chaffjs 0.27.1 → 0.29.0

package/dist/html/html-elements.js

@@ -0,0 +1,91 @@
+// The pieces of an HTML page the converters look at: tags, links, attributes and elements matched to their own
+// closing tag. Pure; a regular-expression reading, not a parser.
+import { decodeEntities } from "./markup-text.js";
+// In a tag a quote opens a value only after "=", and the value ends only at its matching quote, whatever it holds.
+const TAG_BODY = String.raw `(?:[^>=]|=\s*"[^"]*"|=\s*'[^']*'|=(?!\s*["']))*`;
+// A comment, a processing instruction, a script and a style are matched whole, so that no tag is read inside them
+// (the converter drops them anyway), and a "<!--" inside a value is escaped before comments are looked for.
+const TAG = new RegExp(String.raw `<!--[\s\S]*?-->|<\?[\s\S]*?\?>|<(script|style)\b${TAG_BODY}>[\s\S]*?<\/\1\s*>|<[a-z][a-z0-9-]*${TAG_BODY}>`, "giu");
+const QUOTED_VALUE = /=(\s*)("[^"]*"|'[^']*')/gu;
+const withValuesEscaped = (tag) => tag.replace(QUOTED_VALUE, (_whole, space, value) => `=${space}${value.replaceAll("<", "&lt;").replaceAll(">", "&gt;")}`);
+/**
+ * Each "<" and ">" inside a quoted attribute value (title="published under the <span>…</span>") as a character
+ * reference, which means the same in a value, so that a scanner reading a tag up to its first ">" reads all of it.
+ */
+export const withAttributeMarkupEscaped = (html) => html.replace(TAG, (tag) => withValuesEscaped(tag));
+/** Apply step until the text stops changing: nested elements are removed from the inside out. */
+export const untilStable = (text, step) => {
+    const next = step(text);
+    return next === text ? text : untilStable(next, step);
+};
+export const stripTags = (html) => html.replace(/<\/?[a-z!][^>]*>/giu, "");
+export const plainText = (html) => decodeEntities(stripTags(html)).replace(/\s+/gu, " ").trim();
+/** Text put back into markup, so that decoding it again gives the same text. */
+export const asMarkup = (text) => text.replaceAll("&", "&amp;").replaceAll("<", "&lt;");
+/** A full stop, question or exclamation mark closing a sentence, also before a closing quote or bracket; not the point in "3.5". */
+export const CLOSED_SENTENCE = /[。．！？]|[.!?][)\]"'”’]*(?=\s|$)/u;
+/** Whether the markup closes a sentence, each tag read as a space so that "Done.</li><li>Next" ends one. */
+export const hasSentence = (html) => CLOSED_SENTENCE.test(plainText(html.replace(/<\/?[a-z!][^>]*>/giu, " ")));
+export const hasNoWords = (text) => !/[\p{L}\p{N}]/u.test(text);
+export const ANY_LINK = /<a\b[^>]*>[\s\S]*?<\/a\s*>/giu;
+const HREF = /\shref\s*=\s*(?:"([^"]*)"|'([^']*)'|([^\s"'>]+))/iu;
+const PAGE_EXTENSION = /^(?:[sx]?html?|php|aspx?|jsp|cgi)$/iu;
+// The host is not part of the path: "https://example.com" names a site, not a file ending in .com.
+const SCHEME_AND_HOST = /^(?:[a-z][a-z0-9+.-]*:)?\/\/[^/?#]*/iu;
+// A link that opens no page (tel:, mailto:, sms:) carries its own content, a number or an address.
+const OTHER_SCHEME = /^\s*(?!https?:)[a-z][a-z0-9+.-]*:(?!\/\/)/iu;
+const hrefOf = (link) => {
+    const href = HREF.exec(link);
+    return href?.[1] ?? href?.[2] ?? href?.[3] ?? "";
+};
+/** The extension of the file a link opens ("pdf", "mp3"), or undefined for a page or a site. */
+export const fileExtension = (link) => {
+    const path = hrefOf(link).replace(SCHEME_AND_HOST, "").split(/[?#]/u)[0] ?? "";
+    const extension = /\.([a-z0-9]{1,5})$/iu.exec(path.split("/").at(-1) ?? "")?.[1];
+    return extension !== undefined && !PAGE_EXTENSION.test(extension) ? extension : undefined;
+};
+/** A link to a file (an appendix as PDF, a table as .xlsx), or one that opens no page, is part of the document. */
+export const isFileLink = (link) => OTHER_SCHEME.test(hrefOf(link)) || fileExtension(link) !== undefined;
+/** Attributes before a given one, each skipped whole so that a quoted value (title="x role=main") is not read as one. */
+export const ATTRIBUTES = String.raw `(?:\s+[^\s"'>=/]+(?:\s*=\s*(?:"[^"]*"|'[^']*'|[^\s"'>]+))?)*?`;
+const closeLast = (html, scan, close) => {
+    const opener = scan.open.at(-1);
+    if (opener === undefined)
+        return scan;
+    const range = {
+        start: opener.index,
+        end: close.index + close[0].length,
+        openTag: opener[0],
+        inner: html.slice(opener.index + opener[0].length, close.index),
+    };
+    return { open: scan.open.slice(0, -1), ranges: [...scan.ranges, range] };
+};
+/** Every <tag>…</tag>, nested ones matched to their own closing tag, in document order (outer before inner). */
+export const elementRanges = (html, tag) => [...html.matchAll(new RegExp(`<(/?)${tag}\\b[^>]*>`, "giu"))]
+    .reduce((scan, match) => (match[1] === "/" ? closeLast(html, scan, match) : { open: [...scan.open, match], ranges: scan.ranges }), {
+    open: [],
+    ranges: [],
+})
+    .ranges.toSorted((left, right) => left.start - right.start);
+export const isInside = (outer, inner) => outer.start < inner.start && inner.end <= outer.end;
+/** The ranges (in document order) each replaced by a space; one inside another cut one goes with it. */
+export const withoutRanges = (html, ranges) => {
+    const chosen = ranges.reduce((kept, range) => {
+        const insideKept = (kept.at(-1)?.end ?? 0) > range.start;
+        if (!insideKept)
+            kept.push(range);
+        return kept;
+    }, []);
+    const cut = chosen.reduce((acc, range) => ({ parts: [...acc.parts, html.slice(acc.from, range.start), " "], from: range.end }), { parts: [], from: 0 });
+    return [...cut.parts, html.slice(cut.from)].join("");
+};
+/** The <tag> elements that are chrome, each replaced by a space; one inside another dropped one goes with it. */
+export const withoutElementsWhere = (html, tag, isChrome) => withoutRanges(html, elementRanges(html, tag).filter(isChrome));
+/** Every <tag> whose opening matches, for each tag that opens that way, replaced by a space. */
+export const withoutElementsOpening = (html, opening, isChrome) => {
+    const tags = new Set([...html.matchAll(opening)].map((match) => (match[1] ?? "").toLowerCase()));
+    return [...tags].reduce((text, tag) => withoutElementsWhere(text, tag, isChrome), html);
+};
+export const HEADING_TAGS = ["h1", "h2", "h3", "h4", "h5", "h6"];
+export const headingRanges = (html) => HEADING_TAGS.flatMap((tag) => elementRanges(html, tag));
+//# sourceMappingURL=html-elements.js.map

