Process ordered list items as a single WebText

Process ordered list items as a single WebText

This preserves ordered lists in the output. Otherwise they are
converted implicitly to unordered lists, because they are produced
without a parent.

BUG=502524

[Martin Robinson, 2015-06-20 14:25:24]

mrobinson@igalia.com changed reviewers:
+ cjhopman@chromium.org, mdjones@chromium.org, wychen@chromium.org

[wychen, 2015-06-20 18:00:44]

Would this work on nested ul and ol?

[Martin Robinson, 2015-06-22 14:37:00]

On 2015/06/20 18:00:44, wychen wrote:
> Would this work on nested ul and ol?

Nice catch! I think it might not be handled properly. Additionally, it only
modifies the behavior of ol. Modifying how ul is represented in the WebDocument
changes test results and specifically how various bits of text are scored. I
don't have access to the test corpus so I tried to make the most conservative
change possible.

I'll update the patch to support nested lists though. If you feel like it would
be better to also add handle ul this way, I can post it as well and perhaps one
of the Google team can test it against the dataset.

[Marcelo Correa, 2015-07-02 17:53:22]

I talked to Martin Robinson and I'm taking over this bug.

Using a similar solution to Martin Robinson, here is my findings about the bug:
I was able to nest lists but it only works when the ol, li, ol.... are indented.

It fails when the code is not indented like:
<ol><li><ol><li>test</li></ol></li></ol>

This happens because the method buildTreeClone() called by WebText when
generating output, creates a minimum tree in the DOM that contains the text
node.
Example: "Foobar" is a child of <li>, so the minimum tree will be
<li>Foobar</li>.

So, when the code is indented Dom Distiller treats white spaces or tabs from
indentation as text nodes.
So to generate the minimum tree containing LI texts and the empty space/tab
("\n", "\t"), Dom Distiller will get the parent of these two elements, which
leads to the first OL.
Example: <li> element is sibling of empty space, both are children of <ol>, so
the minimum tree will be <ol><li>foobar</li>""</ol>.

Still trying to figure out how to solve this issue without adding empty space as
child of ol element.
Any sugestions?

[Martin Robinson, 2015-07-09 16:18:12]

On 2015/07/02 17:53:22, Marcelo Correa wrote:
> I talked to Martin Robinson and I'm taking over this bug.


Marcelo is taking over this issue here:
https://codereview.chromium.org/1230583006