Migrating to turbohtml¶

turbohtml replaces the HTML libraries it benchmarks against. None is API-compatible, so porting is a translation: turbohtml uses one name per concept and a typed shape where those libraries spread the work across aliases, methods, and treebuilder choices. This page maps each library to turbohtml; BeautifulSoup gets the deepest treatment because it shares the most surface.

The guides are grouped by the turbohtml namespace that replaces each library – parse & DOM, detect, query, clean, convert, extract, build, serialize, the same order the Reference and How-to guides use – so a library that spans namespaces sits under the one it maps to most directly. Within each group they are ordered by adoption, most to least monthly PyPI downloads, from the pepy.tech badge on each row; where two libraries share a download tier the order follows that tier rather than a precise count. The all-time totals and each library’s documentation sit alongside for context.

Parse & DOM¶

These libraries parse HTML into a document tree. turbohtml.parse() builds the tree a browser would, so malformed input recovers the WHATWG way, and every node shares the navigation, query, and mutation surface in Nodes.

#	Library	Docs	Monthly downloads	Total downloads
1	beautifulsoup	docs
2	lxml	docs
3	html5lib	docs
4	selectolax	docs
5	resiliparse	docs
6	mechanicalsoup	docs
7	html5-parser	docs
8	stdlib	docs	Bundled with Python	–
9	parse5	docs	JavaScript (npm)	–
10	jsdom	docs	JavaScript (npm)	–
11	lol-html	docs	Rust / WASM	–

parse5 is the reference JavaScript WHATWG parser rather than a Python library; its guide is a cross-language reference for porting a parse5 pipeline – and its sourceCodeLocationInfo model in particular – onto turbohtml. jsdom is the reference JavaScript DOM built on parse5; its guide maps the DOM traversal (TreeWalker, NodeIterator) and range (Range, StaticRange) APIs onto turbohtml’s. lol-html is Cloudflare’s Rust streaming HTML rewriter; its guide maps its DOM-less HTMLRewriter and Element edit API onto turbohtml.rewrite.rewrite().

Detect¶

turbohtml.detect.detect() sniffs the character encoding of raw bytes with the same C pipeline turbohtml.parse() runs – the WHATWG sniff, then Firefox’s chardetng scoring – so it answers what these libraries answer, with the result a browser would pick.

#	Library	Docs	Monthly downloads	Total downloads
1	charset-normalizer	docs
2	chardet	docs

Query¶

These libraries select nodes with CSS or XPath, or read the CSS Object Model. turbohtml matches CSS selectors with turbohtml.Node.select(), evaluates XPath 1.0 with turbohtml.Node.xpath(), exposes a soupsieve-shaped matching surface in Query, and resolves the CSS cascade to a computed style in CSSOM.

#	Library	Docs
1	soupsieve	docs
2	parsel	docs
3	pyquery	docs
4	cssutils	docs

Clean¶

These libraries scrub, sanitize, or shrink markup. turbohtml.clean sanitizes against an allowlist, autolinks bare URLs, and minifies HTML with minify(), CSS with minify_css(), and JavaScript with minify_js() – every minifier value-safe.

#	Library	Docs
1	linkify-it-py	docs
2	bleach	docs
3	nh3	docs
4	lxml-html-clean	docs
5	rcssmin	docs
6	rjsmin	docs
7	jsmin	docs
8	minify-html	docs
9	htmlmin	docs
10	csscompressor	docs
11	html-sanitizer	docs
12	calmjs.parse	docs
13	html5validator	docs
14	lightningcss	docs
15	sanitize-html	docs
16	DOMPurify	docs

Convert¶

turbohtml.convert.css_to_xpath() translates a CSS selector into the equivalent XPath 1.0 expression, the job cssselect does for lxml, parsel, and pyquery, so a system that speaks only XPath can run a CSS selector.

#	Library	Docs	Monthly downloads	Total downloads
1	cssselect	docs

Extract¶

These libraries pull the article, its metadata, or embedded structured data out of a page. turbohtml.Node.main_content() isolates the article body, turbohtml.Element.records() reads a table into records, and turbohtml.Document.structured_data() collects JSON-LD, Microdata, OpenGraph, and RDFa.

#	Library	Docs
1	pandas	docs
2	w3lib	docs
3	trafilatura	docs
4	courlan	docs
5	extruct	docs
6	justext	docs
7	readability-lxml	docs
8	readabilipy	docs
9	metadata_parser	docs
10	newspaper3k	docs
11	boilerpy3	docs
12	goose3	docs
13	microdata	docs
14	news-please	docs
15	opengraph	docs
16	htmldate	docs
17	feedparser	docs

Build¶

These libraries construct HTML from Python. turbohtml.build.E builds a real element tree that queries, edits, and serializes like a parsed one.

#	Library	Docs
1	dominate	docs
2	yattag	docs
3	htbuilder	docs
4	htpy	docs
5	airium	docs
6	markyp	docs
7	fast-html	docs
8	simple-html	docs
9	hyperpython	docs

Serialize¶

These libraries render a tree back out – as escaped HTML, Markdown, or plain text. turbohtml.escape() and the turbohtml.migration.markupsafe drop-in match markupsafe byte for byte, turbohtml.Node.to_markdown() emits GitHub-Flavored Markdown, and turbohtml.Node.to_text() extracts rendered text.

#	Library	Docs
1	markupsafe	docs
2	markdownify	docs
3	html2text	docs
4	inscriptis	docs
5	html-text	docs