👷‍♀️ [RUM 5088] Reduce INP Null Target

[cy-moi]

Motivation

Save the selector in a map with timestamp as key to mitigate the null target situation.
More information in RFC.

Changes

After mitigation, about 75% of null target INP would have a selector.

Testing

• Local
• Staging
• Unit
• End to end

---

I have gone over the contributing documentation.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 94.73684% with 1 line in your changes missing coverage. Please review.

Project coverage is 93.66%. Comparing base (74609cc) to head (6e862a6).
Report is 5 commits behind head on main.

Files with missing lines	Patch %	Lines
...core/src/domain/action/interactionSelectorCache.ts	90.90%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2950      +/-   ##
==========================================
+ Coverage   93.19%   93.66%   +0.47%     
==========================================
  Files         275      276       +1     
  Lines        7610     7626      +16     
  Branches     1707     1711       +4     
==========================================
+ Hits         7092     7143      +51     
+ Misses        518      483      -35     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[cit-pr-commenter]

Bundles Sizes Evolution

📦 Bundle Name	Base Size	Local Size	𝚫	𝚫%	Status
Rum	160.95 KiB	161.21 KiB	267 B	+0.16%	✅
Logs	56.00 KiB	56.00 KiB	0 B	0.00%	✅
Rum Slim	109.53 KiB	109.79 KiB	267 B	+0.24%	✅
Worker	25.21 KiB	25.21 KiB	0 B	0.00%	✅

🚀 CPU Performance

Action Name	Base Average Cpu Time (ms)	Local Average Cpu Time (ms)	𝚫
addglobalcontext	0.001	0.002	0.000
addaction	0.030	0.040	0.011
addtiming	0.001	0.001	-0.000
adderror	0.031	0.034	0.003
startstopsessionreplayrecording	0.846	0.887	0.041
startview	0.924	1.047	0.123
logmessage	0.019	0.020	0.001

🧠 Memory Performance

Action Name	Base Consumption Memory (bytes)	Local Consumption Memory (bytes)	𝚫 (bytes)
addglobalcontext	7.67 KiB	10.08 KiB	2.41 KiB
addaction	38.77 KiB	40.74 KiB	1.97 KiB
addtiming	6.05 KiB	7.72 KiB	1.68 KiB
adderror	42.53 KiB	45.44 KiB	2.91 KiB
startstopsessionreplayrecording	3.72 KiB	4.02 KiB	303 B
startview	458.87 KiB	467.43 KiB	8.55 KiB
logmessage	39.88 KiB	40.29 KiB	415 B

🔗 RealWorld

[cy-moi]

/to-staging

[BenoitZugmeyer]

💬 suggestion: ‏instead of checking for set / get/delete calls on Map, why not check if the expected value is actually in interactionSelectorMap directly?

Suggested change

	const setInteractionMapSpy = spyOn(Map.prototype, 'set').and.callThrough()
	emulateClick({ activity: { on: 'pointerdown' } })
	expect(setInteractionMapSpy).toHaveBeenCalled()
	emulateClick({ activity: { on: 'pointerdown' }, timeStamp: xxx })
	expect(interactionMap.get(xxx)).toBe(xxx)

I think it would be more explicit and a bit "safer" (we are sure the value was set on the right Map)

[cy-moi]

Thank you! The tests should be fixed with checking the map directly. This way is indeed more effective! TIL