rapier-markdown-kit 1.1.78 → 1.1.80

package/dist/kit/render-sanitize.mjs

@@ -1,52 +1,263 @@
 // SPDX-License-Identifier: MIT
 const RAPIER_SANITIZE_FORBID_TAGS = Object.freeze(['form', 'style', 'dialog', 'template', 'iframe']);
 const RAPIER_SANITIZE_FORBID_ATTR = Object.freeze(['action', 'formaction', 'form', 'for']);
+// The document profiles use one allowlist in a browser and in an inert server parser.
+const RAPIER_SANITIZE_PROFILES = Object.freeze({
+	htmlTags: Object.freeze(["a","abbr","acronym","address","area","article","aside","audio","b","bdi","bdo","big","blink","blockquote","body","br","button","canvas","caption","center","cite","code","col","colgroup","content","data","datalist","dd","decorator","del","details","dfn","dialog","dir","div","dl","dt","element","em","fieldset","figcaption","figure","font","footer","form","h1","h2","h3","h4","h5","h6","head","header","hgroup","hr","html","i","img","input","ins","kbd","label","legend","li","main","map","mark","marquee","menu","menuitem","meter","nav","nobr","ol","optgroup","option","output","p","picture","pre","progress","q","rp","rt","ruby","s","samp","search","section","select","shadow","slot","small","source","spacer","span","strike","strong","style","sub","summary","sup","table","tbody","td","template","textarea","tfoot","th","thead","time","tr","track","tt","u","ul","var","video","wbr"]),
+	svgTags: Object.freeze(["svg","a","altglyph","altglyphdef","altglyphitem","animatecolor","animatemotion","animatetransform","circle","clippath","defs","desc","ellipse","enterkeyhint","exportparts","filter","font","g","glyph","glyphref","hkern","image","inputmode","line","lineargradient","marker","mask","metadata","mpath","part","path","pattern","polygon","polyline","radialgradient","rect","stop","style","switch","symbol","text","textpath","title","tref","tspan","view","vkern"]),
+	mathTags: Object.freeze(["math","menclose","merror","mfenced","mfrac","mglyph","mi","mlabeledtr","mmultiscripts","mn","mo","mover","mpadded","mphantom","mroot","mrow","ms","mspace","msqrt","mstyle","msub","msup","msubsup","mtable","mtd","mtext","mtr","munder","munderover","mprescripts"]),
+	htmlAttributes: Object.freeze(["accept","action","align","alt","autocapitalize","autocomplete","autopictureinpicture","autoplay","background","bgcolor","border","capture","cellpadding","cellspacing","checked","cite","class","clear","color","cols","colspan","command","commandfor","controls","controlslist","coords","crossorigin","datetime","decoding","default","dir","disabled","disablepictureinpicture","disableremoteplayback","download","draggable","enctype","enterkeyhint","exportparts","face","for","headers","height","hidden","high","href","hreflang","id","inert","inputmode","integrity","ismap","kind","label","lang","list","loading","loop","low","max","maxlength","media","method","min","minlength","multiple","muted","name","nonce","noshade","novalidate","nowrap","open","optimum","part","pattern","placeholder","playsinline","popover","popovertarget","popovertargetaction","poster","preload","pubdate","radiogroup","readonly","rel","required","rev","reversed","role","rows","rowspan","spellcheck","scope","selected","shape","size","sizes","slot","span","srclang","start","src","srcset","step","style","summary","tabindex","title","translate","type","usemap","valign","value","width","wrap","xmlns"]),
+	svgAttributes: Object.freeze(["accent-height","accumulate","additive","alignment-baseline","amplitude","ascent","attributename","attributetype","azimuth","basefrequency","baseline-shift","begin","bias","by","class","clip","clippathunits","clip-path","clip-rule","color","color-interpolation","color-interpolation-filters","color-profile","color-rendering","cx","cy","d","dx","dy","diffuseconstant","direction","display","divisor","dominant-baseline","dur","edgemode","elevation","end","exponent","fill","fill-opacity","fill-rule","filter","filterunits","flood-color","flood-opacity","font-family","font-size","font-size-adjust","font-stretch","font-style","font-variant","font-weight","fx","fy","g1","g2","glyph-name","glyphref","gradientunits","gradienttransform","height","href","id","image-rendering","in","in2","intercept","k","k1","k2","k3","k4","kerning","keypoints","keysplines","keytimes","lang","lengthadjust","letter-spacing","kernelmatrix","kernelunitlength","lighting-color","local","marker-end","marker-mid","marker-start","markerheight","markerunits","markerwidth","maskcontentunits","maskunits","max","mask","mask-type","media","method","mode","min","name","numoctaves","offset","operator","opacity","order","orient","orientation","origin","overflow","paint-order","path","pathlength","patterncontentunits","patterntransform","patternunits","pointer-events","points","preservealpha","preserveaspectratio","primitiveunits","r","rx","ry","radius","refx","refy","repeatcount","repeatdur","restart","result","rotate","scale","seed","shape-rendering","slope","specularconstant","specularexponent","spreadmethod","startoffset","stddeviation","stitchtiles","stop-color","stop-opacity","stroke-dasharray","stroke-dashoffset","stroke-linecap","stroke-linejoin","stroke-miterlimit","stroke-opacity","stroke","stroke-width","style","surfacescale","systemlanguage","tabindex","tablevalues","targetx","targety","transform","transform-origin","text-anchor","text-decoration","text-orientation","text-rendering","textlength","type","u1","u2","unicode","values","vector-effect","viewbox","visibility","version","vert-adv-y","vert-origin-x","vert-origin-y","width","word-spacing","wrap","writing-mode","xchannelselector","ychannelselector","x","x1","x2","xmlns","y","y1","y2","z","zoomandpan"]),
+	mathAttributes: Object.freeze(["accent","accentunder","align","bevelled","close","columnalign","columnlines","columnspacing","columnspan","denomalign","depth","dir","display","displaystyle","encoding","fence","frame","height","href","id","largeop","length","linethickness","lquote","lspace","mathbackground","mathcolor","mathsize","mathvariant","maxsize","minsize","movablelimits","notation","numalign","open","rowalign","rowlines","rowspacing","rowspan","rspace","rquote","scriptlevel","scriptminsize","scriptsizemultiplier","selection","separator","separators","stretchy","subscriptshift","supscriptshift","symmetric","voffset","width","xmlns"]),
+	xmlAttributes: Object.freeze(["xlink:href","xml:id","xlink:title","xml:space","xmlns:xlink"]),
+});
+const RAPIER_SANITIZE_DROP_CONTENTS = Object.freeze('annotation-xml.audio.colgroup.desc.foreignobject.head.iframe.math.mi.mn.mo.ms.mtext.noembed.noframes.noscript.plaintext.script.selectedcontent.style.svg.template.thead.title.video.xmp'.split('.'));
+function _rapierSanitizeProfile(...kinds) {
+	return {ALLOWED_TAGS: Object.freeze(['#text', ...new Set(kinds.flatMap(kind => RAPIER_SANITIZE_PROFILES[kind + 'Tags']))]),
+		ALLOWED_ATTR: Object.freeze([...new Set(kinds.flatMap(kind => RAPIER_SANITIZE_PROFILES[kind + 'Attributes']).concat(kinds.some(kind => kind !== 'html') ? RAPIER_SANITIZE_PROFILES.xmlAttributes : []))]),
+		FORBID_CONTENTS: RAPIER_SANITIZE_DROP_CONTENTS};
+}
 const RAPIER_SANITIZE_OPTIONS = Object.freeze({
 	render: Object.freeze({
-		USE_PROFILES: { html: true, svg: true, mathMl: true },
+		..._rapierSanitizeProfile('html', 'svg', 'math'),
 		FORBID_TAGS: RAPIER_SANITIZE_FORBID_TAGS,
 		FORBID_ATTR: RAPIER_SANITIZE_FORBID_ATTR,
 	}),
 	paste: Object.freeze({
-		USE_PROFILES: { html: true },
+		..._rapierSanitizeProfile('html'),
 		FORBID_TAGS: RAPIER_SANITIZE_FORBID_TAGS,
 		FORBID_ATTR: RAPIER_SANITIZE_FORBID_ATTR,
 	}),
 	snippet: Object.freeze({
-		USE_PROFILES: { html: true },
+		..._rapierSanitizeProfile('html'),
 		FORBID_TAGS: RAPIER_SANITIZE_FORBID_TAGS,
 		FORBID_ATTR: Object.freeze([...'style']),
 	}),
 	diagram: Object.freeze({
-		USE_PROFILES: { html: true, svg: true },
+		..._rapierSanitizeProfile('html', 'svg'),
 		FORBID_TAGS: Object.freeze(['form', 'dialog', 'template', 'iframe', 'script', 'foreignobject']),
 		FORBID_ATTR: RAPIER_SANITIZE_FORBID_ATTR,
 	}),
 });
 
