################################## Check a document for conformance ################################## .. currentmodule:: turbohtml.conformance The WHATWG parser recovers from bad markup and records its :class:`~turbohtml.ParseError`\\s, but many authoring mistakes leave a perfectly parseable tree: an image with no alt text, an obsolete ````, a duplicate id, an undefined ARIA role. :func:`check_html` parses a markup string and runs the HTML5 authoring rules over it, returning a :class:`ConformanceReport`. ****************** Read the verdict ****************** A report is truthy when the document is valid -- when nothing it found is an error. Warnings and info notes are advice and never change the verdict. Each :class:`ConformanceMessage` carries a stable ``code``, a ``severity``, a description, and the source ``line`` and ``column`` of the node that triggered it: .. testcode:: from turbohtml.conformance import check_html report = check_html( "

" ) print(report.valid) for message in report.messages: print(message.severity, message.code, "--", message.message) .. testoutput:: False error img-missing-alt -- img element has no alt attribute warning aria-redundant-role -- role navigation duplicates the element's implicit role warning missing-lang -- html element has no lang attribute error missing-title -- document has no non-empty title element ******************* Split by severity ******************* The report's :attr:`~ConformanceReport.errors`, :attr:`~ConformanceReport.warnings`, and :attr:`~ConformanceReport.infos` views partition the messages, so a gate can fail the build on errors while only logging the rest: .. testcode:: if not report: print("errors:", [message.code for message in report.errors]) print("advice:", [message.code for message in report.warnings + report.infos]) .. testoutput:: errors: ['img-missing-alt', 'missing-title'] advice: ['aria-redundant-role', 'missing-lang'] ****************************** Check an already-parsed tree ****************************** When you have already parsed the document, pass it to :func:`check` to skip a reparse. Handing :func:`check` a subtree runs only the per-element rules -- the document-level title and language rules apply to a whole :class:`~turbohtml.Document`: .. testcode:: from turbohtml import parse from turbohtml.conformance import check document = parse("Docs

Hi

") print(check(document).valid) fragment = parse("

").find("section") print([message.code for message in check(fragment).messages]) .. testoutput:: True ['section-no-heading', 'img-missing-alt']