Handle image lazy loading on Wikipedia

Handle image lazy loading on Wikipedia

Hopefully Wikipedia will use <img> instead of <span> as placeholder.

Score changes (as of @150267738):

https://x20web.corp.google.com/~wychen/domdistillerscore/wiki-lazy/top-mobile-mhtml.html
  3 image entries changed (all wikipedia pages):
   Precision remains 1.0
   Recall 0.077 → 0.538
   F1     0.143 → 0.700

Performance changes:
  distillable-desktop-mhtml is 3.5% slower.

BUG=647667
R=mdjones@chromium.org

Committed: 8de0cacfed0621f309015703dd78909a25fefd6e

[wychen, 2017-03-03 17:53:47]

Patchset #1 (id:1) has been deleted

[wychen, 2017-03-14 01:39:05]

wychen@chromium.org changed reviewers:
+ mdjones@chromium.org

[wychen, 2017-03-14 01:39:05]

PTAL.

Benchmark showed 3.5% speed regression on the desktop MHTML dataset. I'm
thinking about using querySelectAll to get all the <span class=...> instead.

[mdjones, 2017-03-14 15:37:52]

On 2017/03/14 01:39:05, wychen wrote:
> PTAL.
> 
> Benchmark showed 3.5% speed regression on the desktop MHTML dataset. I'm
> thinking about using querySelectAll to get all the <span class=...> instead.

lgtm; I'm less concerned about speed if this makes wikipedia usable. If you use
querySelectorAll this code would be in a different file right?

[wychen, 2017-03-14 21:52:10]

On 2017/03/14 15:37:52, mdjones wrote:
> On 2017/03/14 01:39:05, wychen wrote:
> > PTAL.
> > 
> > Benchmark showed 3.5% speed regression on the desktop MHTML dataset. I'm
> > thinking about using querySelectAll to get all the <span class=...> instead.
> 
> lgtm; I'm less concerned about speed if this makes wikipedia usable. If you
use
> querySelectorAll this code would be in a different file right?

I might replace getRelevantTagNames() with a HashSet<Element> containing
querySelectAll results for the particular embed.
This would also enable the possibility of extracting images using
background-image instead of <img>.

[wychen, 2017-03-15 23:58:59]

Description was changed from

==========
Handle image lazy loading on Wikipedia

There are potential performance concerns.
Hopefully Wikipedia will use <img> instead of <span> as placeholder.

BUG=647667
==========

to

==========
Handle image lazy loading on Wikipedia

Hopefully Wikipedia will use <img> instead of <span> as placeholder.

Score changes (as of @150259105):

https://x20web.corp.google.com/~wychen/domdistillerscore/wiki-lazy/top-mobile...
  3 image entries changed (all wikipedia pages):
   Precision remains 1.0
   Recall 0.077 → 0.538
   F1     0.143 → 0.700

Performance changes:
  distillable-desktop-mhtml is 3.5% slower.

BUG=647667
==========

[wychen, 2017-03-16 00:46:14]

Description was changed from

==========
Handle image lazy loading on Wikipedia

Hopefully Wikipedia will use <img> instead of <span> as placeholder.

Score changes (as of @150259105):

https://x20web.corp.google.com/~wychen/domdistillerscore/wiki-lazy/top-mobile...
  3 image entries changed (all wikipedia pages):
   Precision remains 1.0
   Recall 0.077 → 0.538
   F1     0.143 → 0.700

Performance changes:
  distillable-desktop-mhtml is 3.5% slower.

BUG=647667
==========

to

==========
Handle image lazy loading on Wikipedia

Hopefully Wikipedia will use <img> instead of <span> as placeholder.

Score changes (as of @150267738):

https://x20web.corp.google.com/~wychen/domdistillerscore/wiki-lazy/top-mobile...
  3 image entries changed (all wikipedia pages):
   Precision remains 1.0
   Recall 0.077 → 0.538
   F1     0.143 → 0.700

Performance changes:
  distillable-desktop-mhtml is 3.5% slower.

BUG=647667
==========

[wychen, 2017-04-11 00:52:48]

Description was changed from

==========
Handle image lazy loading on Wikipedia

Hopefully Wikipedia will use <img> instead of <span> as placeholder.

Score changes (as of @150267738):

https://x20web.corp.google.com/~wychen/domdistillerscore/wiki-lazy/top-mobile...
  3 image entries changed (all wikipedia pages):
   Precision remains 1.0
   Recall 0.077 → 0.538
   F1     0.143 → 0.700

Performance changes:
  distillable-desktop-mhtml is 3.5% slower.

BUG=647667
==========

to

==========
Handle image lazy loading on Wikipedia

Hopefully Wikipedia will use <img> instead of <span> as placeholder.

Score changes (as of @150267738):

https://x20web.corp.google.com/~wychen/domdistillerscore/wiki-lazy/top-mobile...
  3 image entries changed (all wikipedia pages):
   Precision remains 1.0
   Recall 0.077 → 0.538
   F1     0.143 → 0.700

Performance changes:
  distillable-desktop-mhtml is 3.5% slower.

BUG=647667
R=mdjones@chromium.org

Committed: 8de0cacfed0621f309015703dd78909a25fefd6e
==========

[wychen, 2017-04-11 00:52:49]

Committed patchset #2 (id:40001) manually as
8de0cacfed0621f309015703dd78909a25fefd6e (presubmit successful).