package/dist/html/html-elements.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"html-elements.js","sourceRoot":"","sources":["../../src/html/html-elements.ts"],"names":[],"mappings":"AAAA,+GAA+G;AAC/G,iEAAiE;AACjE,OAAO,EAAE,cAAc,EAAE,MAAM,kBAAkB,CAAC;AAElD,mHAAmH;AACnH,MAAM,QAAQ,GAAG,MAAM,CAAC,GAAG,CAAA,iDAAiD,CAAC;AAE7E,kHAAkH;AAClH,4GAA4G;AAC5G,MAAM,GAAG,GAAG,IAAI,MAAM,CAAC,MAAM,CAAC,GAAG,CAAA,mDAAmD,QAAQ,sCAAsC,QAAQ,GAAG,EAAE,KAAK,CAAC,CAAC;AAEtJ,MAAM,YAAY,GAAG,2BAA2B,CAAC;AAEjD,MAAM,iBAAiB,GAAG,CAAC,GAAW,EAAU,EAAE,CAChD,GAAG,CAAC,OAAO,CAAC,YAAY,EAAE,CAAC,MAAc,EAAE,KAAa,EAAE,KAAa,EAAE,EAAE,CAAC,IAAI,KAAK,GAAG,KAAK,CAAC,UAAU,CAAC,GAAG,EAAE,MAAM,CAAC,CAAC,UAAU,CAAC,GAAG,EAAE,MAAM,CAAC,EAAE,CAAC,CAAC;AAEnJ;;;GAGG;AACH,MAAM,CAAC,MAAM,0BAA0B,GAAG,CAAC,IAAY,EAAU,EAAE,CAAC,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE,CAAC,GAAW,EAAE,EAAE,CAAC,iBAAiB,CAAC,GAAG,CAAC,CAAC,CAAC;AAE/H,iGAAiG;AACjG,MAAM,CAAC,MAAM,WAAW,GAAG,CAAC,IAAY,EAAE,IAA8B,EAAU,EAAE;IAClF,MAAM,IAAI,GAAG,IAAI,CAAC,IAAI,CAAC,CAAC;IACxB,OAAO,IAAI,KAAK,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,WAAW,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC;AACxD,CAAC,CAAC;AAEF,MAAM,CAAC,MAAM,SAAS,GAAG,CAAC,IAAY,EAAU,EAAE,CAAC,IAAI,CAAC,OAAO,CAAC,qBAAqB,EAAE,EAAE,CAAC,CAAC;AAE3F,MAAM,CAAC,MAAM,SAAS,GAAG,CAAC,IAAY,EAAU,EAAE,CAAC,cAAc,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,OAAO,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;AAEhH,gFAAgF;AAChF,MAAM,CAAC,MAAM,QAAQ,GAAG,CAAC,IAAY,EAAU,EAAE,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,EAAE,OAAO,CAAC,CAAC,UAAU,CAAC,GAAG,EAAE,MAAM,CAAC,CAAC;AAExG,oIAAoI;AACpI,MAAM,CAAC,MAAM,eAAe,GAAG,iCAAiC,CAAC;AAEjE,4GAA4G;AAC5G,MAAM,CAAC,MAAM,WAAW,GAAG,CAAC,IAAY,EAAW,EAAE,CAAC,eAAe,CAAC,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,OAAO,CAAC,qBAAqB,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC;AAEhI,MAAM,CAAC,MAAM,UAAU,GAAG,CAAC,IAAY,EAAW,EAAE,CAAC,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;AAEjF,MAAM,CAAC,MAAM,QAAQ,GAAG,+BAA+B,CAAC;AAExD,MAAM,IAAI,GAAG,oDAAoD,CAAC;AAElE,MAAM,cAAc,GAAG,sCAAsC,CAAC;AAE9D,mGAAmG;AACnG,MAAM,eAAe,GAAG,uCAAuC,CAAC;AAEhE,mGAAmG;AACnG,MAAM,YAAY,GAAG,6CAA6C,CAAC;AAEnE,MAAM,MAAM,GAAG,CAAC,IAAY,EAAU,EAAE;IACtC,MAAM,IAAI,GAAG,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAC7B,OAAO,IAAI,EAAE,CAAC,CAAC,CAAC,IAAI,IAAI,EAAE,CAAC,CAAC,CAAC,IAAI,IAAI,EAAE,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;AACnD,CAAC,CAAC;AAEF,gGAAgG;AAChG,MAAM,CAAC,MAAM,aAAa,GAAG,CAAC,IAAY,EAAsB,EAAE;IAChE,MAAM,IAAI,GAAG,MAAM,CAAC,IAAI,CAAC,CAAC,OAAO,CAAC,eAAe,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;IAC/E,MAAM,SAAS,GAAG,sBAAsB,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;IACjF,OAAO,SAAS,KAAK,SAAS,IAAI,CAAC,cAAc,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC;AAC5F,CAAC,CAAC;AAEF,mHAAmH;AACnH,MAAM,CAAC,MAAM,UAAU,GAAG,CAAC,IAAY,EAAW,EAAE,CAAC,YAAY,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,IAAI,aAAa,CAAC,IAAI,CAAC,KAAK,SAAS,CAAC;AAE1H,yHAAyH;AACzH,MAAM,CAAC,MAAM,UAAU,GAAG,MAAM,CAAC,GAAG,CAAA,+DAA+D,CAAC;AAMpG,MAAM,SAAS,GAAG,CAAC,IAAY,EAAE,IAAe,EAAE,KAAsB,EAAa,EAAE;IACrF,MAAM,MAAM,GAAG,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;IAChC,IAAI,MAAM,KAAK,SAAS;QAAE,OAAO,IAAI,CAAC;IACtC,MAAM,KAAK,GAAG;QACZ,KAAK,EAAE,MAAM,CAAC,KAAK;QACnB,GAAG,EAAE,KAAK,CAAC,KAAK,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,MAAM;QAClC,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC;QAClB,KAAK,EAAE,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,KAAK,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,MAAM,EAAE,KAAK,CAAC,KAAK,CAAC;KAChE,CAAC;IACF,OAAO,EAAE,IAAI,EAAE,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,GAAG,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC,EAAE,CAAC;AAC3E,CAAC,CAAC;AAEF,gHAAgH;AAChH,MAAM,CAAC,MAAM,aAAa,GAAG,CAAC,IAAY,EAAE,GAAW,EAAkB,EAAE,CACzE,CAAC,GAAG,IAAI,CAAC,QAAQ,CAAC,IAAI,MAAM,CAAC,QAAQ,GAAG,WAAW,EAAE,KAAK,CAAC,CAAC,CAAC;KAC1D,MAAM,CAAY,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,KAAK,GAAG,CAAC,CAAC,CAAC,SAAS,CAAC,IAAI,EAAE,IAAI,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,CAAC,GAAG,IAAI,CAAC,IAAI,EAAE,KAAK,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM,EAAE,CAAC,EAAE;IAC5I,IAAI,EAAE,EAAE;IACR,MAAM,EAAE,EAAE;CACX,CAAC;KACD,MAAM,CAAC,QAAQ,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,CAAC;AAEhE,MAAM,CAAC,MAAM,QAAQ,GAAG,CAAC,KAAmB,EAAE,KAAmB,EAAW,EAAE,CAAC,KAAK,CAAC,KAAK,GAAG,KAAK,CAAC,KAAK,IAAI,KAAK,CAAC,GAAG,IAAI,KAAK,CAAC,GAAG,CAAC;AAEnI,wGAAwG;AACxG,MAAM,CAAC,MAAM,aAAa,GAAG,CAAC,IAAY,EAAE,MAA+B,EAAU,EAAE;IACrF,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,CAAiB,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE;QAC3D,MAAM,UAAU,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG,IAAI,CAAC,CAAC,GAAG,KAAK,CAAC,KAAK,CAAC;QACzD,IAAI,CAAC,UAAU;YAAE,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QAClC,OAAO,IAAI,CAAC;IACd,CAAC,EAAE,EAAE,CAAC,CAAC;IACP,MAAM,GAAG,GAAG,MAAM,CAAC,MAAM,CACvB,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC,EAAE,KAAK,EAAE,CAAC,GAAG,GAAG,CAAC,KAAK,EAAE,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,EAAE,KAAK,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,EAAE,IAAI,EAAE,KAAK,CAAC,GAAG,EAAE,CAAC,EACpG,EAAE,KAAK,EAAE,EAAE,EAAE,IAAI,EAAE,CAAC,EAAE,CACvB,CAAC;IACF,OAAO,CAAC,GAAG,GAAG,CAAC,KAAK,EAAE,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;AACvD,CAAC,CAAC;AAEF,iHAAiH;AACjH,MAAM,CAAC,MAAM,oBAAoB,GAAG,CAAC,IAAY,EAAE,GAAW,EAAE,QAA0C,EAAU,EAAE,CACpH,aAAa,CAAC,IAAI,EAAE,aAAa,CAAC,IAAI,EAAE,GAAG,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC;AAEjE,gGAAgG;AAChG,MAAM,CAAC,MAAM,sBAAsB,GAAG,CAAC,IAAY,EAAE,OAAe,EAAE,QAA0C,EAAU,EAAE;IAC1H,MAAM,IAAI,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC;IACjG,OAAO,CAAC,GAAG,IAAI,CAAC,CAAC,MAAM,CAAC,CAAC,IAAI,EAAE,GAAG,EAAE,EAAE,CAAC,oBAAoB,CAAC,IAAI,EAAE,GAAG,EAAE,QAAQ,CAAC,EAAE,IAAI,CAAC,CAAC;AAC1F,CAAC,CAAC;AAEF,MAAM,CAAC,MAAM,YAAY,GAAG,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC,CAAC;AAEjE,MAAM,CAAC,MAAM,aAAa,GAAG,CAAC,IAAY,EAAkB,EAAE,CAAC,YAAY,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,aAAa,CAAC,IAAI,EAAE,GAAG,CAAC,CAAC,CAAC"}

