the name is wrong. It depends on `aPoint` whether the `<br>` element is
visible or invisible because it just scans preceding `<br>` element but
stop doing it when it meets a visible content. Therefore, this patch renames
it to explain what it does.
Differential Revision: https://phabricator.services.mozilla.com/D85528
CLOSED TREE
We don't need these macros anymore, for two reasons:
1. We have static analysis to provide the same sort of checks via `MOZ_RAII`
and friends.
2. clang now warns for the "temporary that should have been a declaration" case.
The extra requirements on class construction also show up during debug tests
as performance problems.
This change was automated by using the following sed script:
```
# Remove declarations in classes.
/MOZ_DECL_USE_GUARD_OBJECT_NOTIFIER/d
/MOZ_GUARD_OBJECT_NOTIFIER_INIT/d
# Remove individual macros, carefully.
{
# We don't have to worry about substrings here because the closing
# parenthesis "anchors" the match.
s/MOZ_GUARD_OBJECT_NOTIFIER_PARAM)/)/g;
s/MOZ_GUARD_OBJECT_NOTIFIER_PARAM_TO_PARENT)/)/g;
s/MOZ_GUARD_OBJECT_NOTIFIER_PARAM_IN_IMPL)/)/g;
s/MOZ_GUARD_OBJECT_NOTIFIER_ONLY_PARAM_IN_IMPL)/)/g;
# Remove the longer identifier first.
s/MOZ_GUARD_OBJECT_NOTIFIER_ONLY_PARAM_TO_PARENT//g;
s/MOZ_GUARD_OBJECT_NOTIFIER_ONLY_PARAM//g;
}
# Remove the actual include.
\@# *include "mozilla/GuardObjects.h"@d
```
and running:
```
find . -name \*.cpp -o -name \*.h | grep -v 'GuardObjects.h' |xargs sed -i -f script 2>/dev/null
mach clang-format
```
Differential Revision: https://phabricator.services.mozilla.com/D85168
We don't need these macros anymore, for two reasons:
1. We have static analysis to provide the same sort of checks via `MOZ_RAII`
and friends.
2. clang now warns for the "temporary that should have been a declaration" case.
The extra requirements on class construction also show up during debug tests
as performance problems.
This change was automated by using the following sed script:
```
# Remove declarations in classes.
/MOZ_DECL_USE_GUARD_OBJECT_NOTIFIER/d
/MOZ_GUARD_OBJECT_NOTIFIER_INIT/d
# Remove individual macros, carefully.
{
# We don't have to worry about substrings here because the closing
# parenthesis "anchors" the match.
s/MOZ_GUARD_OBJECT_NOTIFIER_PARAM)/)/g;
s/MOZ_GUARD_OBJECT_NOTIFIER_PARAM_TO_PARENT)/)/g;
s/MOZ_GUARD_OBJECT_NOTIFIER_PARAM_IN_IMPL)/)/g;
s/MOZ_GUARD_OBJECT_NOTIFIER_ONLY_PARAM_IN_IMPL)/)/g;
# Remove the longer identifier first.
s/MOZ_GUARD_OBJECT_NOTIFIER_ONLY_PARAM_TO_PARENT//g;
s/MOZ_GUARD_OBJECT_NOTIFIER_ONLY_PARAM//g;
}
# Remove the actual include.
\@# *include "mozilla/GuardObjects.h"@d
```
and running:
```
find . -name \*.cpp -o -name \*.h | grep -v 'GuardObjects.h' |xargs sed -i -f script 2>/dev/null
mach clang-format
```
Differential Revision: https://phabricator.services.mozilla.com/D85168
The previous patch changed target of white-space normalization. That caused
`editor/libeditor/tests/test_bug1425997.html` starting to failed with new
exception. What happens in the test is, selection is collapsed into comment
node (I think that this behavior is also a bug though) after second
`document.execCommand("delete")`, and then, failed to delete anything
and throws the exception. For fixing this issue, the most simple approach is
make `HTMLEditor::HandleDeleteAroundCollapsedSelection()` stop handling
deletion when caret position is not editable. So, unfortunately, this
patch changes traditional path too.
