We rewrote our Rust WASM Parser in TypeScript – and it got 3x Faster

4 hours ago

The openui-lang parser was initially built in Rust and compiled to WASM for performance, but optimization efforts revealed the WASM boundary overhead was the bottleneck.
The parser pipeline consists of six stages: Autocloser, Lexer, Splitter, Parser, Resolver, and Mapper.
Attempts to optimize by skipping JSON serialization with serde-wasm-bindgen resulted in 30% slower performance due to fine-grained conversions across runtime boundaries.
The real solution was porting the parser to TypeScript, eliminating the WASM boundary entirely, resulting in 2.2x to 4.6x faster per-call performance.
An O(N²) streaming inefficiency was addressed by implementing statement-level incremental caching, reducing total streaming cost by 2.6x to 3.3x.
WASM is beneficial for compute-bound tasks with minimal interop, portable native libraries, but not for frequently-called functions on small inputs.
Key takeaways include profiling actual time spent, understanding the costs of WASM-JS boundary crossings, and prioritizing algorithmic improvements over language-level optimizations.

Hasty Briefsbeta