package/dist/html/html-footnote-marks.d.ts

@@ -0,0 +1,2 @@
+export declare const withoutFootnoteMarks: (html: string) => string;
+//# sourceMappingURL=html-footnote-marks.d.ts.map

package/dist/html/html-footnote-marks.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"html-footnote-marks.d.ts","sourceRoot":"","sources":["../../src/html/html-footnote-marks.ts"],"names":[],"mappings":"AAgBA,eAAO,MAAM,oBAAoB,GAAI,MAAM,MAAM,KAAG,MAG8E,CAAC"}

package/dist/html/html-footnote-marks.js

@@ -0,0 +1,15 @@
+// A footnote's reference mark: a superscript that is a link (<sup><a …>1</a></sup>), or a link to a place on the same
+// page holding nothing but a superscript (<a href="#fn1"><sup>1</sup></a>). Read as text the number sticks to the
+// sentence it follows ("recessions.1 Over"), so the mark is dropped; the note itself stays where the page has it. A
+// superscript that is a link is a mark wherever the link goes (the notes may be on another page); a link around a
+// superscript is one only when it stays on this page, since "Brand<a href=…><sup>TM</sup></a>" is a sign. A superscript
+// with no link ("25<sup>th</sup>") is text. Pure.
+const LINK_IN_SUPERSCRIPT = /<sup\b[^>]*>\s*<a\b[^>]*>[^<]*<\/a\s*>\s*<\/sup\s*>/giu;
+const SUPERSCRIPT_LINK = /<a\b([^>]*)>\s*<sup\b[^>]*>([^<]*)<\/sup\s*>\s*<\/a\s*>/giu;
+// The fragment must name something: href="#" alone goes nowhere and is a control.
+const IN_PAGE_HREF = /\shref\s*=\s*(?:"#[^"\s]+"|'#[^'\s]+'|#[^\s"'>]+)/iu;
+const isFootnoteLink = (attributes, mark) => IN_PAGE_HREF.test(attributes) && mark.trim() !== "";
+export const withoutFootnoteMarks = (html) => html
+    .replace(LINK_IN_SUPERSCRIPT, "")
+    .replace(SUPERSCRIPT_LINK, (link, attributes, mark) => (isFootnoteLink(attributes, mark) ? "" : link));
+//# sourceMappingURL=html-footnote-marks.js.map