-// Snapshot current math text before a Markdown writer collapses whitespace. The parsed copy
-// stays inert; the writer still admits raw source through its shared math grammar.
+const DIAGRAM_LABEL_HTML = new Set('div span p br strong em b i s u del code sub sup'.split(' '));
+const DIAGRAM_LABEL_MATH = new Set(('math mrow mi mn mo mtext mspace ms msub msup msubsup mfrac msqrt mroot mstyle merror '
+	+ 'mpadded mphantom mfenced menclose munder mover munderover mtable mtr mtd mlabeledtr mmultiscripts mprescripts none semantics').split(' '));
+const DIAGRAM_LABEL_MATH_ATTRIBUTES = new Set(('display mathvariant mathsize mathcolor mathbackground dir accent accentunder '
+	+ 'columnalign columnlines columnspacing columnspan columnwidth depth displaystyle equalcolumns equalrows fence frame framespacing '
+	+ 'height largeop linebreak linethickness lspace maxsize minsize movablelimits notation rowalign rowlines rowspacing rowspan rspace '
+	+ 'scriptlevel scriptminsize scriptsizemultiplier separator stretchy subscriptshift superscriptshift valign voffset width').split(' '));
+const DIAGRAM_LABEL_CLASSES = new Set('nodeLabel edgeLabel labelBkg markdown-node-label katex messageText'.split(' '));
+const DIAGRAM_LABEL_CSS = new Set(('color fill stroke background-color font-family font-size font-style font-weight font-variant '
+	+ 'font-variant-ligatures font-feature-settings line-height letter-spacing word-spacing text-align vertical-align white-space '
+	+ 'word-break overflow-wrap text-wrap text-transform text-decoration text-decoration-line text-decoration-style text-decoration-color '
+	+ 'text-decoration-thickness text-underline-offset align-items align-content justify-content justify-items flex-direction flex-wrap '
+	+ 'width height min-width min-height max-width max-height margin margin-top margin-right margin-bottom margin-left '
+	+ 'padding padding-top padding-right padding-bottom padding-left').split(' '));
+const DIAGRAM_LABEL_SVG_NAMESPACE = 'http://www.w3.org/2000/svg';
+const DIAGRAM_LABEL_HTML_NAMESPACE = 'http://www.w3.org/1999/xhtml';
+const DIAGRAM_LABEL_MATH_NAMESPACE = 'http://www.w3.org/1998/Math/MathML';
+
+// Snapshot current math text before a Markdown writer collapses whitespace. Rendered SVG
+// glyphs have no textContent: put their retained source in this inert copy so the writer's
+// blank-node rule cannot discard an equation or its containing paragraph/list item.
 function prepareMathSource(input, DOMParser) {
 	const html = typeof input === 'string';
-	const MARK = 'span[data-rapier-math-source]';
-	if (html ? !/data-rapier-math-source/i.test(input) : typeof input?.cloneNode !== 'function' ||
+	const MARK = 'span[data-rapier-math-source],span.math-rendered[data-math-src]';
+	if (html ? !/data-(?:rapier-math-source|math-src)/i.test(input) : typeof input?.cloneNode !== 'function' ||
 		!(input.matches?.(MARK) || input.querySelector?.(MARK))) return input;
 	const root = html ? new DOMParser().parseFromString(
 		'<x-turndown id="turndown-root">' + input + '</x-turndown>', 'text/html').getElementById('turndown-root') : input.cloneNode(true);
 	const nodes = Array.from(root.querySelectorAll(MARK));
-	const sourceRoot = root.nodeName === 'SPAN' && root.hasAttribute('data-rapier-math-source');
+	const sourceRoot = root.nodeName === 'SPAN' && root.matches(MARK);
 	if (sourceRoot) nodes.push(root);
 	for (const node of nodes) {
+		if (node.classList.contains('math-rendered') && node.hasAttribute('data-math-src')) {
+			try { node.textContent = decodeURIComponent(node.getAttribute('data-math-src')); } catch (_) {}
+			continue;
+		}
 		try { node.setAttribute('data-rapier-math-source', encodeURIComponent(node.textContent || '')); }
 		catch (_) { node.removeAttribute('data-rapier-math-source'); }
 	}
 	return sourceRoot ? root.outerHTML : root;
 }
 
