[devbite-enzo]

• I have followed (at least) the [PR section of the contributing guide]
  (https://github.com/mui/material-ui/blob/HEAD/CONTRIBUTING.md#sending-a-pull-request).

Same issue as #49213, in Autocomplete.

The chip, clearIndicator, paper, popper and popupIndicator slot props are typed as SlotProps<React.ElementType<Partial<XProps>>, …>. Every prop of Partial<XProps> is optional, so it fits almost every intrinsic element. React.ElementType<…> then expands to a union of ~476 HTML/SVG tags, and SlotComponentProps builds the props of each one. Whenever two different AutocompleteProps instantiations are compared, TypeScript has to measure the variance of all six type parameters through these unions. That happens with wrapper components that spread props or slotProps, which is the common pattern for form-library wrappers.

This PR types the slots by the component they render, as #49213 did for AvatarGroup. IconButton, Paper and Popper use typeof Component. chip uses React.JSXElementConstructor<Partial<ChipProps<ChipComponent>>> so it keeps the ChipComponent generic. That matches how AutocompleteSlots already types the same slots.

Performance gain

A wrapper in the usual shape:

type WrapperProps = Omit<
  AutocompleteProps<unknown, boolean | undefined, boolean | undefined, boolean | undefined>,
  'renderInput'
>;

function Wrapper({ slotProps, ...other }: WrapperProps) {
  return (
    <Autocomplete
      {...other}
      renderInput={() => null}
      slotProps={{ ...slotProps, chip: { size: 'small', ...slotProps?.chip } }}
    />
  );
}

tsc --noEmit --extendedDiagnostics on that file alone, inside packages/mui-material:

	master	this PR
Check time	3.39 s	1.31 s
Instantiations	2,048,830	1,778,514
Memory (RSS)	1,347 MB	813 MB

In a production Next.js app (~1,400 source files, TypeScript 7), computing the variance of AutocompleteProps took 2.9 s per checker. A full tsc --noEmit went from 20.9 s / 63 s CPU / 7.3 GB to 13.1 s / 29 s CPU / 3.0 GB with this change applied as a local patch (and --checkers 2). Since tsgo computes the variance in each checker separately, the cost multiplies with parallelism.

Compatibility

The props each slot accepts are unchanged, including component on paper. Only the props of unrelated intrinsic elements drop out of the union. The new spec cases cover the accepted props and the rejection of unknown ones. pnpm proptypes and pnpm docs:api produce no changes.

🤖 Generated with Claude Code

[code-infra-dashboard]

Deploy preview

https://deploy-preview-49634--material-ui.netlify.app/

Bundle size

Bundle	Parsed size	Gzip size
@mui/material	0B(0.00%)	0B(0.00%)
@mui/lab	0B(0.00%)	0B(0.00%)
@mui/private-theming	0B(0.00%)	0B(0.00%)
@mui/system	0B(0.00%)	0B(0.00%)
@mui/utils	0B(0.00%)	0B(0.00%)

Details of bundle changes

Performance

No regressions in 32 benchmarks · ran 3m 55s · details

---

Check out the code infra dashboard for more information about this PR.