package/dist/html/html-footnote-marks.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"html-footnote-marks.js","sourceRoot":"","sources":["../../src/html/html-footnote-marks.ts"],"names":[],"mappings":"AAAA,sHAAsH;AACtH,kHAAkH;AAClH,oHAAoH;AACpH,kHAAkH;AAClH,wHAAwH;AACxH,kDAAkD;AAElD,MAAM,mBAAmB,GAAG,wDAAwD,CAAC;AAErF,MAAM,gBAAgB,GAAG,4DAA4D,CAAC;AAEtF,kFAAkF;AAClF,MAAM,YAAY,GAAG,qDAAqD,CAAC;AAE3E,MAAM,cAAc,GAAG,CAAC,UAAkB,EAAE,IAAY,EAAW,EAAE,CAAC,YAAY,CAAC,IAAI,CAAC,UAAU,CAAC,IAAI,IAAI,CAAC,IAAI,EAAE,KAAK,EAAE,CAAC;AAE1H,MAAM,CAAC,MAAM,oBAAoB,GAAG,CAAC,IAAY,EAAU,EAAE,CAC3D,IAAI;KACD,OAAO,CAAC,mBAAmB,EAAE,EAAE,CAAC;KAChC,OAAO,CAAC,gBAAgB,EAAE,CAAC,IAAY,EAAE,UAAkB,EAAE,IAAY,EAAE,EAAE,CAAC,CAAC,cAAc,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC"}

package/dist/html/html-heading-links.d.ts

@@ -0,0 +1,6 @@
+/**
+ * A heading's link to its own section goes when it stands just after the heading ("Copy link to Context"), or inside
+ * it as a mark without words ("Scope ¶"). A link with words inside the heading is its title and stays.
+ */
+export declare const withoutHeadingSelfLinks: (html: string) => string;
+//# sourceMappingURL=html-heading-links.d.ts.map

package/dist/html/html-heading-links.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"html-heading-links.d.ts","sourceRoot":"","sources":["../../src/html/html-heading-links.ts"],"names":[],"mappings":"AA8EA;;;GAGG;AACH,eAAO,MAAM,uBAAuB,GAAI,MAAM,MAAM,KAAG,MAUtD,CAAC"}

package/dist/html/html-heading-links.js

@@ -0,0 +1,71 @@
+// A heading's link to its own section (a permalink beside or after the title), taken out of an HTML page. Pure.
+import { ANY_LINK, ATTRIBUTES, elementRanges, hasNoWords, plainText } from "./html-elements.js";
+import { decodeEntities } from "./markup-text.js";
+const ID = new RegExp(String.raw `<([a-z][a-z0-9-]*)${ATTRIBUTES}\s+id\s*=\s*(?:"([^"]*)"|'([^']*)'|([^\s"'>]+))`, "giu");
+/** Where each id's element opens, and which element it is; the first one when an id is repeated. */
+const idTargets = (html) => [...html.matchAll(ID)].reduce((targets, match) => {
+    const id = decodeEntities(match[2] ?? match[3] ?? match[4] ?? "");
+    return targets.has(id) ? targets : targets.set(id, { start: match.index, tag: (match[1] ?? "").toLowerCase() });
+}, new Map());
+const FRAGMENT = /^<a\b[^>]*\bhref\s*=\s*["']?[^"'#\s>]*#([^"'\s>]+)/iu;
+/** The id a link's fragment names, its character references and percent escapes undone. */
+const fragmentId = (link) => {
+    const fragment = FRAGMENT.exec(link)?.[1];
+    if (fragment === undefined)
+        return undefined;
+    const decoded = decodeEntities(fragment);
+    try {
+        return decodeURIComponent(decoded);
+    }
+    catch {
+        return decoded;
+    }
+};
+const HEADING_OPENING = /<h[1-6]\b/iu;
+/** The target opens before the heading and closes after it, with no other heading opening in between. */
+const isSectionOf = (target, place) => place.rangesOf(target.tag).some((range) => range.start === target.start && range.end >= place.end) &&
+    !HEADING_OPENING.test(place.html.slice(target.start, place.start));
+/** A link to the heading's own section: the heading, something inside it, or the block it opens (a permalink, "Copy link to Context"). */
+const isSelfLink = (link, place) => {
+    const id = fragmentId(link);
+    const target = id === undefined ? undefined : place.ids.get(id);
+    if (target === undefined || target.start >= place.end)
+        return false;
+    return target.start >= place.start || isSectionOf(target, place);
+};
+/** A permalink mark inside the heading (¶, #, an icon); a self-link with words is the title, or part of it. */
+const isPermalinkMark = (link, place) => hasNoWords(plainText(link)) && isSelfLink(link, place);
+const HEADING_THEN_LINK = /(<h([1-6])\b[^>]*>)([\s\S]*?)(<\/h\2\s*>)(\s*<a\b[^>]*>[\s\S]*?<\/a\s*>)?/giu;
+/** The heading as matched, without its permalink marks or the self-link standing after it. */
+const headingWithoutSelfLinks = (match, page) => {
+    const [, open = "", , inside = "", close = "", after] = match;
+    const place = { ...page, start: match.index, end: match.index + open.length + inside.length + close.length };
+    const kept = inside.replace(ANY_LINK, (link) => (isPermalinkMark(link, place) ? " " : link));
+    const next = after === undefined || isSelfLink(after.trim(), place) ? "" : after;
+    return `${open}${kept}${close}${next}`;
+};
+/** elementRanges for each tag, computed once. */
+const rangesByTag = (html) => {
+    const cache = new Map();
+    return (tag) => {
+        const known = cache.get(tag);
+        if (known !== undefined)
+            return known;
+        const ranges = elementRanges(html, tag);
+        cache.set(tag, ranges);
+        return ranges;
+    };
+};
+/**
+ * A heading's link to its own section goes when it stands just after the heading ("Copy link to Context"), or inside
+ * it as a mark without words ("Scope ¶"). A link with words inside the heading is its title and stays.
+ */
+export const withoutHeadingSelfLinks = (html) => {
+    const page = { ids: idTargets(html), html, rangesOf: rangesByTag(html) };
+    const cut = [...html.matchAll(HEADING_THEN_LINK)].reduce((acc, match) => ({
+        parts: [...acc.parts, html.slice(acc.from, match.index), headingWithoutSelfLinks(match, page)],
+        from: match.index + match[0].length,
+    }), { parts: [], from: 0 });
+    return [...cut.parts, html.slice(cut.from)].join("");
+};
+//# sourceMappingURL=html-heading-links.js.map

