Add HTML parser UMA for rewinds and chunk queue statistics.

Add HTML parser UMA for rewinds and chunk queue statistics.

This patch adds uma for how many tokens we had to rewind the BG parser,
as well as the peak size of the pending chunk queue, in chunk length and
token length. We also log the time it takes to create a chunk/checkpoint,
which can involve lengthy string copies.

This is in preparation for an experiment where the parser's behavior with
the queue will be modified.

BUG=603274
Committed: https://crrev.com/cfb243c66139d26bccc8056f34a006d2588e86d3
Cr-Commit-Position: refs/heads/master@{#401819}

[Charlie Harrison, 2016-06-17 06:11:27]

csharrison@chromium.org changed reviewers:
+ kouhei@chromium.org

[Charlie Harrison, 2016-06-17 06:11:27]

Kouhei, can you give me some feedback on these statistics? I'm not sure if we
are measuring the right things here. In particular:

1. Should we be measuring aggregate metrics on every non fragment detach()? This
often measures stats that aren't obvious (cnn.com logs like 13 histogram
values). What are all the use cases of HTMLDocumentParser? Should we be
restricting them to just the main html?

2. I did not include a memory calculation, but rather just a peak chunk / token
count as a rough proxy. Do you suppose that's good enough?

3. I included how far back we had to rewind in terms of tokens. Should we
include any other rewind stat? I'm curious about the ratio of successful
document.writes to unsuccessful document.writes (if the ratio is poor, the
feature could be considered for removal). That may be a separate change though.

[Charlie Harrison, 2016-06-17 06:12:01]

Description was changed from

==========
Add HTML parser UMA for rewinds and chunk queue statistics.

This patch adds uma for how many tokens we had to rewind the BG parser,
as well as the peak size of the pending chunk queue, in chunk length and
token length. We also log the time it takes to create a chunk/checkpoint,
which can involve lengthy string copies.

This is in preparation for an experiment where the parser's behavior with
the queue will be modified.

BUG=603274
==========

to

==========
Add HTML parser UMA for rewinds and chunk queue statistics.

This patch adds uma for how many tokens we had to rewind the BG parser,
as well as the peak size of the pending chunk queue, in chunk length and
token length. We also log the time it takes to create a chunk/checkpoint,
which can involve lengthy string copies.

This is in preparation for an experiment where the parser's behavior with
the queue will be modified.

BUG=603274
==========

[kouhei (in TOK), 2016-06-17 08:42:27]

On 2016/06/17 06:11:27, csharrison wrote:
> Kouhei, can you give me some feedback on these statistics? I'm not sure if we
> are measuring the right things here. In particular:
> 
> 1. Should we be measuring aggregate metrics on every non fragment detach()?
This
> often measures stats that aren't obvious (http://cnn.com logs like 13
histogram
> values). What are all the use cases of HTMLDocumentParser? Should we be
> restricting them to just the main html?
I think we should measure this for subframes too, assuming 13 is from ads
iframes?

> 2. I did not include a memory calculation, but rather just a peak chunk /
token
> count as a rough proxy. Do you suppose that's good enough?
Yes. I think this is good enough.

> 3. I included how far back we had to rewind in terms of tokens. Should we
> include any other rewind stat? I'm curious about the ratio of successful
> document.writes to unsuccessful document.writes (if the ratio is poor, the
> feature could be considered for removal). That may be a separate change
though.
Your plan sgtm

[kouhei (in TOK), 2016-06-17 08:42:50]

lgtm

core/ lgtm

[Charlie Harrison, 2016-06-17 10:36:36]

csharrison@chromium.org changed reviewers:
+ holte@chromium.org

[Charlie Harrison, 2016-06-17 10:36:37]

Thanks!

holte@, can you look at histograms.xml?

[kouhei (in TOK), 2016-06-17 11:20:14]

kouhei@chromium.org changed reviewers:
+ caseq@chromium.org

[kouhei (in TOK), 2016-06-17 11:20:57]

https://codereview.chromium.org/2078723002/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/html/parser/HTMLDocumentParser.cpp (right):

https://codereview.chromium.org/2078723002/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/html/parser/HTMLDocumentParser.cpp:392:
TRACE_EVENT0("blink,loading",
"HTMLDocumentParser::discardSpeculationsAndResumeFrom");
caseq: Would you review this TRACE_EVENT?

