Prevent to measure the whole word when break-word

Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it lets overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as the sum of
such effects is within 2em.

BUG=603398
Committed: https://crrev.com/41fe145a8629f1f7002b41a8e12687b97d5e635f
Cr-Commit-Position: refs/heads/master@{#387974}

[kojii, 2016-04-18 06:29:00]

Description was changed from

==========
break-word

BUG=603398
==========

to

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it let overflow by 50px, and rewind the overflow.

This prevents the line breaker to measure the whole word, while
preventing to measure ligatures/kernings incorrectly as long as it
occurs within the 50px buffer.

BUG=603398
==========

[kojii, 2016-04-18 06:30:20]

kojii@chromium.org changed reviewers:
+ drott@chromium.org, eae@chromium.org

[kojii, 2016-04-18 06:30:21]

Not fully tested yet, but here's another proposal to solve the perf issue in
break-word.

Does this look better idea, assuming I can make it to pass all the tests?

[kojii, 2016-04-18 06:34:32]

Description was changed from

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it let overflow by 50px, and rewind the overflow.

This prevents the line breaker to measure the whole word, while
preventing to measure ligatures/kernings incorrectly as long as it
occurs within the 50px buffer.

BUG=603398
==========

to

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it let overflow by 50px, and rewind the overflow.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as it occurs
within the 50px buffer from the line end.

BUG=603398
==========

[kojii, 2016-04-18 06:38:24]

To help you to review the diff; the "if" block is simply a revert of
crrev.com/385693 except that it has additional "+ 50".

[kojii, 2016-04-18 06:50:27]

For the test case in crbug.com/603398 (9233 bytes of hex string) on a Linux
release+buildtype=Official
This CL: 124ms
https://codereview.chromium.org/1894863002 : 315ms

[kojii, 2016-04-18 07:24:28]

Description was changed from

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it let overflow by 50px, and rewind the overflow.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as it occurs
within the 50px buffer from the line end.

BUG=603398
==========

to

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it let overflow by 2em, and rewind the overflow.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as it occurs
within the 2em from the line end.

BUG=603398
==========

[kojii, 2016-04-18 07:24:59]

Description was changed from

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it let overflow by 2em, and rewind the overflow.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as it occurs
within the 2em from the line end.

BUG=603398
==========

to

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it let overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as it occurs
within the 2em from the line end.

BUG=603398
==========

[drott, 2016-04-18 07:27:50]

I took a look at this, Koji, but I am afraid I can't judge this and would defer
the review to eae@.

[kojii, 2016-04-18 07:40:00]

Description was changed from

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it let overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as it occurs
within the 2em from the line end.

BUG=603398
==========

to

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it lets overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as it occurs
within the 2em from the line end.

BUG=603398
==========

[kojii, 2016-04-18 13:08:21]

Description was changed from

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it lets overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as it occurs
within the 2em from the line end.

BUG=603398
==========

to

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it lets overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as the sum of
the effects is within 2em in a line.

BUG=603398
==========

[kojii, 2016-04-18 13:08:52]

Description was changed from

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it lets overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as the sum of
the effects is within 2em in a line.

BUG=603398
==========

to

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it lets overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as the sum of
such effects is within 2em.

BUG=603398
==========

[eae, 2016-04-18 17:55:12]

LGTM

[kojii, 2016-04-18 18:54:55]

The CQ bit was checked by kojii@chromium.org

[commit-bot: I haz the power, 2016-04-18 18:55:42]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1900513002/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1900513002/40001

[commit-bot: I haz the power, 2016-04-18 19:02:40]

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

[commit-bot: I haz the power, 2016-04-18 19:02:41]

Failed to apply the patch.

[commit-bot: I haz the power, 2016-04-18 19:02:49]

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

[commit-bot: I haz the power, 2016-04-18 19:03:41]

Description was changed from

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it lets overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as the sum of
such effects is within 2em.

BUG=603398
==========

to

==========
Prevent to measure the whole word when break-word

crrev.com/385693 changed the line breaker to measure word-by-word then
rewind if break-word/break-all. This causes a performance regression
when a word is extraordinary long, such as minified JS/CSS or
hex/base64 data.

This patch fixes to measure character-by-character, but unlike before
r385693, it lets overflow by 2em then rewind to the correct position.

This prevents the line breaker to measure the whole word, while
still measuring ligatures/kernings correctly as long as the sum of
such effects is within 2em.

BUG=603398

Committed: https://crrev.com/41fe145a8629f1f7002b41a8e12687b97d5e635f
Cr-Commit-Position: refs/heads/master@{#387974}
==========

[commit-bot: I haz the power, 2016-04-18 19:03:42]

Patchset 3 (id:??) landed as
https://crrev.com/41fe145a8629f1f7002b41a8e12687b97d5e635f
Cr-Commit-Position: refs/heads/master@{#387974}