package/dist/html/html-heading-links.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"html-heading-links.js","sourceRoot":"","sources":["../../src/html/html-heading-links.ts"],"names":[],"mappings":"AAAA,gHAAgH;AAChH,OAAO,EAAE,QAAQ,EAAE,UAAU,EAAE,aAAa,EAAE,UAAU,EAAE,SAAS,EAAqB,MAAM,oBAAoB,CAAC;AACnH,OAAO,EAAE,cAAc,EAAE,MAAM,kBAAkB,CAAC;AAElD,MAAM,EAAE,GAAG,IAAI,MAAM,CAAC,MAAM,CAAC,GAAG,CAAA,qBAAqB,UAAU,iDAAiD,EAAE,KAAK,CAAC,CAAC;AAIzH,oGAAoG;AACpG,MAAM,SAAS,GAAG,CAAC,IAAY,EAAyB,EAAE,CACxD,CAAC,GAAG,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,OAAO,EAAE,KAAK,EAAE,EAAE;IAC/C,MAAM,EAAE,GAAG,cAAc,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC;IAClE,OAAO,OAAO,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,EAAE,EAAE,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,GAAG,EAAE,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,WAAW,EAAE,EAAE,CAAC,CAAC;AAClH,CAAC,EAAE,IAAI,GAAG,EAAoB,CAAC,CAAC;AAElC,MAAM,QAAQ,GAAG,sDAAsD,CAAC;AAExE,2FAA2F;AAC3F,MAAM,UAAU,GAAG,CAAC,IAAY,EAAsB,EAAE;IACtD,MAAM,QAAQ,GAAG,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;IAC1C,IAAI,QAAQ,KAAK,SAAS;QAAE,OAAO,SAAS,CAAC;IAC7C,MAAM,OAAO,GAAG,cAAc,CAAC,QAAQ,CAAC,CAAC;IACzC,IAAI,CAAC;QACH,OAAO,kBAAkB,CAAC,OAAO,CAAC,CAAC;IACrC,CAAC;IAAC,MAAM,CAAC;QACP,OAAO,OAAO,CAAC;IACjB,CAAC;AACH,CAAC,CAAC;AAEF,MAAM,eAAe,GAAG,aAAa,CAAC;AAUtC,yGAAyG;AACzG,MAAM,WAAW,GAAG,CAAC,MAAgB,EAAE,KAAmB,EAAW,EAAE,CACrE,KAAK,CAAC,QAAQ,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,KAAK,KAAK,MAAM,CAAC,KAAK,IAAI,KAAK,CAAC,GAAG,IAAI,KAAK,CAAC,GAAG,CAAC;IAClG,CAAC,eAAe,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,KAAK,EAAE,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC;AAErE,0IAA0I;AAC1I,MAAM,UAAU,GAAG,CAAC,IAAY,EAAE,KAAmB,EAAW,EAAE;IAChE,MAAM,EAAE,GAAG,UAAU,CAAC,IAAI,CAAC,CAAC;IAC5B,MAAM,MAAM,GAAG,EAAE,KAAK,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;IAChE,IAAI,MAAM,KAAK,SAAS,IAAI,MAAM,CAAC,KAAK,IAAI,KAAK,CAAC,GAAG;QAAE,OAAO,KAAK,CAAC;IACpE,OAAO,MAAM,CAAC,KAAK,IAAI,KAAK,CAAC,KAAK,IAAI,WAAW,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;AACnE,CAAC,CAAC;AAEF,+GAA+G;AAC/G,MAAM,eAAe,GAAG,CAAC,IAAY,EAAE,KAAmB,EAAW,EAAE,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,IAAI,UAAU,CAAC,IAAI,EAAE,KAAK,CAAC,CAAC;AAE/H,MAAM,iBAAiB,GAAG,8EAA8E,CAAC;AAEzG,8FAA8F;AAC9F,MAAM,uBAAuB,GAAG,CAAC,KAAsB,EAAE,IAAa,EAAU,EAAE;IAChF,MAAM,CAAC,EAAE,IAAI,GAAG,EAAE,EAAE,AAAD,EAAG,MAAM,GAAG,EAAE,EAAE,KAAK,GAAG,EAAE,EAAE,KAAK,CAAC,GAAG,KAAK,CAAC;IAC9D,MAAM,KAAK,GAAG,EAAE,GAAG,IAAI,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,GAAG,EAAE,KAAK,CAAC,KAAK,GAAG,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,GAAG,KAAK,CAAC,MAAM,EAAE,CAAC;IAC7G,MAAM,IAAI,GAAG,MAAM,CAAC,OAAO,CAAC,QAAQ,EAAE,CAAC,IAAY,EAAE,EAAE,CAAC,CAAC,eAAe,CAAC,IAAI,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC;IACrG,MAAM,IAAI,GAAG,KAAK,KAAK,SAAS,IAAI,UAAU,CAAC,KAAK,CAAC,IAAI,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC;IACjF,OAAO,GAAG,IAAI,GAAG,IAAI,GAAG,KAAK,GAAG,IAAI,EAAE,CAAC;AACzC,CAAC,CAAC;AAEF,iDAAiD;AACjD,MAAM,WAAW,GAAG,CAAC,IAAY,EAA8C,EAAE;IAC/E,MAAM,KAAK,GAAG,IAAI,GAAG,EAAmC,CAAC;IACzD,OAAO,CAAC,GAAW,EAAE,EAAE;QACrB,MAAM,KAAK,GAAG,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;QAC7B,IAAI,KAAK,KAAK,SAAS;YAAE,OAAO,KAAK,CAAC;QACtC,MAAM,MAAM,GAAG,aAAa,CAAC,IAAI,EAAE,GAAG,CAAC,CAAC;QACxC,KAAK,CAAC,GAAG,CAAC,GAAG,EAAE,MAAM,CAAC,CAAC;QACvB,OAAO,MAAM,CAAC;IAChB,CAAC,CAAC;AACJ,CAAC,CAAC;AAEF;;;GAGG;AACH,MAAM,CAAC,MAAM,uBAAuB,GAAG,CAAC,IAAY,EAAU,EAAE;IAC9D,MAAM,IAAI,GAAG,EAAE,GAAG,EAAE,SAAS,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,CAAC,IAAI,CAAC,EAAE,CAAC;IACzE,MAAM,GAAG,GAAG,CAAC,GAAG,IAAI,CAAC,QAAQ,CAAC,iBAAiB,CAAC,CAAC,CAAC,MAAM,CACtD,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC;QACf,KAAK,EAAE,CAAC,GAAG,GAAG,CAAC,KAAK,EAAE,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,EAAE,KAAK,CAAC,KAAK,CAAC,EAAE,uBAAuB,CAAC,KAAK,EAAE,IAAI,CAAC,CAAC;QAC9F,IAAI,EAAE,KAAK,CAAC,KAAK,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,MAAM;KACpC,CAAC,EACF,EAAE,KAAK,EAAE,EAAE,EAAE,IAAI,EAAE,CAAC,EAAE,CACvB,CAAC;IACF,OAAO,CAAC,GAAG,GAAG,CAAC,KAAK,EAAE,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;AACvD,CAAC,CAAC"}