+// An inert parser and a string sanitizer use the same profile as DOMPurify. Neither port may
+// execute document scripts or fetch resources. HTML5 parsing precedes filtering so foreign
+// content, entities and malformed table markup cross the same boundary as a browser's parser.
+function createInertHtmlSanitizer({parseFragment, serialize, sanitizeHtml, safeRasterDataUrl}) {
+	if (![parseFragment, serialize, sanitizeHtml, safeRasterDataUrl].every(port => typeof port === 'function'))
+		throw new TypeError('The inert HTML sanitizer requires every parser and filtering port');
+	const uri = /^(?:(?:(?:f|ht)tps?|mailto|tel|callto|sms|cid|xmpp|matrix):|[^a-z]|[a-z+.\-]+(?:[^a-z+.\-:]|$))/i;
+	const uriSafe = new Set(['alt', 'class', 'for', 'id', 'label', 'name', 'pattern', 'placeholder', 'role', 'summary', 'title', 'value', 'style', 'xmlns']);
+	const schemes = ['http', 'https', 'ftp', 'ftps', 'mailto', 'tel', 'callto', 'sms', 'cid', 'xmpp', 'matrix'];
+	const htmlNamespace = 'http://www.w3.org/1999/xhtml', svgNamespace = 'http://www.w3.org/2000/svg', mathNamespace = 'http://www.w3.org/1998/Math/MathML';
+	const svgTags = new Set(RAPIER_SANITIZE_PROFILES.svgTags), mathTags = new Set(RAPIER_SANITIZE_PROFILES.mathTags);
+	const mathIntegration = new Set(['mi', 'mo', 'mn', 'ms', 'mtext']), htmlIntegration = new Set(['annotation-xml']);
+	const sharedHtmlSvg = new Set(['title', 'style', 'font', 'a', 'script']);
+	// DOMPurify's default namespace admission applies in addition to its tag profile. In
+	// particular a MathML name parsed as HTML, or HTML inside SVG desc, is not admitted.
+	const validNamespace = (node, parent) => {
+		const tag = node.tagName.toLowerCase(), parentTag = (parent.tagName || 'template').toLowerCase();
+		const namespace = node.namespaceURI, parentNamespace = parent.namespaceURI || htmlNamespace;
+		if (namespace === svgNamespace) return parentNamespace === htmlNamespace ? tag === 'svg'
+			: parentNamespace === mathNamespace ? tag === 'svg' && (parentTag === 'annotation-xml' || mathIntegration.has(parentTag)) : svgTags.has(tag);
+		if (namespace === mathNamespace) return parentNamespace === htmlNamespace ? tag === 'math'
+			: parentNamespace === svgNamespace ? tag === 'math' && htmlIntegration.has(parentTag) : mathTags.has(tag);
+		return namespace === htmlNamespace && !(parentNamespace === svgNamespace && !htmlIntegration.has(parentTag))
+			&& !(parentNamespace === mathNamespace && !mathIntegration.has(parentTag))
+			&& !mathTags.has(tag) && (sharedHtmlSvg.has(tag) || !svgTags.has(tag));
+	};
+	return function sanitize(value, context = 'render') {
+		const selected = context === 'raw' || context === 'export' || context === 'source' || RAPIER_SANITIZE_OPTIONS[context] ? context : 'render';
+		const options = RAPIER_SANITIZE_OPTIONS[selected] || RAPIER_SANITIZE_OPTIONS.render;
+		const forbiddenTags = new Set(options.FORBID_TAGS || []), forbiddenAttributes = new Set(options.FORBID_ATTR || []);
+		const tags = new Set(options.ALLOWED_TAGS.filter(tag => tag !== '#text' && !forbiddenTags.has(tag)));
+		const attributes = new Set(options.ALLOWED_ATTR);
+		const tree = parseFragment(String(value == null ? '' : value));
+		const dropContents = new Set(RAPIER_SANITIZE_DROP_CONTENTS);
+		const admitTree = parent => {
+			for (let index = 0; index < (parent.childNodes || []).length;) {
+				const node = parent.childNodes[index], tag = node.tagName?.toLowerCase();
+				if (tag && !tags.has(tag)) {
+					const children = dropContents.has(tag) ? [] : node.childNodes || [];
+					for (const child of children) child.parentNode = parent;
+					parent.childNodes.splice(index, 1, ...children);
+					continue;
+				}
+				if (tag && !validNamespace(node, parent) || node.nodeName === '#comment') {
+					parent.childNodes.splice(index, 1); continue;
+				}
+				admitTree(node); index++;
+			}
+		};
+		admitTree(tree);
+		const admitNames = node => {
+			if (node.tagName && tags.has(node.tagName.toLowerCase())) tags.add(node.tagName);
+			for (const child of node.childNodes || []) admitNames(child);
+		};
+		admitNames(tree);
+		const cleaned = sanitizeHtml(serialize(tree), {
+			allowedTags: [...tags], allowedAttributes: false,
+			disallowedTagsMode: 'discard', nonTextTags: RAPIER_SANITIZE_DROP_CONTENTS,
+			allowedSchemes: schemes, allowedSchemesByTag: {img: [...schemes, 'data']},
+			allowedSchemesAppliedToAttributes: ['href', 'src', 'xlink:href', 'cite', 'background', 'poster'],
+			parseStyleAttributes: false,
+			parser: {lowerCaseTags: false, lowerCaseAttributeNames: false, decodeEntities: true},
+			transformTags: {'*': (tagName, input) => {
+				const attribs = Object.create(null), tag = tagName.toLowerCase();
+				for (const [name, inputValue] of Object.entries(input)) {
+					const key = name.toLowerCase();
+					const value = name === 'value' ? inputValue : inputValue.trim();
+					if (forbiddenAttributes.has(key) || key.startsWith('on')) continue;
+					if (/((--!?|])>)|<\/(style|script|title|xmp|textarea|noscript|iframe|noembed|noframes)/i.test(value)
+						|| key === 'attributename' && value.includes('href')) continue;
+					const data = /^data-[\-\w.\u00b7-\uffff]+$/.test(key), aria = /^aria-[\-\w]+$/.test(key);
+					const pageBreak = key === 'contenteditable' && value === 'false' && tag === 'div' && input['data-md-break'] === 'page';
+					if (!(data && selected !== 'raw') && !aria && !pageBreak && !attributes.has(key)) continue;
+					const compact = value.replace(/[\u0000-\u0020\u00a0\u1680\u180e\u2000-\u2029\u205f\u3000]/g, '');
+					if (!data && !aria && !pageBreak && !uriSafe.has(key) && compact) {
+						if (/^data:/i.test(compact)) {
+							if (key !== 'src' || tag !== 'img' || !safeRasterDataUrl(compact)) continue;
+						} else if (!uri.test(compact)) continue;
+					}
+					attribs[name] = value;
+				}
+				if ((tag === 'a' || tag === 'area') && attribs.target?.trim().toLowerCase() === '_blank') attribs.rel = 'noopener noreferrer';
+				return {tagName, attribs};
+			}},
+		});
+		return serialize(parseFragment(cleaned));
+	};
+}
+
 // The document's renderer. DOM, codecs and host state are explicit inputs.
 function createRenderSanitizer(runtime) {
-  const {CSSStyleSheet, DOMPurify, RAPIER_RASTER_DATA_URL_RE, URL, _rapierCssDeclaration, _rapierDropRemoteDeclarations, _rapierRemoteContent, _rapierRuleDescriptorIsRemote, _rapierSanitizeRuntime, _rapierVerifyRasterBytes, document, globalThis, location} = runtime;
+  const {CSSStyleSheet, DOMPurify, RAPIER_RASTER_DATA_URL_RE, URL, _rapierCssDeclaration, _rapierDropRemoteDeclarations, _rapierRemoteContent, _rapierRuleDescriptorIsRemote, _rapierSanitizeRuntime, document, globalThis, location} = runtime;
 
