Description

T317070 and T215000 both dealt with non-linear performance of getElementsByTagName in PHP DOM's implementation.

Unless someone wants to take on fixing PHP DOM's implementation, we should replace its use in our code bases with Zest::find or Parsoid's DOMCompat::querySelectorAll or DOMCompat::getElementsByTagName both of which are wrappers around Zest.

DOMCompat::findElementsByTagName seems the simplest replacement since it leaves open the possibility of easy switching back if/when PHP DOM perf gets fixed.

Code search reveals Flow, DonationsInterface, Translate, GWToolset, TextExtracts all use this method (https://codesearch.wmcloud.org/search/?q=getElementsByTagName&i=nope&files=php&excludeFiles=)

@Esanders also says: "And also a lint rule to prevent it being used in the future."

Details

	Project	Branch	Lines +/-	Subject
	mediawiki/core	master	+2 -1	VueComponentParser: Use Zest's getElementsByTagName() rather than PHP's

Customize query in gerrit

Related Objects
Search...

		Status	Subtype	Assigned	Task
		Open		None	T317255 Migrate code away from use of PHPDOM's getElementsByTagName which has non-linear perf characteristics
		Resolved		Wangombe	T317523 Migrate Translate extension away from use of PHPDOM's getElementsByTagName

Event Timeline

ssastry created this task.Sep 7 2022, 10:23 PM

Restricted Application added a project: Growth-Team. · View Herald TranscriptSep 7 2022, 10:23 PM

Restricted Application added a subscriber: Aklapper. · View Herald Transcript

Maintenance_bot added a project: Commons.Sep 7 2022, 10:30 PM

matmarex added a subscriber: matmarex.Sep 7 2022, 11:01 PM

ssastry updated the task description. (Show Details)Sep 8 2022, 12:21 AM

The uses in Translate are a maintenance script and Xliff file format handler which is currently unused: https://codesearch.wmcloud.org/search/?q=getElementsByTagName&i=nope&files=php&excludeFiles=&repos=Extension:Translate

What replacements are present in MediaWiki core version 1.37 and above?

Esanders updated the task description. (Show Details)Sep 8 2022, 11:27 AM

In T317255#8219789, @Nikerabbit wrote:

The uses in Translate are a maintenance script and Xliff file format handler which is currently unused: https://codesearch.wmcloud.org/search/?q=getElementsByTagName&i=nope&files=php&excludeFiles=&repos=Extension:Translate

What replacements are present in MediaWiki core version 1.37 and above?

The mentioned Zest::find is present in 1.37 (indeed, since 1.35.0-wmf.20):

https://gerrit.wikimedia.org/r/plugins/gitiles/mediawiki/vendor/+/refs/heads/REL1_37/wikimedia/zest-css/src/Zest.php#41

The easy fix is to replace $node->getElementsByTagName( $tag ) with Zest::getElementsByTagName( $node, $tag ) (which unlike find has identical signature).