Consider what to do with validation
Closed, ResolvedPublic
Actions

Assigned To

Authored By

	cmassaro
	Jan 4 2026, 7:24 PM

Project Tags

Referenced Files

None

Subscribers

cmassaro

GrounderUK

Jdforrester-WMF

Description

Validation has been largely deprecated, except in the case of custom validator functions. Consider whether even those functions are adding value.

ensure that validation functions are run
ensure that these functions are run only once (i.e., once validated, an object is not validated again)

Completion checklist

Before closing this task, review one by one the checklist available here: https://www.mediawiki.org/wiki/Abstract_Wikipedia_team/Definition_of_Done#Back-end_Task/Bug_completion_checklist

Details

Related Changes in Gerrit:

	Subject	Author	Repo	Branch	Lines +/-
	wikifunctions: Upgrade orchestrator from 2026-02-18-140059 to 2026-02-25-124326	Jforrester	operations/deployment-charts	master	+1 -1

Customize query in gerrit

Related Changes in GitLab:

	Title	Reference	Author	Source Branch	Dest Branch
	Add support for validation in composition language v2.	repos/abstract-wiki/wikifunctions/function-orchestrator!560	apine	apine-validation-v2	main

Customize query in GitLab

Related Objects
Search...

Status	Assigned	Task
Resolved	cmassaro	T419051 Refine Composition Language V2
Resolved	Jdforrester-WMF	T413729 Ensure that all current unit tests pass with the new /v2/evaluate
Resolved	cmassaro	T413731 Consider what to do with validation
Resolved	cmassaro	T413732 Build the initial transpilation system and wire it up for use in a new endpoint

Event Timeline

cmassaro created this task.Jan 4 2026, 7:24 PM

cmassaro added a parent task: T413729: Ensure that all current unit tests pass with the new /v2/evaluate.

cmassaro added a subtask: T413732: Build the initial transpilation system and wire it up for use in a new endpoint.

cmassaro closed subtask T413732: Build the initial transpilation system and wire it up for use in a new endpoint as Resolved.Jan 7 2026, 4:47 PM

Jdforrester-WMF edited projects, added: Abstract Wikipedia team (26Q3 (Jan–Mar)), OKR-Work; removed: Abstract Wikipedia team.Jan 7 2026, 6:39 PM

Jdforrester-WMF moved this task from Incoming to Ready on the Abstract Wikipedia team (26Q3 (Jan–Mar)) board.Jan 7 2026, 6:48 PM

DSantamaria triaged this task as High priority.Jan 8 2026, 3:15 PM

cmassaro claimed this task.Feb 10 2026, 2:09 PM

I’ve raised a couple of tickets relating to type-guarantee failures: T411565 and T417631. These guarantees are currently costly to enforce in user functions. See, for example apply two-argument function: validated arguments (Z31684), with this test.

In T413731#11627238, @GrounderUK wrote:

I’ve raised a couple of tickets relating to type-guarantee failures: T411565 and T417631. These guarantees are currently costly to enforce in user functions. See, for example apply two-argument function: validated arguments (Z31684), with this test.

This is really helpful!!! We don't have a good way to assess how useful validation is, so examples like these are important.

In your experience, is validation mostly useful for inputs? Or do people expect it to be run on a function's return value, too?

Most critically, I would say it is at the interface between functions. A called function benefits from static guarantees provided by the composition itself, but if the interface is not strictly typed on both sides, that guarantee can fail at runtime, with downstream consequences that are difficult to trace (or even identify as having occurred), especially when the interface is encapsulated within another function.

At present, incorporating intermediate type-checking is particularly inefficient because there is no built-in type-equality function, nor a direct object-typing primitive. Our type of object (Z16829) function is currently implemented via value by key (safer) (Z22475) to avoid failures for string objects in Z803 T389487 but, not surprisingly, this has a noticeable performance impact.

In T413731#11627321, @cmassaro wrote:

In T413731#11627238, @GrounderUK wrote:

I’ve raised a couple of tickets relating to type-guarantee failures: T411565 and T417631. These guarantees are currently costly to enforce in user functions. See, for example apply two-argument function: validated arguments (Z31684), with this test.

This is really helpful!!! We don't have a good way to assess how useful validation is, so examples like these are important.

In your experience, is validation mostly useful for inputs? Or do people expect it to be run on a function's return value, too?

cmassaro updated the task description. (Show Details)Feb 19 2026, 11:07 AM

cmassaro changed the task status from Open to In Progress.Feb 19 2026, 8:48 PM

cmassaro moved this task from Ready to In Engineering on the Abstract Wikipedia team (26Q3 (Jan–Mar)) board.

Ah! I think I misunderstood something. The validator functions currently used in practice are not the same as what the backend formally defines as validators. I still plan to implement this functionality, but it looks like it's never used in practice.

apine opened https://gitlab.wikimedia.org/repos/abstract-wiki/wikifunctions/function-orchestrator/-/merge_requests/560

Add support for validation in composition language v2.

cmassaro moved this task from In Engineering to In Code review on the Abstract Wikipedia team (26Q3 (Jan–Mar)) board.Feb 20 2026, 8:52 PM

ecarg merged https://gitlab.wikimedia.org/repos/abstract-wiki/wikifunctions/function-orchestrator/-/merge_requests/560

Add support for validation in composition language v2.

Maintenance_bot removed a project: Patch-For-Review.Feb 23 2026, 8:32 PM