Originated from: https://github.com/openzim/mwoffliner/issues/359
We're finding that some WikiTables are returned as source, rather than parsed.
An example is here:
http://en.wikipedia.org/api/rest_v1/page/mobile-sections/Molecular_geometry
The returned section with ID 7 contains an un-parsed wikitable (:{| class="wikitable"...) along with multiple properly parsed tables.
Could be that there is no whitespace before it?
Reducible to the following edge case:
[subbu@earth:~/work/wmf/parsoid] cat /tmp/wt
:{|
|foo
|}
{|
|bar
|}
:{|
|bar
|}
:x
:x
[subbu@earth:~/work/wmf/parsoid] php ../mediawiki/maintenance/parse.php < /tmp/wt
<dl><dd><table>
<tbody><tr>
<td>foo
</td></tr></tbody></table></dd></dl>
<table>
<tbody><tr>
<td>bar
</td></tr></tbody></table>
<dl><dd><table>
<tbody><tr>
<td>bar
</td></tr></tbody></table></dd></dl>
<dl><dd>x</dd></dl>
<pre> :x
</pre>
[subbu@earth:~/work/wmf/parsoid] parse.js --normalize < /tmp/wt
<pre> :{|
|foo
|}</pre>
<table>
<tbody>
<tr>
<td>bar</td>
</tr>
</tbody>
</table>
<dl>
<dd>
<table>
<tbody>
<tr>
<td>bar</td>
</tr>
</tbody>
</table>
</dd>
</dl>
<dl>
<dd>x</dd>
</dl>
<pre> :x</pre>So, an indented table parses just fine and a colon-prefixed table parses just fine, but, an indented colon-prefixed table is not parsed as a table in Parsoid. The PHP parser behavior is somewhat odd given that you cannot space-indent a colon normally. So, it breaks that behavior for tables? I am inclined to just fix the wikitext and treat this as unsupported in Parsoid. We could probably lint this wikitext in the corpus and have editors fix this edge case.