Depends on D84948
Differential Revision: https://phabricator.services.mozilla.com/D84949
Blink normalizes preceding white-spaces of deleting range only when the range is
NOT remove last character or there are no following text nodes. If removing
last character at first text node which is followed by another text node, Blink
normalize white-spaces at start of the following text node.
This behavior seems odd, but we should follow the Blink's behavior for better
compatibility.
Note that new test failures in
`white-spaces-after-execCommand-forwarddelete.tentative.html` is caused by
that forward deletion is handled by existing path instead of the new handler.
Therefore, it's no problem for now.
Additionally, this change causes a mochitest
(`editor/libeditor/tests/test_bug1425997.html`) and a crash test
(`editor/libeditor/crashtests/1424450.html`) becoming oranges since their
DOM tree becomes different after applying this patch. The oranges will
be fixed by the following patches.
Depends on D84947
Differential Revision: https://phabricator.services.mozilla.com/D84948
The difference between `HTMLEditor::HandleDeleteCollapsedSelectionAtTextNode()`
and `HTMLEditor::HandleDeleteCollapsedSelectionAtWhiteSpace()` is whether
it treats surrogate pair or not. And it's easy to handle it in `WSRunScanner`.
Therefore, they can handle all deletion in a text node.
Some new failures are caused by that we normalize preceding white-spaces of
deletion range but Blink does not to it. These failures will be fixed by
part 5.
Depends on D84946
Differential Revision: https://phabricator.services.mozilla.com/D84947
For the following patches, it's nicer that the API has optional behavior which
removes new empty text node or empty parent elements because directly removing
parent element can reduce the number of transactions and avoid storing text
with `DeleteTextTransaction()`.
Depends on D84944
Differential Revision: https://phabricator.services.mozilla.com/D84945
`InputEvent.getTargetRange()` should have actual delete range. It means that
the range should contain the invisible leading/trailing white-spaces which
need to be removed for keep them invisible, but should not contain the range
of normalizing white-space sequence because they are not target of edit action
indicated by `InputEvent.inputType`.
So, we can use new path which uses the new white-space normalizer for
computing the value of `InputEvent.getTargetRanges()` because difference of
white-space normalizer shouldn't affect the deleting ranges (although, some
existing path calls `DeleteNodeIfInvisibleAndEditableTextNode()` later so that
the new method, `ComputeRangeInTextNodesContainingInvisibleWhiteSpaces()`, does
not exactly same thing, but the result shouldn't become different in usual
cases). This new path can test with some WPTs under `editing/other`.
This patch creates new backspace/delete key handler when caret is at next
to a white-space as `HTMLEditor::HandleDeleteTextAroundCollapsedSelection()`
and creates helper methods of `WSRunScanner` to treat invisible leading and
trailing white-spaces.
Note that new failures are caused by the difference whether adjacent white-space
sequence at deletion is normalized or not in edge cases. They will be fixed
by the part.5.
Depends on D84943
Differential Revision: https://phabricator.services.mozilla.com/D84944
With the previous patch, 2 bugs become visible.
One is a `MOZ_ASSERT` in
`WhiteSpaceVisibilityKeeper::MakeSureToKeepVisibleStateOfWhiteSpacesAroundDeletingRange()`
is wrong. When `TextFragmentData::GetReplaceRangeDataAtEndOfDeletionRange()`
returns a replace range, it may set the start to inclusive next character of
end to delete. So, start of replace range may start after end to delete when
container of end to delete is not a text node and its next editable char point
is an ASCII white-space. Therefore, this patch rewrite it with
`MOZ_ASSERT_IF` and `EditorDOMPoint::IsInTextNode()`.
The other is `TextFragmentData::GetReplaceRangeDataAtEndOfDeletionRange()`
and `TextFragmentData::GetReplaceRangeDataAtStartOfDeletionRange()` may call
`TextFragmentData::GetFirstASCIIWhiteSpacePointCollapsedTo()` and
`TextFragmentData::GetEndOfCollapsibleASCIIWhiteSpaces()` with a point in
preformatted text node. Therefore, this patch adds the check into them.