package/dist/html/html-lost-tags.d.ts

@@ -0,0 +1,3 @@
+/** The HTML with each lost paragraph start tag in its text put back as <p>. */
+export declare const withLostParagraphTagsRestored: (html: string) => string;
+//# sourceMappingURL=html-lost-tags.d.ts.map

package/dist/html/html-lost-tags.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"html-lost-tags.d.ts","sourceRoot":"","sources":["../../src/html/html-lost-tags.ts"],"names":[],"mappings":"AAWA,+EAA+E;AAC/E,eAAO,MAAM,6BAA6B,GAAI,MAAM,MAAM,KAAG,MAC2C,CAAC"}

package/dist/html/html-lost-tags.js

@@ -0,0 +1,11 @@
+// A paragraph's start tag that lost its "<" upstream and so reached the page as text: the eCFR serves 16 CFR 310.4(b)
+// with "… 45 CFR 160.103. P&gt;(2) It is an abusive …", where "P>" is what is left of the <P> that opened paragraph (2).
+// Read as text it glues the paragraph to the one before and leaves "P>" in the sentence. Only a P and a ">" standing
+// after a sentence's end and right before a paragraph's designation ("(2)", "(iv)", "(A)") and a capital are taken for the lost
+// tag; "P>0.05", "P > (2)" and a formula opening a sentence ("P>(2) follows") stay text. Pure.
+const LOST_PARAGRAPH_TAG = /(?<=[.;:][)"'”’]?[ \t])[Pp](?:&gt;|>)(?=\([0-9A-Za-z]{1,4}\)[ \t]\p{Lu})/gu;
+/** Whether at is inside a tag. An attribute's ">" is escaped by then (withAttributeMarkupEscaped), so a tag has none before its end. */
+const isInsideTag = (html, at) => html.lastIndexOf("<", at) > html.lastIndexOf(">", at - 1);
+/** The HTML with each lost paragraph start tag in its text put back as <p>. */
+export const withLostParagraphTagsRestored = (html) => html.replace(LOST_PARAGRAPH_TAG, (lost, at) => (isInsideTag(html, at) ? lost : "<p>"));
+//# sourceMappingURL=html-lost-tags.js.map

package/dist/html/html-lost-tags.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"html-lost-tags.js","sourceRoot":"","sources":["../../src/html/html-lost-tags.ts"],"names":[],"mappings":"AAAA,sHAAsH;AACtH,yHAAyH;AACzH,qHAAqH;AACrH,gIAAgI;AAChI,+FAA+F;AAE/F,MAAM,kBAAkB,GAAG,4EAA4E,CAAC;AAExG,wIAAwI;AACxI,MAAM,WAAW,GAAG,CAAC,IAAY,EAAE,EAAU,EAAW,EAAE,CAAC,IAAI,CAAC,WAAW,CAAC,GAAG,EAAE,EAAE,CAAC,GAAG,IAAI,CAAC,WAAW,CAAC,GAAG,EAAE,EAAE,GAAG,CAAC,CAAC,CAAC;AAErH,+EAA+E;AAC/E,MAAM,CAAC,MAAM,6BAA6B,GAAG,CAAC,IAAY,EAAU,EAAE,CACpE,IAAI,CAAC,OAAO,CAAC,kBAAkB,EAAE,CAAC,IAAY,EAAE,EAAU,EAAE,EAAE,CAAC,CAAC,WAAW,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC"}

package/dist/html/html-markdown.d.ts

@@ -0,0 +1,2 @@
+export declare const htmlToMarkdown: (html: string) => string;
+//# sourceMappingURL=html-markdown.d.ts.map

package/dist/html/html-markdown.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"html-markdown.d.ts","sourceRoot":"","sources":["../../src/html/html-markdown.ts"],"names":[],"mappings":"AA+cA,eAAO,MAAM,cAAc,GAAI,MAAM,MAAM,KAAG,MAiB7C,CAAC"}