[kouhei (in TOK), 2016-06-17 11:21:08]

kouhei@chromium.org changed required reviewers:
+ caseq@chromium.org

[Steven Holte, 2016-06-17 19:21:18]

lgtm

[Charlie Harrison, 2016-06-17 20:38:44]

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

[commit-bot: I haz the power, 2016-06-17 20:39:20]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2078723002/1

[commit-bot: I haz the power, 2016-06-17 21:18:24]

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

[commit-bot: I haz the power, 2016-06-17 21:18:25]

Dry run: This issue passed the CQ dry run.

[Charlie Harrison, 2016-06-22 14:51:42]

On 2016/06/17 21:18:25, commit-bot: I haz the power wrote:
> Dry run: This issue passed the CQ dry run.

caseq: friendly ping

[Charlie Harrison, 2016-06-24 03:55:47]

csharrison@chromium.org changed reviewers:
- caseq@chromium.org

[Charlie Harrison, 2016-06-24 03:55:47]

csharrison@chromium.org changed required reviewers:
- caseq@chromium.org

[Charlie Harrison, 2016-06-24 03:55:48]

Removed the trace event and removed caseq@ from review.

I'll upload a followup with this trace and more :)

[Charlie Harrison, 2016-06-24 03:55:49]

Removed the trace event and removed caseq@ from review.

I'll upload a followup with this trace and more :)

[Charlie Harrison, 2016-06-24 03:56:20]

The CQ bit was checked by csharrison@chromium.org

[Charlie Harrison, 2016-06-24 03:56:20]

The patchset sent to the CQ was uploaded after l-g-t-m from kouhei@chromium.org,
holte@chromium.org
Link to the patchset: https://codereview.chromium.org/2078723002/#ps20001
(title: "Remove trace event")

[commit-bot: I haz the power, 2016-06-24 03:56:33]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2078723002/20001

[commit-bot: I haz the power, 2016-06-24 05:31:10]

Description was changed from

==========
Add HTML parser UMA for rewinds and chunk queue statistics.

This patch adds uma for how many tokens we had to rewind the BG parser,
as well as the peak size of the pending chunk queue, in chunk length and
token length. We also log the time it takes to create a chunk/checkpoint,
which can involve lengthy string copies.

This is in preparation for an experiment where the parser's behavior with
the queue will be modified.

BUG=603274
==========

to

==========
Add HTML parser UMA for rewinds and chunk queue statistics.

This patch adds uma for how many tokens we had to rewind the BG parser,
as well as the peak size of the pending chunk queue, in chunk length and
token length. We also log the time it takes to create a chunk/checkpoint,
which can involve lengthy string copies.

This is in preparation for an experiment where the parser's behavior with
the queue will be modified.

BUG=603274
==========

[commit-bot: I haz the power, 2016-06-24 05:31:11]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-06-24 05:33:23]

Description was changed from

==========
Add HTML parser UMA for rewinds and chunk queue statistics.

This patch adds uma for how many tokens we had to rewind the BG parser,
as well as the peak size of the pending chunk queue, in chunk length and
token length. We also log the time it takes to create a chunk/checkpoint,
which can involve lengthy string copies.

This is in preparation for an experiment where the parser's behavior with
the queue will be modified.

BUG=603274
==========

to

==========
Add HTML parser UMA for rewinds and chunk queue statistics.

This patch adds uma for how many tokens we had to rewind the BG parser,
as well as the peak size of the pending chunk queue, in chunk length and
token length. We also log the time it takes to create a chunk/checkpoint,
which can involve lengthy string copies.

This is in preparation for an experiment where the parser's behavior with
the queue will be modified.

BUG=603274

Committed: https://crrev.com/cfb243c66139d26bccc8056f34a006d2588e86d3
Cr-Commit-Position: refs/heads/master@{#401819}
==========

[commit-bot: I haz the power, 2016-06-24 05:33:24]

Patchset 2 (id:??) landed as
https://crrev.com/cfb243c66139d26bccc8056f34a006d2588e86d3
Cr-Commit-Position: refs/heads/master@{#401819}