Reduce canonicalization when dragging

Reduce canonicalization when dragging

This CL avoids VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms to 17ms).

Review-Url: https://codereview.chromium.org/2741173002
Cr-Commit-Position: refs/heads/master@{#457376}
Committed: https://chromium.googlesource.com/chromium/src/+/d51024e6850f785d2bb364af8d63f742126f7a91

[yoichio, 2017-03-10 08:27:23]

The CQ bit was checked by yoichio@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-03-10 08:27:44]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[yoichio, 2017-03-10 08:30:57]

Description was changed from

==========
dragg perf

BUG=
==========

to

==========
dragg perf

cc::ProxyMain::BeginMainFrame	
content::RenderWidget::WillBeginCompositorFrame

1. rootEditableElement without VisibleSelection
it is ok because canonicalization don't cross editing boundary

2. Avoid FlatTree creation if DomTree selection is none


BUG=
==========

[commit-bot: I haz the power, 2017-03-10 09:53:17]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2017-03-10 09:53:18]

Dry run: Try jobs failed on following builders:
  win_chromium_rel_ng on master.tryserver.chromium.win (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_...)

[yoichio, 2017-03-15 02:26:46]

The CQ bit was checked by yoichio@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-03-15 02:27:12]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[yoichio, 2017-03-15 02:30:56]

Description was changed from

==========
dragg perf

cc::ProxyMain::BeginMainFrame	
content::RenderWidget::WillBeginCompositorFrame

1. rootEditableElement without VisibleSelection
it is ok because canonicalization don't cross editing boundary

2. Avoid FlatTree creation if DomTree selection is none


BUG=
==========

to

==========
dragg perf

cc::ProxyMain::BeginMainFrame	
content::RenderWidget::WillBeginCompositorFrame

1. rootEditableElement without VisibleSelection
it is ok because canonicalization don't cross editing boundary

2. Avoid FlatTree creation if DomTree selection is none


BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 2.5 times(from 80ms
to 30ms).
==========

[yoichio, 2017-03-15 03:54:30]

Description was changed from

==========
dragg perf

cc::ProxyMain::BeginMainFrame	
content::RenderWidget::WillBeginCompositorFrame

1. rootEditableElement without VisibleSelection
it is ok because canonicalization don't cross editing boundary

2. Avoid FlatTree creation if DomTree selection is none


BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 2.5 times(from 80ms
to 30ms).
==========

to

==========
Reduce canonicalization when dragging

This CL does
1. Use rootEditableElementOf(Position) instead of 
  computeVisiblePosition.rootEditableElement().
(InputMethodController.cpp, FrameCaret.cpp)
This conversion is O.K. because canonicalicalization doesn't go over
 EditingBoundary in which rootEditableElement traverses to find root element.

In addition, move calling rootEditableElement below early returns which return
 also WebTextInputTypeNone to reduce rootEditableElement calls.

2. Avoid VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

Unfortunately, 
 WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
 measure this performance improvement because it goes through another
 code path to original issue's.

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 2.5 times(from 80ms
to 30ms).
==========

[commit-bot: I haz the power, 2017-03-15 04:14:55]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2017-03-15 04:14:56]

Dry run: Try jobs failed on following builders:
  win_chromium_rel_ng on master.tryserver.chromium.win (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_...)

[yoichio, 2017-03-15 08:34:43]

Description was changed from

==========
Reduce canonicalization when dragging

This CL does
1. Use rootEditableElementOf(Position) instead of 
  computeVisiblePosition.rootEditableElement().
(InputMethodController.cpp, FrameCaret.cpp)
This conversion is O.K. because canonicalicalization doesn't go over
 EditingBoundary in which rootEditableElement traverses to find root element.

In addition, move calling rootEditableElement below early returns which return
 also WebTextInputTypeNone to reduce rootEditableElement calls.

2. Avoid VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

Unfortunately, 
 WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
 measure this performance improvement because it goes through another
 code path to original issue's.