+// Only the diagram reader and its styled export admit static foreignObject labels. The
+// ordinary SVG reader still rejects HTML subdocuments. The export class selects scope;
+// every descendant is checked again, so a copied class never confers authority.
+function diagramLabelRoot(node) {
+	for (let current = node; current?.nodeType === 1; current = current.parentElement) {
+		if (current.localName?.toLowerCase() === 'foreignobject') return current;
+	}
+	return null;
+}
+
+function diagramLabelStyle(property, value) {
+	property = String(property).toLowerCase();
+	value = String(value).trim();
+	if (property === 'display') return /^(?:none|block|inline|inline-block|table|table-cell|flex|inline-flex)$/.test(value);
+	if (/^overflow(?:-[xy])?$/.test(property)) return /^(?:hidden|clip)$/.test(value);
+	if (property === 'contain') return value === 'content';
+	if (property === 'isolation') return value === 'isolate';
+	if (!DIAGRAM_LABEL_CSS.has(property)) return false;
+	const declaration = document.createElement('span').style;
+	declaration.setProperty(property, value);
+	const parsed = declaration.getPropertyValue(property);
+	if (!parsed || /(?:url|image-set|element|paint|attr)\s*\(/i.test(parsed)) return false;
+	// SVG paint accepts resource references; a custom property must not smuggle one in.
+	return !/^(?:fill|stroke)$/.test(property) || !/var\s*\(/i.test(parsed);
+}
+
+function diagramLabelElement(node) {
+	if (node.nodeType !== 1) return;
+	const label = diagramLabelRoot(node);
+	if (!label) return;
+	if (node === label) {
+		const svg = node.parentElement?.closest('svg');
+		if (node.namespaceURI !== DIAGRAM_LABEL_SVG_NAMESPACE || !svg || diagramLabelRoot(node.parentElement) ||
+				(_rapierSanitizeRuntime.context !== 'diagram' && !svg.classList.contains('rapier-diagram'))) node.remove();
+		return;
+	}
+	const name = node.localName?.toLowerCase();
+	if (!(node.namespaceURI === DIAGRAM_LABEL_HTML_NAMESPACE && DIAGRAM_LABEL_HTML.has(name)) &&
+			!(node.namespaceURI === DIAGRAM_LABEL_MATH_NAMESPACE && DIAGRAM_LABEL_MATH.has(name))) node.remove();
+}
+
+function diagramLabelAttribute(node, data) {
+	const label = diagramLabelRoot(node);
+	if (!label) return;
+	data.forceKeepAttr = false;
+	const name = String(data.attrName).toLowerCase();
+	const value = String(data.attrValue);
+	if (name === 'style') {
+		const declaration = document.createElement('span').style;
+		declaration.cssText = value;
+		for (const property of Array.from(declaration)) {
+			if (!diagramLabelStyle(property, declaration.getPropertyValue(property))) declaration.removeProperty(property);
+		}
+		data.attrValue = declaration.cssText;
+		data.keepAttr = !!data.attrValue;
+		return;
+	}
+	if (name === 'class') {
+		data.attrValue = value.split(/\s+/).filter(token => DIAGRAM_LABEL_CLASSES.has(token)).join(' ');
+		data.keepAttr = !!data.attrValue;
+		return;
+	}
+	if (name === 'xmlns') {
+		data.keepAttr = value === node.namespaceURI;
+		return;
+	}
+	if (node === label) {
+		data.keepAttr = name === 'transform' || /^(?:x|y|width|height)$/.test(name) &&
+			/^[-+]?(?:\d+(?:\.\d*)?|\.\d+)(?:e[-+]?\d+)?(?:px|%)?$/i.test(value) && Number.isFinite(parseFloat(value)) &&
+			(!/^(?:width|height)$/.test(name) || parseFloat(value) >= 0);
+		return;
+	}
+	data.keepAttr = node.namespaceURI === DIAGRAM_LABEL_MATH_NAMESPACE
+		? DIAGRAM_LABEL_MATH_ATTRIBUTES.has(name)
+		: name === 'dir' && /^(?:ltr|rtl|auto)$/.test(value);
+}
+
+function diagramLabelContained(node) {
+	const label = diagramLabelRoot(node);
+	if (!label) return;
+	if (node === label) node.style.setProperty('overflow', 'hidden', 'important');
+	else if (node.parentElement === label) node.style.setProperty('contain', 'content', 'important');
+}
 
