Page MenuHomePhabricator

[timebox: 3 days] Impact module: Support larger wgGEUserImpactMaxEdits
Open, MediumPublic3 Estimated Story Points

Description

Background

The Growth-Team tried increasing the maximum number of edits in Impact module to 10,000 edits at the English Wikipedia, see T341599: Impact Module: improvements for former newcomers. Unfortunately, that increase triggered a lot of fatal errors, see T398418: TypeError: array_map(): Argument #2 ($array) must be of type array, int given. We fixed them by reverting the changes, but we need to adjust the Impact module in a way so that it works with larger values for wgGEUserImpactMaxEdits (10,000 is the current goal).

The underlying technical problem is related to the refreshUserImpactJob. On every edit, we:

  1. compute user impact data,
  2. store it into the database,
  3. pass the computed value to the refreshUserImpactJob,
  4. the job computes its own version of user impact data and updates the database

The issue happens, because the job (generated at step 3) is larger than the maximum size, and is refused by the event gate.

Questions to resolve
  1. Why are we using the "compute-schedule-compute-again" approach? If we want the computation to happen within the job, wouldn't it make more sense to just schedule the job and let it compute?
  2. Is it necessary for us to pass the impact data blob to the job? Can we avoid that to ensure the job is not becoming too large?
  3. What is the maximum size of a job params?
Acceptance Criteria

Determine the next steps that are needed to increase the wgGEUsereImpactMaxEdits value.

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald Transcript
Michael renamed this task from Impact module: Support larger wgGEUserImpactMaxEdits to [timebox: 3 days] Impact module: Support larger wgGEUserImpactMaxEdits.Jul 7 2025, 4:07 PM
Michael triaged this task as High priority.
Michael set the point value for this task to 3.
KStoller-WMF lowered the priority of this task from High to Medium.Jul 30 2025, 4:15 PM

Is there an estimate of when this task might move forward?

@kostajh - we don't have an estimate, as this keeps getting pushed off since it doesn't directly relate to our annual plan commitments.

That being said, I think it's a valuable improvement for junior and experienced editors, and if this is something that is high priority for Trust & Safety Product, then perhaps Growth can consider a small hypothesis under WE4.2 to support the User Info card priority? Do you think that makes sense?

@kostajh - we don't have an estimate, as this keeps getting pushed off since it doesn't directly relate to our annual plan commitments.

That being said, I think it's a valuable improvement for junior and experienced editors, and if this is something that is high priority for Trust & Safety Product, then perhaps Growth can consider a small hypothesis under WE4.2 to support the User Info card priority? Do you think that makes sense?

Sorry I missed replying to this. I'm not sure about having this as a hypothesis in WE4.2. Could we try increasing the limit on some production Wikipedias (e.g. enwiki) and seeing if it results in any issues?

Could we try increasing the limit on some production Wikipedias (e.g. enwiki) and seeing if it results in any issues?

I know when we attempted this previously T341599: Impact Module: improvements for former newcomers we ran into issues: T398418: TypeError: array_map(): Argument #2 ($array) must be of type array, int given.

This has been on hold until we wrap up some annual plan efforts, but I would love to see this move forward!