Investigate Production Errors from ComputedUserImpactLookup::getPageViewData
Closed, ResolvedPublic3 Estimated Story Points
Actions

Assigned To

Authored By

	Michael
	Aug 1 2025, 2:13 PM

Description

We have a number of recent-ish production errors all related to the ComputedUserImpactLookup class, and in particular its getPageViewData method.

T399092: PHP Deprecated: DateTime::__construct(): Passing null to parameter #1 ($datetime) of type string is deprecated
T400574: PHP Warning: Undefined array key "mos:LABEL"
T400706: PHP Warning: Trying to access array offset on value of type null (via ComputedUserImpactLookup)

While they don't seem to be directly related to our efforts in T341599: Impact Module: improvements for former newcomers/T398500: [timebox: 3 days] Impact module: Support larger wgGEUserImpactMaxEdits, we probably want to address them before scaling the feature further and potentially worsen existing issues.

Since it seems plausible that there is a common cause, this tasks exists to consolidate the exploration of all three.

Acceptance criteria:

We know what's causing these errors
We have identified next steps, or if they were simple to fix, we fixed them.
If we haven't fixed them yet, we understand their potential impact on T341599

Details

Related Changes in Gerrit:

	Subject	Repo	Branch	Lines +/-
	fix(Impact): skip mismatched title in pageview data	mediawiki/extensions/GrowthExperiments	master	+12 -0

Customize query in gerrit

Related Objects
Search...

Status	Subtype	Assigned	Task
Open		None	T376276 PHP 8.4: Implicitly nullable parameter declarations deprecated
Open		None	T379587 PHP 8.4: Implicitly nullable parameter declarations deprecated in 3rd party vendor libs
Resolved		Reedy	T378424 PHP 8.4: Implicitly nullable parameter declarations deprecated in qrcode vendor libs
Open		None	T317255 PHP's getElementsByTagName is slow prior to PHP 8.3
Open		None	T358667 Drop PHP 8.2 support from MediaWiki
Open		None	T366644 Raise MediaWiki's MariaDB requirement to 10.6
Open		None	T366646 Raise Quibble jobs' tested version of MariaDB to 10.6
Declined		None	T405346 Remove okvpn/clock-lts post WMF PHP 8.3 upgrade
Resolved		Reedy	T405344 Upgrade web-auth/webauthn-lib to >= 5.2.3
Resolved		Krinkle	T358666 Drop PHP 8.1 support from MediaWiki
Resolved		Krinkle	T360995 Migrate Wikimedia production from PHP 8.1 to PHP 8.3
Resolved		cscott	T401855 ☂ PHP 8.3 issues found during WMF rollout
Resolved	PRODUCTION ERROR	Michael	T408852 PHP Deprecated: DateTime::__construct(): Passing null to parameter #1 ($datetime) of type string is deprecated
Resolved	PRODUCTION ERROR	Michael	T399092 PHP Deprecated: DateTime::__construct(): Passing null to parameter #1 ($datetime) of type string is deprecated
Resolved	PRODUCTION ERROR	Michael	T400706 PHP Warning: Trying to access array offset on value of type null (via ComputedUserImpactLookup)
Resolved	PRODUCTION ERROR	Michael	T400574 PHP Warning: Undefined array key "mos:LABEL"
Resolved		Michael	T401005 Investigate Production Errors from ComputedUserImpactLookup::getPageViewData

Event Timeline

Michael created this task.Aug 1 2025, 2:13 PM

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptAug 1 2025, 2:13 PM

Michael triaged this task as High priority.Aug 1 2025, 3:04 PM

Michael added parent tasks: T399092: PHP Deprecated: DateTime::__construct(): Passing null to parameter #1 ($datetime) of type string is deprecated, T400706: PHP Warning: Trying to access array offset on value of type null (via ComputedUserImpactLookup), T400574: PHP Warning: Undefined array key "mos:LABEL".