 function _rapierInstallSanitizeHooks() {
 	DOMPurify.addHook('uponSanitizeAttribute', (node, data) => {
@@ -148,11 +359,27 @@ function sanitizeRapierHtml(value, context = 'render') {
 	// (the hook) without an id the chrome owns. 'source' is Markdown source HTML, written and never shown: the render
 	// profile's admission, and the block on remote content stays the reader's (a remote picture's URL is the person's words).
 	const selected = context === 'raw' || context === 'export' || context === 'source' || RAPIER_SANITIZE_OPTIONS[context] ? context : 'render';
-	const options = selected === 'raw' ? { ...RAPIER_SANITIZE_OPTIONS.render, ALLOW_DATA_ATTR: false } : RAPIER_SANITIZE_OPTIONS[selected] || RAPIER_SANITIZE_OPTIONS.render;
+	let options = selected === 'raw' ? { ...RAPIER_SANITIZE_OPTIONS.render, ALLOW_DATA_ATTR: false } : RAPIER_SANITIZE_OPTIONS[selected] || RAPIER_SANITIZE_OPTIONS.render;
+	const labels = selected === 'diagram' || selected === 'export';
+	if (labels) {
+		// DOMPurify's USE_PROFILES replaces explicit allowlists. Add MathML to the
+		// selected profile without discarding its admitted HTML and SVG vocabulary.
+		options = {...options, ALLOWED_TAGS: [...new Set([...options.ALLOWED_TAGS, ...RAPIER_SANITIZE_PROFILES.mathTags])],
+			ALLOWED_ATTR: [...new Set([...options.ALLOWED_ATTR, ...RAPIER_SANITIZE_PROFILES.mathAttributes])], ADD_TAGS: ['foreignobject'],
+			HTML_INTEGRATION_POINTS: {foreignobject: true}, FORBID_TAGS: options.FORBID_TAGS.filter(tag => tag !== 'foreignobject')};
+		DOMPurify.addHook('uponSanitizeElement', diagramLabelElement);
+		DOMPurify.addHook('uponSanitizeAttribute', diagramLabelAttribute);
+		DOMPurify.addHook('afterSanitizeAttributes', diagramLabelContained);
+	}
 	_rapierSanitizeRuntime.context = selected;
 	try {
 		return DOMPurify.sanitize(value == null ? '' : String(value), options);
 	} finally {
+		if (labels) {
+			DOMPurify.removeHook('uponSanitizeElement', diagramLabelElement);
+			DOMPurify.removeHook('uponSanitizeAttribute', diagramLabelAttribute);
+			DOMPurify.removeHook('afterSanitizeAttributes', diagramLabelContained);
+		}
 		_rapierSanitizeRuntime.context = 'render';
 	}
 }
@@ -237,6 +464,18 @@ function _rapierDropRemoteRules(rules, owner) {
 	}
 }
 
