Compile-time experiment: table-driven dispatch for giant per-property match statements

[nicoburns]

Summary

This is a compile-time experiment, not a functional change. It replaces the largest mako-generated per-property match statements with static arrays of function pointers indexed by property-id discriminant, to test the hypothesis that giant 200+-arm matches are what makes the stylo crate slow to compile.

Converted (largest generated matches by line count in the generated properties.rs):

• AnimationValue::from_declaration (~10,300 generated lines, incl. the nested CSSWideKeyword match) → FROM_DECL / FROM_KEYWORD tables of per-property fns
• ComputedValues::computed_or_resolved_declaration (~3,200 lines) → TO_DECL table
• ComputedValues::computed_or_resolved_value (~2,800 lines) → TO_CSS table
• impl Animate for AnimationValue (~2,700 lines) → ANIMATE table
• ComputedValues::property_value_to_typed_value_list (~1,800 lines) → TO_TYPED table
• AnimationValue::from_computed_values, AnimationValue::uncompute, AnimationValue::set_in_style_for_servo, AnimationValue::is_different_for, impl ComputeSquaredDistance for AnimationValue → corresponding tables

Each table is [fn(...); property_counts::LONGHANDS] with one generated function per (physical) longhand; non-applicable slots point at a shared None/unreachable stub. Dispatch reads the enum tag (same repr tricks the existing code already uses) and indexes the table. Behavior and public API are unchanged.

Measurements (2-core machine, rustc 1.95.0, default servo feature, debug profile)

Metric	Before	After
Cold cargo build (workspace)	94 s	91 s
stylo unit time (cargo --timings)	50.8 s	51.6 s
touch style/lib.rs && cargo check (CARGO_INCREMENTAL=0, 3 runs)	19.0 / 19.4 / 19.7 s	19.7 / 19.7 / 19.8 s
Generated properties.rs lines	123,136	140,085

Result: neutral. On this toolchain, converting these matches to table-driven dispatch does not measurably improve compile times (differences are within run-to-run noise; the generated file actually grows ~17k lines from the per-property functions). The giant matches do not appear to be the frontend bottleneck for this crate on rustc 1.95.

Verification

• cargo check passes with no new warnings.
• cargo test --workspace --exclude stylo passes. (stylo's own lib tests fail to compile on main too — pre-existing PseudoElement::MozRubyText/FirstLine test breakage in style/selector_parser.rs, unrelated to this change.)

Link to Devin session: https://app.devin.ai/sessions/c1053e755c5f431990f0fd9ec986740c
Requested by: @nicoburns

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR that start with 'DevinAI' or '@devin'.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring