Implementation Comparison

The shared comparison runner lives in scripts/compare-impls.mjs because this repo owns the corpus. It compares sibling implementation checkouts against the same .crv / .html pairs and reports default conformance, optional Tier-2 adapter coverage, rough CLI timing, and the extension hook surface each implementation exposes.

Snapshot (2026-05-31)

Historical run from 2026-05-31, when the corpus held 154 pairs. The corpus has since grown (now ~200 pairs); regenerate with npm run compare:impls for current figures. The numbers below are from that dated run.

154 / 154Rust corpus pass

154 / 154JS corpus pass

154 / 154PHP corpus pass

0cross-implementation diffs

Implementation	Commit	Corpus	Mismatches	Errors	Avg CLI ms/file
Rust	07ea8ea	154 / 154	0	0	33.83
JS	48c45e0	154 / 154	0	0	58.92
PHP	b1fa677	154 / 154	0	0	73.96

Spec commit: 840b64f

Optional Tier-2 Profile

The optional profile enables a shared adapter per feature where each implementation exposes one. Unsupported feature/implementation combinations are reported as skipped, not failures.

Feature	Rust	JS	PHP
Social link templates	pass	pass	pass
Emoji map	pass	pass	skipped
German smart quotes	skipped	skipped	pass
Bare URL autolink	skipped	skipped	pass

Implementation	Optional pass	Skipped	Mismatches	Errors	Avg CLI ms/file
Rust	2 / 2	2	0	0	63.93
JS	2 / 2	2	0	0	94.26
PHP	3 / 3	1	0	0	103.30

Optional cross-implementation diffs: 0

CLI Timing

These timings include process startup and should be read as smoke-level CLI performance, not parser microbenchmarks.

Rust

33.83 ms

JS

58.92 ms

PHP

73.96 ms

Extension Surface

The comparison run is default/no-opt-in, so extension behavior is not yet exercised across every min/max profile. This matrix records the hook surface available in each implementation today.

Capability	Rust	JS	PHP
Inline matcher	yes	no	yes
Block matcher	yes	no	yes
After-parse transform	yes	yes	yes
Before-render transform	yes	yes	yes
Inline extension renderer	yes	yes	yes
Block extension renderer / render listener	yes	no	yes
Converter-level registration	no	no	yes

Running It

npm run compare:impls
npm run compare:impls -- --corpus=optional
npm run compare:impls -- --limit=20 --bench

By default the script expects sibling checkouts:

• ../carve-rs
• ../carve-js
• ../carve-php

Override those paths with CARVE_RS_DIR, CARVE_JS_DIR, and CARVE_PHP_DIR.

The documented snapshot used:

CARVE_RS_DIR=/media/mark/data/work/git/carve-rs \
CARVE_JS_DIR=/media/mark/data/work/git/carve-js \
CARVE_PHP_DIR=/media/mark/data/work/git/carve-php \
node scripts/compare-impls.mjs

Default raw output:

Implementation summary
profile=default/no-opt-in corpus=core corpus_pairs=154
rust: pass=154/154 mismatch=0 error=0 skipped=0 avg_ms=33.83
js: pass=154/154 mismatch=0 error=0 skipped=0 avg_ms=58.92
php: pass=154/154 mismatch=0 error=0 skipped=0 avg_ms=73.96
cross_impl_diffs=0

Extension capability matrix
rust: inline matcher, block matcher, after_parse, before_render, inline extension renderer, block extension renderer
js: afterParse, beforeRender, inline extension renderer
php: inline matcher, block matcher, parsed-document hook, before-render hook, render listeners, converter registration
extension_profile_note=this run compares default/no-opt-in output. Use --corpus=optional for Tier-2 opt-in adapters.

Optional raw output:

Implementation summary
profile=optional/opt-in corpus=optional corpus_pairs=4
rust: pass=2/2 mismatch=0 error=0 skipped=2 avg_ms=63.93
js: pass=2/2 mismatch=0 error=0 skipped=2 avg_ms=94.26
php: pass=3/3 mismatch=0 error=0 skipped=1 avg_ms=103.30
cross_impl_diffs=0

Optional feature coverage
social-link-templates: rust, js, php
emoji-map: rust, js
smart-quotes-locale-de: php
bare-url-autolink: php

Scope

The tool has two profiles:

• It runs the mandatory Tier-1 corpus in tests/corpus.
• It runs optional Tier-2 adapters in tests/corpus-optional with --corpus=optional.
• It compares byte-identical output after trimming.
• It reports CLI-level average time per corpus file.
• It reports extension system surface area.

Tier-3 app-extension max profiles still need language-specific adapter fixtures. That means a small runner per implementation that enables the same test extension in each language, then feeds those through the same comparison loop.