-  return {_rapierInstallSanitizeHooks, sanitizeRapierHtml, escapeRapierHtmlText, _rapierChromeOwnsId, _rapierSafeRasterDataUrl, _rapierCssPresentationIsRemote, _rapierRemoteSubresourceOrigin, _rapierStyleWithoutRemoteUrls, _rapierStylesheetWithoutRemoteUrls, _rapierDropRemoteRules};
+function _rapierVerifyRasterBytes(bytes, mime) {
+	if (mime === 'image/jxl') return globalThis.RapierImageAssets.isJxl(bytes);
+
+	// declaration, doctype or generator/license comment before <svg itself, of whatever length its
+
+	if (mime === 'image/svg+xml') return /<svg[\s>]/i.test(new TextDecoder('utf-8', {fatal: false}).decode(bytes));
+	if (mime === 'image/png') return bytes.length >= 8 && bytes[0] === 0x89 && bytes[1] === 0x50 && bytes[2] === 0x4e && bytes[3] === 0x47 && bytes[4] === 0x0d && bytes[5] === 0x0a && bytes[6] === 0x1a && bytes[7] === 0x0a;
+	if (mime === 'image/jpeg') return bytes.length >= 3 && bytes[0] === 0xff && bytes[1] === 0xd8 && bytes[2] === 0xff;
+	if (mime === 'image/webp') return bytes.length >= 12 && String.fromCharCode(...bytes.slice(0, 4)) === 'RIFF' && String.fromCharCode(...bytes.slice(8, 12)) === 'WEBP';
+	return false;
+}
+
+  return {_rapierVerifyRasterBytes, _rapierInstallSanitizeHooks, sanitizeRapierHtml, diagramLabelStyle, escapeRapierHtmlText, _rapierChromeOwnsId, _rapierSafeRasterDataUrl, _rapierCssPresentationIsRemote, _rapierRemoteSubresourceOrigin, _rapierStyleWithoutRemoteUrls, _rapierStylesheetWithoutRemoteUrls, _rapierDropRemoteRules};
 }
-export {RAPIER_SANITIZE_FORBID_TAGS as forbidTags, RAPIER_SANITIZE_FORBID_ATTR as forbidAttributes, createRenderSanitizer, RAPIER_SANITIZE_OPTIONS as options, prepareMathSource};
+export {RAPIER_SANITIZE_FORBID_TAGS as forbidTags, RAPIER_SANITIZE_FORBID_ATTR as forbidAttributes, createRenderSanitizer, createInertHtmlSanitizer, RAPIER_SANITIZE_OPTIONS as options, prepareMathSource};

package/dist/kit/render-styles.mjs

@@ -2,12 +2,12 @@
 // The document's renderer. DOM, codecs and host state are explicit inputs.
 function createRenderStyles(runtime) {
   const {_rapierArtifactAccent, _rapierStyleText} = runtime;
-function _rapierArtifactThemeCss() {
+function _rapierArtifactThemeCss(houseDefaults = false) {
 	// Geometry and the selected accent belong to this page; palette and type belong to the
 	// MIT reference sheet. Only the optional code lexer still reads the app token names.
-	const accent = _rapierArtifactAccent();
+	const accent = houseDefaults ? '' : ':root,.md-render{--md-color-accent:' + _rapierArtifactAccent() + '}';
 	return `
-:root,.md-render{--md-color-accent:${accent}}
+${accent}
 html[data-rapier-theme="dark"]{color-scheme:dark}
 html[data-rapier-theme="light"]{color-scheme:light}
 html[data-rapier-theme="system"]{color-scheme:light dark}
@@ -71,11 +71,11 @@ function _rapierDocumentPrintCss(hostSelector) {
 }`;
 }
 
-function _rapierArtifactStyles(theme, includeFonts, printMode) {
+function _rapierArtifactStyles(theme, includeFonts, printMode, houseDefaults = false) {
 	const parts = [];
 	if (includeFonts) parts.push(_rapierStyleText('rapier-font-style'));
 	parts.push(_rapierStyleText('rapier-content-style'));
-	parts.push(_rapierArtifactThemeCss());
+	parts.push(_rapierArtifactThemeCss(houseDefaults));
 	// Where the faces do not ride (the written page's policy carries no font-src), the reader's own sans is
 	// matched to Geist's metrics: one family per system face, src:local() only (fetched by nothing, so the
 	// policy is untouched), size-adjust from Geist's weighted average advance over the face's own (Geist 467

package/dist/kit/render.mjs

@@ -32,7 +32,7 @@ async function _rapierBuildArtifact(options, providedContext) {
 
