Edoardo Luppi

@mike-lischke you were right about machine differences. I've cloned your repo and run the benchmarks. I've updated the results, and they're comparable now, see table above.

> ANTLR started out with split arrays :-D A while ago there were sections that got concatenated at load time. Using a string however is all but optimal, as it...

I'll open a PR as soon as the others are merged, just to avoid too much stacking. I've removed all parse error catching, so if it fails it's unrecoverable, and...

> You mean all catch clauses in generated code Oh no no. Initially I had copied the try-catch you have in the parsing service, which accumulates errors. I did get...

I've just added a benchmark for the WASM (`wasmJs`) target. See updated [table](https://github.com/Strumenta/antlr-kotlin/issues/165#issuecomment-1914879063). Overall it seems WebAssembly on Node.js cuts the times in half or more compared to JS.

> Can you explain more how the WASM target is structured? The runtime and the benchmark code are both compiled down to [optimized](https://github.com/WebAssembly/binaryen) WebAssembly binaries. The only place where I...

> However a, say, C++ developer, who wants the parser code to be in C++ cannot use this WASM binary, right? You can use the binary if you have a...

This really shows how much we need https://youtrack.jetbrains.com/issue/KT-47038. This type of issue would be entirely avoidable.

You can assign this one to me in case. I'll take a look in a couple weeks.

@JaneWardSandy what's stopping you here? Maybe I can help in clarifying how to go further.