KStoller-WMF moved this task from Backlog to Estimated tasks backlog on the Growth-Team board.Aug 26 2025, 4:54 PM

KStoller-WMF set the point value for this task to 3.Sep 8 2025, 4:21 PM

DMburugu added a project: Essential-Work.Oct 6 2025, 4:24 PM

KStoller-WMF moved this task from Estimated tasks backlog to Needs Estimation on the Growth-Team board.Oct 17 2025, 10:48 PM

KStoller-WMF moved this task from Needs Estimation to Estimated tasks backlog on the Growth-Team board.

KStoller-WMF lowered the priority of this task from High to Medium.Nov 18 2025, 5:54 PM

KStoller-WMF moved this task from Estimated tasks backlog to FY2025-26 Q2 Sprint 4 on the Growth-Team board.

KStoller-WMF edited projects, added Growth-Team (FY2025-26 Q2 Sprint 4); removed Growth-Team.

Michael added a subtask: T408852: PHP Deprecated: DateTime::__construct(): Passing null to parameter #1 ($datetime) of type string is deprecated.Nov 21 2025, 3:01 PM

Michael removed a subtask: T408852: PHP Deprecated: DateTime::__construct(): Passing null to parameter #1 ($datetime) of type string is deprecated.Nov 21 2025, 3:09 PM

Michael added a parent task: T408852: PHP Deprecated: DateTime::__construct(): Passing null to parameter #1 ($datetime) of type string is deprecated.

Looking into this a bit, the errors mentioned in T400574: PHP Warning: Undefined array key "mos:LABEL" seem to be the most fruitful to investigate. In particular, I notice that all the titles that come up here as undefined array keys seem to contain a colon, like an interwiki link or sometimes a redirect:

mos:LABEL (siwiki)
R:Dúniya (crhwiki)
mos:DASH (siwiki)

Also, these errors seem to be rare:

I suggest skipping processing such a title when encountering it, and logging a warning to allow further investigation when it happens again. After that change is merged, I also recommend closing the related production errors because they should not happen again. Though this task should be kept open for further investigation.

There should be virtually no impact on T341599: in some very rare cases, a page might be missing pageview data, that's all.

Change #1208382 had a related patch set uploaded (by Michael Große; author: Michael Große):

[mediawiki/extensions/GrowthExperiments@master] fix(Impact): skip mismatched title in pageview data

https://gerrit.wikimedia.org/r/1208382

gerritbot added a project: Patch-For-Review.Nov 21 2025, 3:48 PM

Michael claimed this task.Nov 21 2025, 3:49 PM

Michael moved this task from Incoming to Code Review on the Growth-Team (FY2025-26 Q2 Sprint 4) board.

Restricted Application added a project: User-Michael. · View Herald TranscriptNov 21 2025, 3:49 PM

Motivated by @Michael's patch, I took a look at this as well. The titles listed by Michael are special: two of them are external titles (as mos is a language code, it refers to mos.wikipedia.org regardless of the Wikipedia it is used on). The last remaining one doesn't have any views logged in AQS:

urbanecm@wmf3345 ~ % curl -s 'https://wikimedia.org/api/rest_v1/metrics/pageviews/per-article/crh.wikipedia/all-access/user/R:D%C3%BAniya/daily/20250920/20251023' | jq .
{
  "detail": "The date(s) you used are valid, but we either do not have data for those date(s), or the project you asked for is not loaded yet. Please check documentation for more information",
  "method": "get",
  "status": 404,
  "title": "Not Found",
  "type": "about:blank",
  "uri": "/metrics/pageviews/per-article/crh.wikipedia/all-access/user/R:D%C3%BAniya/daily/20250920/20251023"
}
urbanecm@wmf3345 ~ %

This made me think "what is the result ComputedUserImpact gets" (the Pinnipedia query is there just to verify I was having the right expectations and did not mess something in the manual call):

>  $t = \MediaWiki\Title\Title::newFromText('R:Dúniya')                                                                                                                                                                                                                       
= MediaWiki\Title\Title {#5335                                                                                                                                                                                                                                                
    +mArticleID: -1,                                                                                                                                                                                                                                                          
    +prefixedText: null,                                                                                                                                                                                                                                                      
    +mRedirect: null,                                                                                                                                                                                                                                                         
  }                                                                                                                                                                                                                                                                           
                                                                                                                                                                                                                                           