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 2.5 times(from 80ms
to 30ms).
==========

to

==========
Reduce canonicalization when dragging

This CL does
1. Use rootEditableElementOf(Position) instead of 
  computeVisiblePosition.rootEditableElement().
(InputMethodController.cpp, FrameCaret.cpp)
This conversion is O.K. because canonicalicalization doesn't go over
 EditingBoundary in which rootEditableElement traverses to find root element.

In addition, move calling rootEditableElement below early returns which return
 also WebTextInputTypeNone to reduce rootEditableElement calls.

2. Avoid VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

Unfortunately, 
 WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
 measure this performance improvement because it goes through another
 code path to original issue's.

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms
to 17ms).
==========

[yoichio, 2017-03-15 08:42:40]

The CQ bit was checked by yoichio@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-03-15 08:42:51]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2017-03-15 11:02:11]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2017-03-15 11:02:12]

Dry run: This issue passed the CQ dry run.

[yoichio, 2017-03-16 01:46:50]

Description was changed from

==========
Reduce canonicalization when dragging

This CL does
1. Use rootEditableElementOf(Position) instead of 
  computeVisiblePosition.rootEditableElement().
(InputMethodController.cpp, FrameCaret.cpp)
This conversion is O.K. because canonicalicalization doesn't go over
 EditingBoundary in which rootEditableElement traverses to find root element.

In addition, move calling rootEditableElement below early returns which return
 also WebTextInputTypeNone to reduce rootEditableElement calls.

2. Avoid VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

Unfortunately, 
 WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
 measure this performance improvement because it goes through another
 code path to original issue's.

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms
to 17ms).
==========

to

==========
Reduce canonicalization when dragging

This CL does
1. Use rootEditableElementOf(Position) instead of 
  computeVisiblePosition.rootEditableElement().
(InputMethodController.cpp, FrameCaret.cpp)
This conversion is O.K. because canonicalicalization doesn't go over
 EditingBoundary in which rootEditableElement traverses to find root element.

In addition, move calling rootEditableElement below the early returns 
which return also WebTextInputTypeNone to reduce
 rootEditableElement calls.

2. Avoid VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

Unfortunately, 
 WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
 measure this performance improvement because it goes through another
 code path to original issue's.

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms
to 17ms).
==========

[yoichio, 2017-03-16 01:47:32]

yoichio@chromium.org changed reviewers:
+ yosin@chromium.org

[yosin_UTC9, 2017-03-16 02:30:05]

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/FrameCaret.cpp (right):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/FrameCaret.cpp:208:
rootEditableElementOf(m_selectionEditor->selectionInDOMTree().base());
Could you move change in another CL?
This change is independent from others.

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/InputMethodController.cpp (left):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/InputMethodController.cpp:1035:
.computeVisibleSelectionInDOMTreeDeprecated()
Let's do mechanical change for

computeVisibleSelectionInDOMDeprecaerd().rootEditableElement() to
rootEditableElement(...selectionInDOMTree().base())

in another patch.

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/InputMethodController.cpp (right):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/InputMethodController.cpp:1187: // It's
important to preserve the equivalence of textInputInfo().type and
Please have another patch to moving this block.

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/SelectionController.cpp (right):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/SelectionController.cpp:390: if
(basePosition.isNull())
Could you move this change in another patch?
with adding |DCHECK(basePosition.isNotNull())| in |applySelectAll()|.

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/SelectionEditor.cpp (right):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/SelectionEditor.cpp:374: if
(m_cachedVisibleSelectionInDOMTree.isNone()) {
Good catch!

Please move this shortcut in another patch.

We would like to split SelectionEditor::udpateCahceVisibleSelectionIfNeeded()
into to one for DOM tree and another for Flat Tree.
We don't need to compute DOM tree version and Flat Tree version same time, e.g.
execCommand() only needs DOM tree version and painting needs only Flat Tree
version.

This is my mistake that having only one cache.

[yosin_UTC9, 2017-03-16 02:32:01]

BTW,

Unfortunately, 
 WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
 measure this performance improvement because it goes through another
 code path to original issue's.