package/dist/html/html-markdown.js

@@ -0,0 +1,347 @@
+// An HTML page (a CRS report as EveryCRSReport serves it, a ministry's page) as plain Markdown: headings, paragraphs,
+// list items and the text of links; ruby keeps its base text and loses its reading. Only the <main> element (else the
+// role="main" element, else a sole <article>, else the block holding the title and nearly every sentence) is read when
+// the page has one. Scripts, styles, the head, navigation (by
+// element or by role, and breadcrumbs), asides, footers, forms, tables without sentences (html-tables.ts), footnote marks, lists and blocks of nothing but
+// links (a menu, a table of contents, previous/next links, a breadcrumb trail, also as a list ending in the page's
+// title), a list of the page in other languages, a block of icons with only a label, a block before the page's title that holds a menu or only links and no sentence (the site's header, with its tagline and
+// labels), buttons outside a heading, controls (html-widgets.ts: tooltips, media players, scripted link buttons, paging
+// bars, unspoken text), hidden elements, text for a screen reader only (a visually-hidden class, a skip
+// link's target), a heading's link to its own section beside or after its title,
+// lines of nothing but in-page or script links (never a heading), a heading drawn as an image unless its alt text is the page's
+// title, blocks of nothing but links closing the page, and a copyright notice closing the page, with an address just
+// before it, are dropped. A paragraph tag that lost its "<" upstream (html-lost-tags.ts) opens its paragraph again. Markup's spacing collapses; U+3000 is text and stays. Pure; a regular-expression reading that
+// is enough for the documents in the corpus, not a parser for any HTML.
+import { ANY_LINK, ATTRIBUTES, CLOSED_SENTENCE, asMarkup, elementRanges, hasNoWords, hasSentence, isFileLink, isInside, headingRanges, plainText, stripTags, withAttributeMarkupEscaped, withoutElementsOpening, withoutElementsWhere, withoutRanges, untilStable, } from "./html-elements.js";
+import { contentBlock } from "./html-content-block.js";
+import { withoutFootnoteMarks } from "./html-footnote-marks.js";
+import { withoutHeadingSelfLinks } from "./html-heading-links.js";
+import { withLostParagraphTagsRestored } from "./html-lost-tags.js";
+import { withoutReaderOnlyText } from "./html-reader-only.js";
+import { withRunInHeadingsRead } from "./html-run-in-headings.js";
+import { withoutWidgets } from "./html-widgets.js";
+import { withPreformattedRestored, withPreformattedStashed } from "./html-preformatted.js";
+import { withTablesRead } from "./html-tables.js";
+import { decodeEntities, tidyLines } from "./markup-text.js";
+const DROPPED = ["script", "style", "head", "nav", "aside", "footer", "form", "noscript", "svg"];
+const withoutElement = (html, tag) => {
+    const innermost = new RegExp(`<${tag}\\b[^>]*>(?:(?!<${tag}\\b)[\\s\\S])*?</${tag}\\s*>`, "giu");
+    return untilStable(html, (text) => text.replace(innermost, " "));
+};
+// HTML lets a ruby's parts omit their closing tags: a reading container (<rtc>) ends at the next <rb> or <rtc>, a
+// reading (<rt>) or its bracket (<rp>) at the next <rb>, <rt> or <rp>; any of them at the end of the <ruby>.
+const RUBY_TEXT_CONTAINER = /<rtc\b[^>]*>[\s\S]*?(?:<\/rtc\s*>|(?=<(?:rb|rtc)\b|$))/giu;
+const RUBY_TEXT = /<(rt|rp)\b[^>]*>[\s\S]*?(?:<\/\1\s*>|(?=<(?:rb|rt|rp)\b|$))/giu;
+/** Ruby as its base text alone: 漢字 with its reading becomes 漢字. */
+const withoutRubyText = (html) => html.replace(/<ruby\b[^>]*>([\s\S]*?)<\/ruby\s*>/giu, (_whole, inside) => inside.replace(RUBY_TEXT_CONTAINER, "").replace(RUBY_TEXT, ""));
+/**
+ * A link that moves within the page (a table of contents, "back to top"), through a series (rel="prev" / "next"), or
+ * goes nowhere and runs a script instead (href="javascript:…", a print or share button).
+ */
+const isChromeLink = (link) => /^<a\b[^>]*\bhref\s*=\s*(?:["']\s*)?(?:#|javascript:)/iu.test(link) || /^<a\b[^>]*\brel\s*=\s*(?:["'][^"']*\b)?(?:prev|next)\b/iu.test(link);
+/** A list whose every item is only a link, such as a site menu or a table of contents. */
+const isNavigation = (body) => [...body.matchAll(ANY_LINK)].length > 0 && stripTags(body.replace(ANY_LINK, "")).trim() === "";
+// A role is a list of tokens and the first one counts: role="main document" is main.
+const MAIN_ROLE = String.raw `${ATTRIBUTES}\s+role\s*=\s*(?:"\s*main(?:\s[^"]*)?"|'\s*main(?:\s[^']*)?'|main(?=[\s/>]))`;
+/** The first element marked role="main" (a CMS's <article id="contents" role="main">), matched to its own closing tag. */
+const mainLandmark = (html) => {
+    const tag = new RegExp(String.raw `<([a-z][a-z0-9]*)${MAIN_ROLE}`, "iu").exec(html)?.[1];
+    if (tag === undefined)
+        return undefined;
+    return elementRanges(html, tag).find((range) => new RegExp(String.raw `^<[a-z][a-z0-9]*${MAIN_ROLE}`, "iu").test(range.openTag))?.inner;
+};
+const isOutermost = (range, _index, all) => !all.some((outer) => isInside(outer, range));
+const holdsEveryTitle = (html, range) => [...html.matchAll(/<h1\b/giu)].every((title) => title.index > range.start && title.index < range.end);
+/**
+ * The one <article> not nested in another, when the page has exactly one and no <h1> outside it: the composition the
+ * page was made for. A page title outside the article means the article is only part of the page's content.
+ */
+const soleArticle = (html) => {
+    const [only, ...others] = elementRanges(html, "article").filter(isOutermost);
+    return only !== undefined && others.length === 0 && holdsEveryTitle(html, only) ? only.inner : undefined;
+};
+/**
+ * The page's own content: inside <main>, else the element marked role="main", else a sole <article>, else the block
+ * holding the title and nearly every sentence (html-content-block.ts), else the whole page.
+ */
+const mainContent = (html) => /<main\b[^>]*>([\s\S]*)<\/main\s*>/iu.exec(html)?.[1] ?? mainLandmark(html) ?? soleArticle(html) ?? contentBlock(html) ?? html;
+const LANDMARK = String.raw `(?:\brole\s*=\s*["']?navigation\b|\baria-label\s*=\s*(?:["'][^"']*|[^\s"'>]*)breadcrumb)`;
+const LANDMARK_OPENING = new RegExp(String.raw `<([a-z][a-z0-9]*)\b[^>]*${LANDMARK}`, "giu");
+const isLandmark = (range) => new RegExp(`^<[^>]*${LANDMARK}`, "iu").test(range.openTag);
+/** Navigation that is not a <nav>: any element with role="navigation", or labelled as a breadcrumb. */
+const withoutNavigationLandmarks = (html) => withoutElementsOpening(html, LANDMARK_OPENING, isLandmark);
+// hidden="until-found" is found by the browser's search and opened, so its text is the page's.
+const HIDDEN = String.raw `${ATTRIBUTES}\s+hidden(?=[\s/>=])(?!\s*=\s*(?:"until-found"|'until-found'|until-found(?=[\s/>])))`;
+const HIDDEN_OPENING = new RegExp(String.raw `<([a-z][a-z0-9-]*)${HIDDEN}`, "giu");
