Improve type check performance

[ypresto]

• I have followed (at least) the PR section of the contributing guide.

Implements performance improvement for:

• makeStyles/createStyles: Can the typings be simplified to improve performance? #19113 (comment)
• styled(): Can the typings be simplified to improve performance? #19113 (comment)
• OverridableComponent: TBD

TODO: "< 10 includes to primitives" issue should be sent to TypeScript.

[mui-pr-bot]

Bundle size will be reported once Azure build #30886 finishes.

Generated by 🚫 dangerJS against bc5041c

[rvion]

what is missing to accept this ?

[mnajdova]

I apologize, but so far we didn't have bandwidth to dive deeper into TS perf issues. @ypresto are you still interested in continuing this effort?

I have started some investigation, the first one being around the OverridableComponent performance. I am using tsc --generateTrace to compare the results of the time it takes to check the types before & after the changes.

I can try the styled() changes in a project and report back the data I will get. I guess this specific change will depend on which params are used when the styled() method is defined.

[ypresto]

I was too busy this year and still busy. 🙏
But it is glad for me if you think my investigation is valuable (...also I'm open for second job 😁...)

I'll organize performance pitfalls and caveats, and try to improve TypeScript itself.

(Note that TS perf fix for unions with many primitives is already landed in v4.5.2. microsoft/TypeScript#45220 )

I guess this specific change will depend on which params are used when the styled() method is defined.

You mean, no AdditionalProps nor template type is used for styled(Button)(HERE)? If so, you are right.
As discussed in #19113, instantiating or comparing complicated types will be cost so much.
So this PR will use simpler alternative types for where it is enough.
But still, where expression which cannot be covered by simpler one is used, this does not improve performance.
More investigation is still necessary for it.

[mnajdova]

But it is glad for me if you think my investigation is valuable (...also I'm open for second job 😁...)

I'll just drop this here - https://mui.com/careers/react-engineer-core/ 😁 😉

You mean, no AdditionalProps nor template type is used for styled(Button)(HERE)? If so, you are right.
As discussed in #19113, instantiating or comparing complicated types will be cost so much.
So this PR will use simpler alternative types for where it is enough.
But still, where expression which cannot be covered by simpler one is used, this does not improve performance.
More investigation is still necessary for it.

Thanks, yeah this answers my question.

This is my investigation on the OverridableComponent -> #32735 would be great to hear your opinion on.

Also if I am being honest, I would focus on improving the types on the new v5 utils at this point, I don't think it's worth spending time on the legacy styling solution as it is not React 18 compatible and I don't expect the usage to be growing in the future.

[mnajdova]

@ypresto thanks for the investigations done in the PR. As there weren't any recent updates, I am closing it. Feel free to re-open a new issue if you still want to investigate some type checks perfs. Also, I would recommend creating PR per module, so that it is easier for review :)