If this description is true, this patch isn't related to issue 668122

[yoichio, 2017-03-16 03:59:08]

The CQ bit was checked by yoichio@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-03-16 03:59:52]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[yoichio, 2017-03-16 04:11:00]

> Unfortunately, 
>  WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
>  measure this performance improvement because it goes through another
>  code path to original issue's.
> 
> If this description is true, this patch isn't related to the issue
No, this fixes the issue.
My mistake is that mouse-move-with-hidden-elements.html doesn't
  measure the issue.

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/FrameCaret.cpp (right):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/FrameCaret.cpp:208:
rootEditableElementOf(m_selectionEditor->selectionInDOMTree().base());
On 2017/03/16 02:30:05, yosin_UTC9 wrote:
> Could you move change in another CL?
> This change is independent from others.

Acknowledged.

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/InputMethodController.cpp (left):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/InputMethodController.cpp:1035:
.computeVisibleSelectionInDOMTreeDeprecated()
On 2017/03/16 02:30:05, yosin_UTC9 wrote:
> Let's do mechanical change for
> 
> computeVisibleSelectionInDOMDeprecaerd().rootEditableElement() to
> rootEditableElement(...selectionInDOMTree().base())
> 
> in another patch.

Acknowledged.

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/InputMethodController.cpp (right):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/InputMethodController.cpp:1187: // It's
important to preserve the equivalence of textInputInfo().type and
On 2017/03/16 02:30:05, yosin_UTC9 wrote:
> Please have another patch to moving this block.

Acknowledged.

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/SelectionController.cpp (right):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/SelectionController.cpp:390: if
(basePosition.isNull())
On 2017/03/16 02:30:05, yosin_UTC9 wrote:
> Could you move this change in another patch?
> with adding |DCHECK(basePosition.isNotNull())| in |applySelectAll()|.
This needs because of SelectionEditor change.
Somehow in L346-350, we get "valid" |targetPosition| but null 
 |basePosition|.
I'm suspicious about this code that we get VSinDOM first and 
 covert it to VSinFlat. We should get VSinFlat directly?
Anyway this should be done in another patch.

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/SelectionEditor.cpp (right):

https://codereview.chromium.org/2741173002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/SelectionEditor.cpp:374: if
(m_cachedVisibleSelectionInDOMTree.isNone()) {
On 2017/03/16 02:30:05, yosin_UTC9 wrote:
> Good catch!
> 
> Please move this shortcut in another patch.
> 
This is core of this CL and needs for performance improvement.

> We would like to split SelectionEditor::udpateCahceVisibleSelectionIfNeeded()
> into to one for DOM tree and another for Flat Tree.
> We don't need to compute DOM tree version and Flat Tree version same time,
e.g.
> execCommand() only needs DOM tree version and painting needs only Flat Tree
> version.
> 
> This is my mistake that having only one cache.
We should do with fixing dragging selection code

[yoichio, 2017-03-16 04:36:58]

Description was changed from

==========
Reduce canonicalization when dragging

This CL does
1. Use rootEditableElementOf(Position) instead of 
  computeVisiblePosition.rootEditableElement().
(InputMethodController.cpp, FrameCaret.cpp)
This conversion is O.K. because canonicalicalization doesn't go over
 EditingBoundary in which rootEditableElement traverses to find root element.

In addition, move calling rootEditableElement below the early returns 
which return also WebTextInputTypeNone to reduce
 rootEditableElement calls.

2. Avoid VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

Unfortunately, 
 WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
 measure this performance improvement because it goes through another
 code path to original issue's.

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms
to 17ms).
==========

to

==========
Reduce canonicalization when dragging

This CL avoids VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

Unfortunately, 
 WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
 measure this performance improvement because it goes through another
 code path to original issue's.

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms
to 17ms).
==========

[commit-bot: I haz the power, 2017-03-16 05:30:16]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2017-03-16 05:30:16]

Dry run: This issue passed the CQ dry run.

[yoichio, 2017-03-16 05:42:20]

Any thoughts?

[yoichio, 2017-03-16 06:37:15]

Description was changed from

==========
Reduce canonicalization when dragging

This CL avoids VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

Unfortunately, 
 WebKit/PerformanceTests/Editing/mouse-move-with-hidden-elements.html doesn't
 measure this performance improvement because it goes through another
 code path to original issue's.