Depends on D84323
Differential Revision: https://phabricator.services.mozilla.com/D84943
Traditionally, `WSRunScanner` (and formerly `WSRunObject`) treats all text
nodes in its range when scan start container node has preformatted style.
This means that when starting start from start or end of preformatted text node
or inline element, it treats adjacent white-spaces which is not preformatted
as preformatted.
This patch fixes this issue. Because of the fix of preceding patches,
`BoundaryData` stops scanning if it meets preformatted text node. So,
`BoundaryData` can store whether the scanners found a preformatted character.
Therefore, if one of `BoundaryData`s is marked as "preformatted", it means
the range contains a preformatted character. And keep referring the style
of scan start point if there is no visible text nodes around it.
Note that the above change causes new failures with `<listing>` element which
should be treated as `<pre>` element, but `HTMLEditUtils` treats it as
non-container inline element.
https://html.spec.whatwg.org/multipage/dom.html#elements-in-the-dom:listing
Therefore, this patch changes the definition of `<listing>` element and
`<xmp>` element which is also mentioned in the above link to container
element etc. `<listing>` element is treated exactly same as `<pre>`,
therefore, the new definition is same as `<pre>`, but `<xmp>` does not
allow any tags after opens it (i.e., even if its close tag). Therefore,
`<xmp>` definition is different from `<listing>` and `<pre>` elements'
definition.
Differential Revision: https://phabricator.services.mozilla.com/D84323
The check does not make sense because `TextFragmentData`'s range may contain
multiple text nodes which may have different style.
With the previous patch, the range won't cross characters in preformatted
text nodes. Therefore, their result won't contain preformatted characters
right now. So, we can just drop the unnecessary check from them.
Differential Revision: https://phabricator.services.mozilla.com/D84320
Simply, if a white-space is in a preformatted text node, we shouldn't modify
it for normalizing. This patch adds such checks to various points in
`WSRunObject.cpp`.
Differential Revision: https://phabricator.services.mozilla.com/D84319
Currently, `TextFragmentData` stores whether the scan start point is
preformatted or not. However, it does not make sense because the class
may contain other text nodes which may have different style in its range.
The main job of `TextFragmentData` is managing white-spaces as visible
sequence or invisible sequence. So, preformatted white-space should be
treated as visible character because of not collapsible with adjacent
formatted ASCII white-spaces.
First of all, this patch its initializer stop scanning white-spaces if
it meets non-empty preformatted text node.
Note that the new failures are caused by the difference whether which
white-space sequence should be normalized when modifying text at text node
or inline element boundary. This difference should be fixed in another
bug because our new normalizer does not handle this same as Blink for now.
Differential Revision: https://phabricator.services.mozilla.com/D84317
Having two classes in the inheritance chain inherit from SupportsWeakPtr
now won't compile, but you can use WeakPtr<Derived> when any base class
inherits from SupportsWeakPtr.
Differential Revision: https://phabricator.services.mozilla.com/D83674
They return copy of cached `EditorDOMRange` instance. So, they can return
const-reference rather than copy of stored data.
Differential Revision: https://phabricator.services.mozilla.com/D83919
With this patch, the creation cost of `TextFragmentData` is minimized, but
`MakeSureToKeepVisibleStateOfWhiteSpacesAroundDeletingRange()` can work
with the latest DOM tree information at touching start of the deleting range
even after touching end of the range.
Depends on D83228
Differential Revision: https://phabricator.services.mozilla.com/D83229
This is what the original method should've done. Mutation event listeners may
change the DOM tree at the first modification, but it hasn't checked whether
the range to delete or replace around start of deleting range is still valid
or not.
Depends on D83226
Differential Revision: https://phabricator.services.mozilla.com/D83228
For representing delete or replace data which have the range and replace string,
this patch creates `ReplaceRangeDataBase` class which is stack-only, has
`EditorDOMRange` or `EditorDOMRangeInTexts` as the range and has replace string
which can be empty string.
