In complex multi-template scenarios, preserve wikitext newline constraints during html->wt of edited content
Closed, ResolvedPublic0 Estimated Story Points
Actions

Assigned To

Authored By

	• Whatamidoing-WMF
	Aug 3 2015, 11:25 PM

Description

https://en.wikipedia.org/w/index.php?diff=prev&oldid=673179474 is the diff in question. The required newline before the "{|" has not been respected during serialization.

This seems to be because a {{DISPLAYTITLE:..}} meta-tag that was part of {{Infobox ship begin|..}} has been hoisted out of the table and has captured templating information for the entire infobox. During serialization, Parsoid sees a meta-tag that is part of a transclusion. It does not see a table and does not know to introduce a newline break there. However, since the meta-tag was not part of the source wikitext, but instead came from a template, Parsoid should have ignored it while computing newline constraints.

Details

	Subject	Repo	Branch	Lines +/-
	Record first wikitext node in multi-template-content-block scenario	mediawiki/services/parsoid	master	+141 -18

Customize query in gerrit

Related Objects

Mentioned Here: T86271: Existing categories that are preceded by new/modified content should be forced onto a new line
T48811: Fix parsing of infobox templates used in table attribute position
T58234: VisualEditor: Hatnote cleanup problem

Event Timeline

• Whatamidoing-WMF created this task.Aug 3 2015, 11:25 PM

• Whatamidoing-WMF raised the priority of this task from to Medium.

• Whatamidoing-WMF updated the task description. (Show Details)

• Whatamidoing-WMF added projects: VisualEditor, Parsoid.

• Whatamidoing-WMF subscribed.

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptAug 3 2015, 11:25 PM

Jdforrester-WMF moved this task from To Triage to TR1: Releases on the VisualEditor board.Aug 4 2015, 7:03 PM

Jdforrester-WMF moved this task from TR1: Releases to TR0: Interrupt on the VisualEditor board.Aug 26 2015, 1:36 AM

Jdforrester-WMF set Security to None.Sep 15 2015, 2:47 AM

Jdforrester-WMF edited a custom field.

ssastry moved this task from Needs Triage to In Progress on the Parsoid board.Oct 6 2015, 3:09 AM

This is not directly related to tables and seems very specific to cases where there are intervening HTML tags before the table.

In this case, there is a {{DISPLAYTITLE:..}} meta tag (which was in Infobox ship begin) that got fostered out of the table. So, when content before the table is deleted, the meta-tag is interfering with the constraint that forces newlines before tables. Because of the complex templating scenario, parsoid never sees a table tag during serialization. It sees a meta-tag that is template-generated. And, since transclusions don't have any newline constraints, there is no newline break introduced here.

So, the good news is that in the general scenario, we won't break wikitext like this. The bad news is that this is somewhat of an edge case and seems a little sticky ... have to figure out what useful information we can use here that doesn't involve some weird hack.

ssastry renamed this task from Don't remove the line feed/carriage return before a table to In complex multi-template scenarios, preserve wikitext newline constraints during html->wt of edited content.Oct 9 2015, 9:10 PM

ssastry updated the task description. (Show Details)

ssastry claimed this task.Oct 16 2015, 2:44 PM

ssastry added a comment.Nov 20 2015, 5:19 AM

This comment was removed by ssastry.

This does seem to be fostering-related, except it is not marked as fostered because the fostered content comes from a transclusion.

I have a solution for this, but this also requires the patch for T86271: Existing categories that are preceded by new/modified content should be forced onto a new line to be reviewed and merged.

Change 254472 had a related patch set uploaded (by Subramanya Sastry):
WIP: Record first tpl node in multi-template-content-block scenario

https://gerrit.wikimedia.org/r/254472

gerritbot added a project: Patch-For-Review.Nov 20 2015, 8:24 PM

Change 254472 merged by jenkins-bot:
Record first wikitext node in multi-template-content-block scenario