perf(ast): box typescript enum variants.

[rzvxa]

Similar to #3058 and #3061 it is a continuation of #3047.

Handles these enum types:

TSEnumMemberName
Variant sizes: 16, 24, 24, 40
Unboxed variants: IdentifierName (struct), StringLiteral (struct), NumericLiteral (struct)
Dependents: TSEnumMember (struct)
=> Box all variants.

TSModuleReference
Variant sizes: 16, 32
Unboxed variants: TSExternalModuleReference (struct)
Dependents: Box in TSImportEqualsDeclaration
=> Box all variants. Replace Box with TSModuleReference in TSImportEqualsDeclaration.

TSTypePredicateName
Variant sizes: 8, 24
Unboxed variants: IdentifierName (struct), TSThisType (struct)
Dependents: TSTypePredicate (struct)
=> Box Identifier variant. Do not box This variant as only 8 bytes (just contains Span).

TSTypeQueryExprName
Variant sizes: 16, 88
Unboxed variants: TSImportType (struct)
Dependents: TSTypeQuery (struct)
=> Box TSImportType variant. Do not box TSTypeName variant, as is another enum.

[rzvxa]

• perf(ast): box typescript enum variants. #3065 👈
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @rzvxa and the rest of your teammates on Graphite

[codspeed-hq]

CodSpeed Performance Report

Merging #3065 will not alter performance

_{Comparing 04-22-perf_ast_box_typescript_enum_variants (3ef34ff) with main (84c43c8)}

Summary

✅ 30 untouched benchmarks

[overlookmotel]

Oh damn it! I made 2 mistakes on my list:

• TSEnumMemberName::ComputedPropertyName should not be boxed as Expression is already an enum.
• TSModuleReference::TypeName should not be boxed as TSTypeName is already an enum.

Would you have time to change those two back in unboxed? Sorry I've wasted your time with my stupid mistakes.

[rzvxa]

Oh damn it! I made 2 mistakes on my list:

* `TSEnumMemberName::ComputedPropertyName` should not be boxed as `Expression` is already an enum.

* `TSModuleReference::TypeName` should not be boxed as `TSTypeName` is already an enum.

Would you have time to change those two back in unboxed? Sorry I've wasted your time with my stupid mistakes.

No worries, It was partially my own fault I didn't check the context and mindlessly just did what was written in the description.

[overlookmotel]

Thanks for making the changes @rzvxa.

As with the other PRs in this group, the benchmarks are inconclusive. It's not making much difference either way. Personally I think there's a benefit to a consistent pattern in the types, so I've reviewed as approved. But please can you give your sign-off @Boshen?