Optimizations to Spellchecking Chunk Generation

Optimizations to Spellchecking Chunk Generation

1. The existing implementation frequently does translations between
offsets and positions, and each translation needs to scan from the
beginning of the text paragraph. The patch avoids such translations
completely.

2. Only the first chunk needs to have both of its start and end
positions expanded to sentence boundaries. For each subsequent chunk, we
only need to expand its end position to sentence boundary. Hence, we can
reduce the number of calls of startOfSentence to 1 regardless of the
number of chunks generated.

BUG=109587
Committed: https://crrev.com/b0f70be05379a033d47c4eb7d5a389b546338044
Cr-Commit-Position: refs/heads/master@{#368035}

[Xiaocheng, 2016-01-06 13:19:09]

Description was changed from

==========
Optimizations to Spellchecking Chunk Generation

1. The existing implementation frequently does translations between
offsets and positions, and each translation needs to scan from the
beginning of the text paragraph. The patch avoids such translations
completely.

2. Only the first chunk needs to have both of its start and end
positions expanded to sentence boundaries. For each subsequent chunk, we
only need to expand its end position to sentence boundary. Hence, we can
reduce the number of calls of startOfSentence to 1 regardless of the
number of chunks generated.

BUG=109587
==========

to

==========
Optimizations to Spellchecking Chunk Generation

1. The existing implementation frequently does translations between
offsets and positions, and each translation needs to scan from the
beginning of the text paragraph. The patch avoids such translations
completely.

2. Only the first chunk needs to have both of its start and end
positions expanded to sentence boundaries. For each subsequent chunk, we
only need to expand its end position to sentence boundary. Hence, we can
reduce the number of calls of startOfSentence to 1 regardless of the
number of chunks generated.

BUG=109587
==========

[Xiaocheng, 2016-01-06 13:19:09]

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

[Xiaocheng, 2016-01-06 13:19:26]

PTAL.

[yosin_UTC9, 2016-01-07 08:34:55]

lgtm

So smart!

>2. Only the first chunk needs to have both of its start and end
>positions expanded to sentence boundaries. For each subsequent chunk, we
>only need to expand its end position to sentence boundary. Hence, we can
>reduce the number of calls of startOfSentence to 1 regardless of the
>number of chunks generated.

Here is my understanding:

sentence_1 sentence_2 sentence_3
012345678901234577890123456789012

When chunk size is 15,
chunk[0] = 0, 10
chunk[1] = 11, 21
chunk[2] = 22, 32

[yosin_UTC9, 2016-01-07 08:34:58]

The CQ bit was checked by yosin@chromium.org

[commit-bot: I haz the power, 2016-01-07 08:35:33]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1563783003/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1563783003/1

[commit-bot: I haz the power, 2016-01-07 08:39:55]

Description was changed from

==========
Optimizations to Spellchecking Chunk Generation

1. The existing implementation frequently does translations between
offsets and positions, and each translation needs to scan from the
beginning of the text paragraph. The patch avoids such translations
completely.

2. Only the first chunk needs to have both of its start and end
positions expanded to sentence boundaries. For each subsequent chunk, we
only need to expand its end position to sentence boundary. Hence, we can
reduce the number of calls of startOfSentence to 1 regardless of the
number of chunks generated.

BUG=109587
==========

to

==========
Optimizations to Spellchecking Chunk Generation

1. The existing implementation frequently does translations between
offsets and positions, and each translation needs to scan from the
beginning of the text paragraph. The patch avoids such translations
completely.

2. Only the first chunk needs to have both of its start and end
positions expanded to sentence boundaries. For each subsequent chunk, we
only need to expand its end position to sentence boundary. Hence, we can
reduce the number of calls of startOfSentence to 1 regardless of the
number of chunks generated.

BUG=109587
==========

[commit-bot: I haz the power, 2016-01-07 08:39:56]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2016-01-07 08:41:06]

Description was changed from

==========
Optimizations to Spellchecking Chunk Generation

1. The existing implementation frequently does translations between
offsets and positions, and each translation needs to scan from the
beginning of the text paragraph. The patch avoids such translations
completely.

2. Only the first chunk needs to have both of its start and end
positions expanded to sentence boundaries. For each subsequent chunk, we
only need to expand its end position to sentence boundary. Hence, we can
reduce the number of calls of startOfSentence to 1 regardless of the
number of chunks generated.

BUG=109587
==========

to

==========
Optimizations to Spellchecking Chunk Generation

1. The existing implementation frequently does translations between
offsets and positions, and each translation needs to scan from the
beginning of the text paragraph. The patch avoids such translations
completely.

2. Only the first chunk needs to have both of its start and end
positions expanded to sentence boundaries. For each subsequent chunk, we
only need to expand its end position to sentence boundary. Hence, we can
reduce the number of calls of startOfSentence to 1 regardless of the
number of chunks generated.

BUG=109587

Committed: https://crrev.com/b0f70be05379a033d47c4eb7d5a389b546338044
Cr-Commit-Position: refs/heads/master@{#368035}
==========

[commit-bot: I haz the power, 2016-01-07 08:41:06]

Patchset 1 (id:??) landed as
https://crrev.com/b0f70be05379a033d47c4eb7d5a389b546338044
Cr-Commit-Position: refs/heads/master@{#368035}

[Xiaocheng, 2016-01-07 08:43:54]

On 2016/01/07 at 08:34:55, yosin wrote:
> lgtm
> 
> So smart!
> 
> >2. Only the first chunk needs to have both of its start and end
> >positions expanded to sentence boundaries. For each subsequent chunk, we
> >only need to expand its end position to sentence boundary. Hence, we can
> >reduce the number of calls of startOfSentence to 1 regardless of the
> >number of chunks generated.
> 
> Here is my understanding:
> 
> sentence_1 sentence_2 sentence_3
> 012345678901234577890123456789012
> 
> When chunk size is 15,
> chunk[0] = 0, 10
> chunk[1] = 11, 21
> chunk[2] = 22, 32

Exactly! :)