 	// The semantic export root is detached and has never entered the live visibility queue.
 	// Await the same materialization owner before any projection clones or serializes it.
-	for (const diagram of context.semanticRoot.querySelectorAll('.diagram-block[data-diagram-src]')) {
+	for (const diagram of opts.kind === 'page' ? [] : context.semanticRoot.querySelectorAll('.diagram-block[data-diagram-src]')) {
 		await _rapierFillDiagram(diagram, true);
 	}
 
@@ -41,21 +41,23 @@ async function _rapierBuildArtifact(options, providedContext) {
 	const styledRoot = _rapierProjectStyledRoot(context.semanticRoot, {
 		metadata,
 		print: !!opts.print,
+		sourceCode: opts.kind === 'page',
 		baseName: base,
 	});
 	// _rapierProjectStyledRoot already returns the sanitized projection.
 	// Print never emits the layout reflow script (see below), so a print artifact has no use for
 	// this annotation -- skip it there rather than pay the cost for numbers nothing will read.
-	if (!opts.print) _rapierAnnotateExportBoxPolygons(styledRoot);
+	if (!opts.print && opts.kind !== 'page') _rapierAnnotateExportBoxPolygons(styledRoot);
 	// Share adds its nearest-side no-script float and long-code cap here. Offline-image
 	// admission above and the page policy below belong to this common writer, not its callers.
 	if (typeof opts.afterRoot === 'function') opts.afterRoot(styledRoot);
 	const carrier = opts.kind === 'standalone' ? await _rapierSharedSourceCarrier(context, styledRoot) : '';
-	const bodyHtml = styledRoot.innerHTML;
+	const bodyHtml = opts.kind === 'page' ? _rapierPageHtml(styledRoot) : styledRoot.innerHTML;
 	const theme = opts.print
 		? 'light'
+		: opts.kind === 'page' ? 'system'
 		: _rapierArtifactPreference('theme', ['system', 'light', 'dark']);
-	const highlights = _rapierArtifactPreference('highlights', ['accent', 'standard']);
+	const highlights = opts.kind === 'page' ? 'standard' : _rapierArtifactPreference('highlights', ['accent', 'standard']);
 	const versionMeta = document.querySelector('meta[name="rapier-version"]');
 	const version = versionMeta && versionMeta.content ? versionMeta.content.trim() : '0.0.0';
 	const bodyClass = theme === 'light' ? 'light' : '';
@@ -79,7 +81,7 @@ async function _rapierBuildArtifact(options, providedContext) {
 	const settings = _rapierDocumentSettingsOf(context.canonical);
 	const settingsCss = _rapierDocumentSettingsCss(settings);
 	const pageTitle = settings && settings.title ? globalThis.RapierMarkdownSpec.documentTitle(settings) : metadata.filename;
-	const css = _rapierArtifactStyles(theme, !!opts.print, !!opts.print || opts.kind === 'standalone')
+	const css = _rapierArtifactStyles(theme, !!opts.print, !!opts.print || opts.kind === 'standalone', opts.kind === 'page')
 		+ (opts.extraCss ? '\n\n' + opts.extraCss : '')
 		+ (willFont ? '\n\n@font-face{font-family:' + willFont.family + ';src:url(' + await _rapierBlobDataUrl(new Blob([willFont.bytes], { type: 'font/ttf' })) + ') format("truetype")}' : '')
 		+ (settingsCss ? '\n' + settingsCss : '');
@@ -90,11 +92,11 @@ async function _rapierBuildArtifact(options, providedContext) {
 	// The reflow script is written first, because whether it exists decides the policy. A page that carries no script says so: script-src 'none' refuses every
 	// script outright. Declaring a nonce a page has no use for is a widening, small but real -- anything that could inject markup into the file could read the
 	// nonce out of it and be admitted.
-	const layoutScript = !opts.print ? _rapierArtifactLayoutScript(styledRoot, nonce) : '';
+	const layoutScript = !opts.print && opts.kind !== 'page' ? _rapierArtifactLayoutScript(styledRoot, nonce) : '';
 	// The lexer for the page's code, under the same nonce, only when a block earned it: the CPU
 	// spans are the first paint, and where the reader's browser has WebGPU the lexer repaints them.
-	const lexerScript = !opts.print ? _rapierArtifactLexerScript(styledRoot, nonce) : '';
-	const inkScript = !opts.print ? _rapierArtifactInkScript(styledRoot, nonce) : '';
+	const lexerScript = !opts.print && opts.kind !== 'page' ? _rapierArtifactLexerScript(styledRoot, nonce) : '';
+	const inkScript = !opts.print && opts.kind !== 'page' ? _rapierArtifactInkScript(styledRoot, nonce) : '';
 	const page = '<!DOCTYPE html>\n'
 		+ '<html lang="en" data-rapier-theme="' + theme + '" data-highlights="' + highlights + '">\n'
 		+ '<head>\n<meta charset="UTF-8">\n' + (carrier ? RapierPageReturnAddress.PAGE_SEED + '\n' : '')
@@ -116,12 +118,31 @@ async function _rapierBuildArtifact(options, providedContext) {
 		html: page,
 		filename: base + '.html',
 		bodyHtml,
+		css,
 		docClass,
 		metadata,
 		documentTitle: settings && settings.title ? pageTitle : '',
 	};
 }
 
