real-world HTML -> html_to_markdown() -> markdown_to_html()

Repository diagnostics · JSON · JSONL

Every fixture here is a real, vendored page (see fixtures/provenance.yaml) run through both conversions, offline -- no live fetch of the real site happens in this report. See tests/test_benchmark_html_roundtrip.py for the exact construct-count assertions this data backs.

Commit 3b0ec39a · main · 2026-10-17T21:10:19+09:00 · Merge pull request #109 from myon-bioinformatics/perf/check-png-before-browser · (dirty)

tohoho_web_home

Source: https://www.tohoho-web.com/ (top page, live site, all rights reserved by its author -- vendored for parser-benchmark purposes only, not redistribution)

tagoriginalround-tripped
h111
h22424
a263263
img22

Original (vendored fixture)

original tohoho_web_home

Round-tripped (html_to_markdown -> markdown_to_html)

round-tripped tohoho_web_home

converted.md · original.html · roundtrip.html