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms
to 17ms).
==========

to

==========
Reduce canonicalization when dragging

This CL avoids VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms
to 17ms).
==========

[yosin_UTC9, 2017-03-16 06:39:18]

lgtm w/ nit

Please move "Unfortunately, ..." from description to crbug.com, it makes me
confusing.

https://codereview.chromium.org/2741173002/diff/40001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/SelectionEditor.cpp (right):

https://codereview.chromium.org/2741173002/diff/40001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/SelectionEditor.cpp:387: m_styleVersion =
document().styleVersion();
Could you move L387 and L388 to L373 and make L374 to early return-style to
reduce indentation and make patch size smaller.

 m_styleVersion = document().styleVersion();
 m_cacheIsDirty = false;
 m_cachedVisibleSelectionInDOMTree = createVisibleSelection(m_selection);
 if (m_cachedVisibleSelectionInDOMTree.isNone()) {
  m_cachedVisibleSelectionInFlatTree = VisibleSelectionInFlatTree();
  return;
 }
 ...

[yosin_UTC9, 2017-03-16 06:40:13]

Also, please update summary to describe this change, e.g.

Make SE::updateXXX not to compute flat tree VS if DOM VS is none.

[yoichio, 2017-03-16 06:56:40]

Thanks

https://codereview.chromium.org/2741173002/diff/40001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/editing/SelectionEditor.cpp (right):

https://codereview.chromium.org/2741173002/diff/40001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/editing/SelectionEditor.cpp:387: m_styleVersion =
document().styleVersion();
On 2017/03/16 06:39:18, yosin_UTC9 wrote:
> Could you move L387 and L388 to L373 and make L374 to early return-style to
> reduce indentation and make patch size smaller.
> 
>  m_styleVersion = document().styleVersion();
>  m_cacheIsDirty = false;
>  m_cachedVisibleSelectionInDOMTree = createVisibleSelection(m_selection);
>  if (m_cachedVisibleSelectionInDOMTree.isNone()) {
>   m_cachedVisibleSelectionInFlatTree = VisibleSelectionInFlatTree();
>   return;
>  }
>  ...

Done.

[yoichio, 2017-03-16 06:56:45]

The CQ bit was checked by yoichio@chromium.org

[yoichio, 2017-03-16 06:56:46]

The patchset sent to the CQ was uploaded after l-g-t-m from yosin@chromium.org
Link to the patchset: https://codereview.chromium.org/2741173002/#ps60001
(title: "update")

[commit-bot: I haz the power, 2017-03-16 06:56:55]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2017-03-16 08:54:18]

CQ is committing da patch.

Bot data: {"patchset_id": 60001, "attempt_start_ts": 1489647405444490,
"parent_rev": "20bd5b1c1e57d402e3e300b4175ec52a7597f64e", "commit_rev":
"d51024e6850f785d2bb364af8d63f742126f7a91"}

[commit-bot: I haz the power, 2017-03-16 08:55:03]

Description was changed from

==========
Reduce canonicalization when dragging

This CL avoids VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms
to 17ms).
==========

to

==========
Reduce canonicalization when dragging

This CL avoids VisibleSelectionInFlatTree creation if VSInDOMTree is none
 in SelectionEditor.cpp.
This reduces redundant canonicalization because
 VSInFlatTree availability == VSInDOMTree availability

BUG=668122
TEST=
 1. Open the page on comment#21 described at the bug.
 2. Open also chrome://tracing/ and record Input Latency.
 3. Confirm Messaseloop::RunTask blocking time gets faster 3.7 times(from 64ms
to 17ms).

Review-Url: https://codereview.chromium.org/2741173002
Cr-Commit-Position: refs/heads/master@{#457376}
Committed:
https://chromium.googlesource.com/chromium/src/+/d51024e6850f785d2bb364af8d63...
==========

[commit-bot: I haz the power, 2017-03-16 08:55:04]

Committed patchset #4 (id:60001) as
https://chromium.googlesource.com/chromium/src/+/d51024e6850f785d2bb364af8d63...