+// Page files have one byte representation across native and inert DOM hosts. Attribute
+// insertion order and a host's optional entity escapes are not document content.
+function _rapierPageHtml(root) {
+	const voidTags = new Set(['area', 'base', 'br', 'col', 'embed', 'hr', 'img', 'input', 'link', 'meta', 'param', 'source', 'track', 'wbr']);
+	const children = node => Array.from(node.childNodes || [], write).join('');
+	function write(node) {
+		if (node.nodeType === 3) return escapeRapierHtmlText(node.nodeValue || '');
+		if (node.nodeType !== 1) return '';
+		// This writer emits text/html; the HTML reader restores SVG's camel-cased names.
+		const tag = node.localName.toLowerCase();
+		const attributes = Array.from(node.attributes).sort((a, b) => a.name < b.name ? -1 : a.name > b.name ? 1 : 0)
+			.map(attribute => ' ' + attribute.name + '="' + escapeRapierHtmlText(attribute.value) + '"').join('');
+		const open = '<' + tag + attributes + '>';
+		return voidTags.has(tag) ? open : open + children(node) + '</' + tag + '>';
+	}
+	return children(root);
+}
+
 function _rapierProjectStyledRoot(semanticRoot, options) {
 	const opts = options || {};
 	const root = semanticRoot.cloneNode(true);
@@ -138,11 +159,12 @@ function _rapierProjectStyledRoot(semanticRoot, options) {
 		if (code) {
 			const lang = _rapierLanguageClass(code) || String(metadata.codeLang || '');
 			code.classList.add('artifact-code');
-			if (metadata.docKind === 'code') { _rapierPaintCode(code, _rapierArtifactHighlight(code.textContent || '', lang)); _rapierArtifactMarkLexed(code, lang); }
+			if (metadata.docKind === 'code' && !opts.sourceCode) { _rapierPaintCode(code, _rapierArtifactHighlight(code.textContent || '', lang)); _rapierArtifactMarkLexed(code, lang); }
 			if (lang) code.classList.add('language-' + String(lang).replace(/[^a-z0-9_-]/gi, ''));
 		}
 	} else {
 		root.querySelectorAll('pre > code').forEach(code => {
+			if (opts.sourceCode) return;
 			const lang = _rapierLanguageClass(code) || 'text';
 			_rapierPaintCode(code, _rapierArtifactHighlight(code.textContent || '', lang));
 			_rapierArtifactMarkLexed(code, lang);
@@ -157,7 +179,7 @@ function _rapierProjectStyledRoot(semanticRoot, options) {
 		});
 
 		root.querySelectorAll('math[display="block"]').forEach(math => {
-			if (math.closest('.math-display-wrap')) return;
+			if (math.closest('svg') || math.closest('.math-display-wrap')) return;
 			const wrap = document.createElement('span');
 			wrap.className = 'math-display-wrap';
 			math.parentNode.insertBefore(wrap, math);
@@ -184,7 +206,7 @@ function _rapierProjectStyledRoot(semanticRoot, options) {
 	_rapierPrefixPortableAnchors(root, opts.baseName || metadata.filename || 'document');
 
 	const clean = document.createElement('div');
-	clean.innerHTML = sanitizeRapierHtml(root.innerHTML, 'export');
+	clean.innerHTML = sanitizeRapierHtml(root.innerHTML, opts.sourceCode ? 'source' : 'export');
 	return clean;
 }
 

package/dist/spec/md-assets.mjs

@@ -370,7 +370,8 @@ export function serializeAsset(asset) {
   return '[' + asset.label + ']: ' + asset.url + title;
 }
 // Pure: re-derived at commit against the moved document; appends only at the text's end.
-export function appendAssetText(source, asset) {
+// A replay supplies the retained definition's line ending instead of choosing one from later prose.
+export function appendAssetText(source, asset, lineEnding) {
   if (typeof source !== 'string' || !asset || normalizeLabel(asset.label) !== asset.id) return fail('image_source_invalid');
   const parsed = documentAssets(source), title = assetTitle(asset.title);
   const existing = [...parsed.assets.values()].find(row => row.url === asset.url && row.title === title);
@@ -382,7 +383,7 @@ export function appendAssetText(source, asset) {
     return {source, id: asset.id, reference: asset.label, added: false, suffix: ''};
   }
   if (parsed.assets.size >= IMAGE_LIMITS.assets) return fail('image_asset_count_limit');
-  const eol = /\r\n|\n|\r/.exec(source)?.[0] || '\n', tail = /(?:\r\n?|\n)[ \t]*$/.exec(source);
+  const eol = lineEnding || /\r\n|\n|\r/.exec(source)?.[0] || '\n', tail = /(?:\r\n?|\n)[ \t]*$/.exec(source);
   const separator = !source || tail && /[\r\n][ \t]*$/.test(source.slice(0, tail.index)) ? '' : tail ? eol : eol + eol;
   const suffix = separator + block + eol, nextSource = source + suffix;
   if (nextSource.length > IMAGE_LIMITS.sourceChars) return fail('image_source_limit');

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "rapier-markdown-kit",
-  "version": "1.1.78",
+  "version": "1.1.80",
   "description": "Read, render, write and style Self-contained Markdown, and carry its optional edit ledger: the shared HTML writer, semantic renderer, stylesheet, layout, marks, pictures, Will/1, replay, authorship and offline merge. MIT, no npm dependencies.",
   "keywords": [
     "markdown",