Provide DOM versions of the OutputTransform pipeline transformations
Open, HighPublic
Actions

Assigned To

Authored By

	ihurbain
	May 13 2025, 8:42 AM

Description

A significant performance issue on Parsoid is the back and forth between text and DOM formats on the output pipeline (@ssastry has a trace that puts that at 9 seconds (!!) on enwiki:Barack_Obama, see discussion on T348254).
To fix this issue, we want to provide a full DOM pipeline for the Parsoid output, which means implementing DOM versions of the current text transformations.

Related Objects
Search...

Status	Assigned	Task
Resolved	ihurbain	T293512 ParserOutput::getText() should be removed from ParserOutput
Open	ihurbain	T394005 Provide DOM versions of the OutputTransform pipeline transformations
Open	None	T405927 Provide a DOM version of AddRedirectHeader OutputTranform
Open	None	T405928 Provide a DOM version of the AddWrapperDivClass OutputTransform
Resolved	ssastry	T405929 Provide a DOM version of the DeduplicateStyles OutputTransform
Open	None	T405930 Provide a DOM version of the ExecutePostCacheTransformHooks OutputTransform
Open	None	T405932 Provide a DOM version of the ExpandToAbsoluteURLs OutputTransform
Resolved	ssastry	T405933 Provide a DOM version of the ExtractBody OutputTransform
Resolved	ihurbain	T405935 Provide a DOM version of the HandleTOCMarkers OutputTransform
Open	None	T405936 Provide a DOM version of the HydrateHeaderPlaceholders OutputTransform
Open	None	T405937 Provide a DOM version of the RenderDebugInfo OutputTransform

Event Timeline

ihurbain created this task.May 13 2025, 8:42 AM

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptMay 13 2025, 8:42 AM

ihurbain moved this task from Backlog to In Progress on the Content-Transform-Team (Work In Progress) board.May 13 2025, 8:43 AM

• mszabo mentioned this in T394059: Post-cache output transforms are expensive on large pages.May 13 2025, 3:29 PM

Alex44019 subscribed.May 14 2025, 12:22 PM

MSantos removed a project: Parsoid.May 22 2025, 2:27 PM

MSantos moved this task from In Progress to Q3 FY25-26 on the Content-Transform-Team (Work In Progress) board.

ssastry triaged this task as High priority.May 22 2025, 5:52 PM

Change #1150712 had a related patch set uploaded (by Isabelle Hurbain-Palatin; author: Isabelle Hurbain-Palatin):

[mediawiki/core@master] Introduce ContentHolderTransformStage

https://gerrit.wikimedia.org/r/1150712

gerritbot added a project: Patch-For-Review.May 26 2025, 3:20 PM

ihurbain removed ihurbain as the assignee of this task.Jul 18 2025, 2:45 PM

Change #1191726 had a related patch set uploaded (by Isabelle Hurbain-Palatin; author: Isabelle Hurbain-Palatin):

[mediawiki/core@master] Add a DOM version of the TOC markers pass

https://gerrit.wikimedia.org/r/1191726

cscott moved this task from Q3 FY25-26 to In Progress on the Content-Transform-Team (Work In Progress) board.Sep 29 2025, 3:27 PM

ihurbain added a subtask: T405928: Provide a DOM version of the AddWrapperDivClass OutputTransform.Sep 29 2025, 4:05 PM

ihurbain added a subtask: T405929: Provide a DOM version of the DeduplicateStyles OutputTransform.Sep 29 2025, 4:08 PM

ihurbain added a subtask: T405930: Provide a DOM version of the ExecutePostCacheTransformHooks OutputTransform.Sep 29 2025, 4:11 PM

ihurbain added a subtask: T405932: Provide a DOM version of the ExpandToAbsoluteURLs OutputTransform.

ihurbain added a subtask: T405933: Provide a DOM version of the ExtractBody OutputTransform.Sep 29 2025, 4:16 PM

ihurbain added a subtask: T405935: Provide a DOM version of the HandleTOCMarkers OutputTransform.

We explicitely exclude HardenNFC from this task as this is a text-only-pass that happens at the end of the pipeline, and makes no sense to convert to DOM.

ihurbain added a subtask: T405936: Provide a DOM version of the HydrateHeaderPlaceholders OutputTransform.Sep 29 2025, 4:20 PM

ihurbain added a subtask: T405937: Provide a DOM version of the RenderDebugInfo OutputTransform.

ihurbain claimed this task.Oct 6 2025, 3:29 PM

ihurbain closed subtask T405935: Provide a DOM version of the HandleTOCMarkers OutputTransform as Resolved.Oct 28 2025, 9:36 AM

ssastry closed subtask T405929: Provide a DOM version of the DeduplicateStyles OutputTransform as Resolved.Nov 10 2025, 4:21 PM

ssastry closed subtask T405933: Provide a DOM version of the ExtractBody OutputTransform as Resolved.

ihurbain moved this task from In Progress to Backlog on the Content-Transform-Team (Work In Progress) board.Dec 16 2025, 12:51 PM