> $l = \MediaWiki\MediaWikiServices::getInstance()->get('GrowthExperimentsUserImpactLookup_Computed')                                                                                                                                      
= GrowthExperiments\UserImpact\ComputedUserImpactLookup {#5355}                                                                                                                                                                            
                                                                                                                                                                                                                                           
>  $titleObjects = []                                                                                                                                                                                                                      
= []                                                                                                                                                                                                                                       
                                                                                                                                                                                                                                           
> $titleObjects['R:Dúniya'] = ['title' => $t]                                                                                                                                                                                              
= [                                                                                                                                                                                                                                        
    "title" => MediaWiki\Title\Title {#5335                                                                                                                                                                                                
      +mArticleID: -1,                                                                                                                                                                                                                     
      +prefixedText: null,                                                                                                                                                                                                                 
      +mRedirect: null,                                                                                                                                                                                                                    
    },                                                                                                                                                                                                                                     
  ]                                                                                                                                                                                                                                        
                                                                                                                                                                                                                                           
> sudo $d =$l->getPageViewDataInJobContext($titleObjects, User::newFromName('Martin Urbanec (WMF)'), 60)                                                                                                                                   
= []     

> $titleObjects['Pinnipedia'] = ['title' => \MediaWiki\Title\Title::newFromText('Pinnipedia')]                                                                                                                                             
= [                                                                                                                                                                                                                                        
    "title" => MediaWiki\Title\Title {#8364                                                                                                                                                                                                
      +mArticleID: -1,                                                                                                                                                                                                                     
      +prefixedText: null,                                                                                                                                                                                                                 
      +mRedirect: null,                                                                                                                                                                                                                    
    },                                                                                                                                                                                                                                     
  ]                                                                                                                                                                                                                                        
                                                                                                                                                                                                                                           
> sudo $d =$l->getPageViewDataInJobContext($titleObjects, User::newFromName('Martin Urbanec (WMF)'), 60)                                                                                                                                   
= [                                                                                                                                                                                                                                        
    "Pinnipedia" => [                                                                                                                                                                                                                      
      "2025-09-27" => 2,                                                                                                                                                                                                                   
      "2025-09-28" => 2,
      "2025-09-29" => 0,
      "2025-09-30" => 1,
      [...]
  ]

>

So, maybe we're just not dealing correctly with pages that were not visited at all in the last 60 days?

Change #1208382 merged by jenkins-bot:

[mediawiki/extensions/GrowthExperiments@master] fix(Impact): skip mismatched title in pageview data

https://gerrit.wikimedia.org/r/1208382

Maintenance_bot removed a project: Patch-For-Review.Nov 26 2025, 12:31 PM

Urbanecm_WMF edited projects, added Growth-Team (FY2025-26 Q2 Sprint 5); removed Growth-Team (FY2025-26 Q2 Sprint 4).Dec 2 2025, 5:51 PM

Urbanecm_WMF moved this task from Incoming to Code Review on the Growth-Team (FY2025-26 Q2 Sprint 5) board.

Not sure why that tag was missing here. Once the train has been fully rolled out, I think we can close the directly related 4 parent production errors as they should no longer happen.

This has rolled out and the error does not seem to have happened since. We'll see it as a Growth warning if it ever does. It will have the message Title {title} not found in allTitleObjects.

Michael mentioned this in T399092: PHP Deprecated: DateTime::__construct(): Passing null to parameter #1 ($datetime) of type string is deprecated.Dec 16 2025, 6:01 PM

Michael mentioned this in T400706: PHP Warning: Trying to access array offset on value of type null (via ComputedUserImpactLookup).

Michael mentioned this in T408852: PHP Deprecated: DateTime::__construct(): Passing null to parameter #1 ($datetime) of type string is deprecated.

Michael mentioned this in T400574: PHP Warning: Undefined array key "mos:LABEL".