Then, `MakeSureToKeepVisibleStateOfWhiteSpacesAroundDeletingRange()` should
modify the DOM tree by itself. With the following patch, the behavior is
corrected for multiple DOM tree changes.
Depends on D83224
Differential Revision: https://phabricator.services.mozilla.com/D83226
This patch is preparation for easier to review.
This patch duplicates the body of
`MakeSureToKeepVisibleStateOfWhiteSpacesAroundDeletingRange()` and calls the
copies from new `MakeSureToKeepVisibleStateOfWhiteSpacesAroundDeletingRange()`.
The one is for handling start of deleting range, and the other is for handling
end of deleting range.
As you see, this patch and the following patches do not work because of the
duplicated code. Perhaps, part 2 works if there is no mutation event listeners.
Part 9 works even if there is mutation event listeners which touch the DOM
tree in the worst case (when it touches both end of delete range and start of
delete range, and around the start of deleting range is modified by JS).
Finally, part 9 takes back the original performance.
Depends on D82715
Differential Revision: https://phabricator.services.mozilla.com/D83217
Now, these classes are used only by `TextFragmentData` and they can be not
exposed. Therefore, we should hide them with making them private nested
classes of `TextFragmentData`.
Differential Revision: https://phabricator.services.mozilla.com/D82714
Although the new name is long, but I have no better idea. The class's purpose
is to keep white-space visibility around modifying DOM position. Therefore,
I use "keeper" for the name.
Differential Revision: https://phabricator.services.mozilla.com/D82713
Similar to the previous patch, `WSRunObject::NormalizeWhiteSpacesAround()` is
a wrapper to create `WSRunObject` instance for calling
`NormalizeWhiteSpacesAtEndOf()`, but it does not need to be `WSRunObject`'s
instance. Therefore, we can merge them.
Note that this renames the merged method to `NormalizeVisibleWhiteSpacesAt`.
Differential Revision: https://phabricator.services.mozilla.com/D82711
Now, it does not need to be a `WSRunObject` instance so that its wrapper to
create `WSRunObject` instance is not necessary.
Differential Revision: https://phabricator.services.mozilla.com/D82710
It's now can work with static helper methods and a `TextFragmentData` instance.
Therefore, this patch makes it a static method.
Note that it's always called with `nsIEditor::eNone` so that we can get rid of
the argument.
Differential Revision: https://phabricator.services.mozilla.com/D82705
Now, `mScanEndPoint` is not used. This patch removes it and clean up the
constructors of `WSRunScanner` and `WSRunObject`.
Differential Revision: https://phabricator.services.mozilla.com/D82703
It's simpler to make `WSRunScanner::InsertText()` take insertion point.
Then, it can do its jobs with `TextFragmentData` instance(s).
Differential Revision: https://phabricator.services.mozilla.com/D82702
`CreateVisibleWhiteSpacesData()` is now called multiple times and maybe called
after the DOM tree is modified even though it's a bug. Therefore, we should
make it store first result and return its reference instead.
Differential Revision: https://phabricator.services.mozilla.com/D82701
Same as the previous patch, it can be split to computation part and
modifying the DOM tree part. Then, the former can be in `TextFragmentData`
and the latter can be done by the caller which is only
`WSRunObject::InsertText()`.
Depends on D82699
Differential Revision: https://phabricator.services.mozilla.com/D82700
It's called by `WSRunObject::InsertText()` and `WSRunObject::InsertBreak()` and
it has 2 parts. One is, considering whether previous char is an NBSP or not
and whether it should be replaced with an ASCII white-space if it's followed
by visible character. The other is, doing the replacement. The latter code
is enough simple. Therefore, we can copy them into the callers. Then,
we can move the check logic into `TextFragmentData`.
Depends on D82698
Differential Revision: https://phabricator.services.mozilla.com/D82699
They work with a `TextFragmentData` instance, and the following patches
require to run it without `WSRunScanner`/`WSRunObject` instances.
Therefore, this patch moves them into `TextFragmentData`.
Depends on D82695
Differential Revision: https://phabricator.services.mozilla.com/D82697
This patch makes `WSRunScanner` have `TextFragmentData const mTextFragmentData`
instead of 2 `BoundaryData`s, a `NoBreakingSpaceData` and a `bool` storing
whether it's preformatted or not.
Depends on D82694
Differential Revision: https://phabricator.services.mozilla.com/D82695
For making `WSRunScanner::BoundaryData` independent from `WSRunScanner`,
its initializer should be in the class itself as static factory methods.
Depends on D82295
Differential Revision: https://phabricator.services.mozilla.com/D82693
Now, we can call `WSFragment` `VisibleWhiteSpacesData`. Then, the methods of
`WSRunObject` which take `WSFragment` as their arguments become clearer what
they mean.
Differential Revision: https://phabricator.services.mozilla.com/D82295
Now, `WSFragment` instances are created only by
`TextFragmentData::CreateWSFragmentForVisibleWhiteSpaces()`. Therefore, it's
always visible.
Additionally, this patch hides `WSFragment` constructors from the others too.
Differential Revision: https://phabricator.services.mozilla.com/D82294
Now, we know that it returns `WSFragment` for visible white-spaces and may
lie position in the line. Therefore, we should rename it as just representing
visible white-spaces.
Differential Revision: https://phabricator.services.mozilla.com/D82292
Now, nobody uses `WSRunScanner::FindNearestRun()` so that we can remove it.
Then, there is no users of `WSRunScanner:mFragments`. Therefore, we can remove
the member, accessors and initializers.
Differential Revision: https://phabricator.services.mozilla.com/D82291
Unfortunately, remaining code which use `beforeRun` and `afterRun` of
`WSRunObject::PreareToDeleteRangePriv()` is completely broken. It tries to
do what the new utility methods say, but as you see in the method, the
checking code does not make sense. For now, we should keep this broken
check even with the expensive DOM point comparisons. I hope that this does
not harm any benchmark score.
Note that I tested this with all automated tests with comparing the result
of these methods with `MOZ_ASSERT()` like this:
https://hg.mozilla.org/try/rev/29cb7840e404473a41d2d1fbdd229f762ccac5d3
So, I think that this is enough safe because the most edge cases are tested
by the first patch in this bug and `editing/run/(forwarddelete|delete).html`
of WPT.
Differential Revision: https://phabricator.services.mozilla.com/D82290
Similar to the previous changes, `WSRunObject::PrepareToDeleteRangePriv()`
can use `TextFragmentData::CreateWSFragmentForVisibleAndMiddleOfLine()` and
result of comparing with deleting range.
Differential Revision: https://phabricator.services.mozilla.com/D82288
For handling composition, `WSRunObject::InsertText()` rescan white-spaces at
end of replacing range. However, in most cases, `mScanStartPoint` and
`mScanEndPoint` are same. Therefore, we can save the runtime cost of the
rescan easier than the old design.
Differential Revision: https://phabricator.services.mozilla.com/D82287
I realized that `WSFragment::MarkAsVisible()` is called only by
`TextFragmentData::CreateWSFragmentForVisibleAndMiddleOfLine()`. Therefore,
`beforeRun && !beforeRun->IsStartOfHardLine() && beforeRun->IsVisible()` can be
handled with `pointPositionWithVisibleWhiteSpacesAtStart` and
`afterRun && !afterRun->IsEndOfHardLine() && afterRun->IsVisible()` can be
handled with `pointPositionWithVisibleWhiteSpacesAtEnd`. So, we can make
`WSRunObject::InsertText()` stop using `FindNearestFragment()` to compute
inserting string.
Differential Revision: https://phabricator.services.mozilla.com/D82286
Similar to `WSRunObject::InsertBreak()`, `WSRunObject::InsertText()` can use
the new APIs for `InsertBreak()` even though the name is odd for `InsertText()`,
but the mismatching is caused by odd logic of it. If replacing range starts
from and/or ends by middle of invisible white-space sequence, all of the
invisible white-spaces should be removed to prevent they become visible.
However, we shouldn't change any behavior in this bug.
Differential Revision: https://phabricator.services.mozilla.com/D82285