+const isHidden = (range) => new RegExp(String.raw `^<[a-z][a-z0-9-]*${HIDDEN}`, "iu").test(range.openTag);
+/** Elements the page does not show (the hidden attribute): a tooltip, a closed menu. */
+const withoutHiddenElements = (html) => withoutElementsOpening(html, HIDDEN_OPENING, isHidden);
+/**
+ * Buttons are controls ("Close", "Share", "Cite this publication"), not prose, except one inside a heading, which is
+ * how an accordion draws its section's title.
+ */
+const withoutButtons = (html) => {
+    const headings = headingRanges(html);
+    return withoutElementsWhere(html, "button", (button) => !headings.some((heading) => isInside(heading, button)));
+};
+const LINK_MARK = "\u0003";
+/** Links joined by ">" or another arrow, then the current page's name as plain text. */
+const BREADCRUMB_TRAIL = new RegExp(`^\\s*(?:${LINK_MARK}\\s*[>›»＞→]\\s*){2,}[^${LINK_MARK}>›»＞→]*$`, "u");
+/** A link wrapping blocks (a card with a title and a summary) carries content, not a way around the site. */
+const isCard = (link) => /<(?:div|p|h[1-6]|ul|ol|dl|section|article|figure)\b/iu.test(link);
+/** The links in html, or none when one of them wraps blocks. */
+const plainLinks = (html) => {
+    const links = [...html.matchAll(ANY_LINK)].map((link) => link[0]);
+    return links.some(isCard) ? [] : links;
+};
+/** The text with each link standing as one mark. */
+const withLinksMarked = (html) => decodeEntities(stripTags(html.replace(ANY_LINK, LINK_MARK)));
+const isBreadcrumbTrail = (html) => plainLinks(html).length > 0 && BREADCRUMB_TRAIL.test(withLinksMarked(html));
+const isOnlyLinks = (html) => hasNoWords(withLinksMarked(html).replaceAll(LINK_MARK, ""));
+const MIN_GROUP_LINKS = 2;
+/** A block made of two or more links and nothing else (a menu, previous and next), or a breadcrumb trail. */
+const isLinkGroup = (range) => {
+    const links = plainLinks(range.inner);
+    if (links.length === 0)
+        return false;
+    return isOnlyLinks(range.inner) ? links.length >= MIN_GROUP_LINKS : isBreadcrumbTrail(range.inner);
+};
+/** A block of links and nothing else, even a single one: before the page's title, a way around the site ("English"). */
+const isLinksOnly = (range) => plainLinks(range.inner).length > 0 && isOnlyLinks(range.inner);
+const withoutLinkGroups = (html) => ["div", "section", "p"].reduce((text, tag) => withoutElementsWhere(text, tag, isLinkGroup), html);
+/** The page's first <h1>: what it reads and where it opens. */
+const pageTitle = (html) => {
+    const title = /<h1\b[^>]*>([\s\S]*?)<\/h1\s*>/iu.exec(html);
+    return title === null ? undefined : { text: plainText(title[1] ?? ""), start: title.index };
+};
+const MIN_TRAIL_LINKS = 2;
+/**
+ * A breadcrumb as a list above the page's title: two or more items that are each only a link, then the title itself as
+ * the last item. The same list further down is the page's own content (steps ending in the one this page is about).
+ */
+const isBreadcrumbList = (body, at, title) => {
+    const items = body.split(/<li\b[^>]*>/iu).slice(1);
+    const current = items.at(-1);
+    const trail = items.slice(0, -1);
+    return (title !== undefined &&
+        at < title.start &&
+        current !== undefined &&
+        plainText(current) === title.text &&
+        trail.length >= MIN_TRAIL_LINKS &&
+        trail.every(isNavigation));
+};
+/** A list with items and not a word in them: buttons drawn as images, such as a text-size switch. */
+const isWordlessList = (body) => /<li\b/iu.test(body) && hasNoWords(plainText(body));
+const HREFLANG = /^<a\b[^>]*\shreflang\s*=/iu;
+/**
+ * A list of this page in other languages: every link names its language (hreflang) and opens a page, and every other
+ * item is only a label, the language being read ("English", "Español"). An item with a sentence, or a link to a file
+ * (a form in each language), makes it the document's.
+ */
+const isLanguageSwitch = (body) => {
+    const links = [...body.matchAll(ANY_LINK)].map((link) => link[0]);
+    const items = body.split(/<li\b[^>]*>/iu).slice(1);
+    return (links.length > 0 &&
+        links.every((link) => HREFLANG.test(link) && !isFileLink(link)) &&
+        items.every((item) => isNavigation(item) || !/<a\b/iu.test(item)) &&
+        !hasSentence(body));
+};
+/** From the inside out, so a nested table of contents goes too once its inner lists are gone. */
+const withoutNavigation = (html) => untilStable(html, (text) => {
+    const title = pageTitle(text);
+    return text.replace(/<(ul|ol)\b[^>]*>((?:(?!<[uo]l\b)[\s\S])*?)<\/\1\s*>/giu, (whole, _tag, body, at) => isNavigation(body) || isWordlessList(body) || isLanguageSwitch(body) || isBreadcrumbList(body, at, title) ? " " : whole);
+});
+const DEFINITION = /<dd\b[^>]*>([\s\S]*?)(?:<\/dd\s*>|(?=<d[dt]\b)|$)/giu;
+const isMenuDefinition = (definition) => isNavigation(definition) || isBreadcrumbTrail(definition);
+/**
+ * A definition list whose every definition is only links or a breadcrumb trail: a menu with its label as the term
+ * ("文字サイズ: 標準 大", "現在位置: トップ > 教育 > この頁").
+ */
+const isLabelledMenu = (list) => [...list.matchAll(DEFINITION)].every((definition) => isMenuDefinition(definition[1] ?? ""));
+const isDefinitionList = (range) => /^<dl\b/iu.test(range.openTag);
+/** What a block holds besides its menus: its link-only lists taken out, or a labelled menu's terms. */
+const besideMenus = (range) => isDefinitionList(range) && isLabelledMenu(range.inner) ? range.inner.replace(DEFINITION, " ") : withoutNavigation(range.inner);
+const holdsMenu = (range) => besideMenus(range) !== range.inner || isLinksOnly(range);
+/** Nothing but labels beside the menus: a sentence beside a menu makes the block the document's. */
+const hasNoSentenceBesideMenus = (range) => !CLOSED_SENTENCE.test(plainText(besideMenus(range)));
+const HEADER_BLOCKS = ["div", "section", "header", "dl"];
+/**
+ * The innermost block that closes before the page's title opens and holds a menu (or only links), with no sentence beside it, is part
+ * of the site's header: it goes whole, with the tagline and labels beside the menu. Only the innermost, so that text
+ * sharing an outer block with a menu block (an agency and docket number) is kept; so is a block with a sentence in it.
+ */
+const withoutSiteHeader = (html) => {
+    const title = pageTitle(html);
+    if (title === undefined)
+        return html;
+    const menus = HEADER_BLOCKS.flatMap((tag) => elementRanges(html, tag)).filter((range) => range.end <= title.start && holdsMenu(range));
+    const innermost = menus.filter((range) => !menus.some((inner) => isInside(range, inner)));
+    return withoutRanges(html, innermost.filter(hasNoSentenceBesideMenus).toSorted((left, right) => left.start - right.start));
+};
+const ICON_MENU_BLOCKS = ["div", "section"];
+const LIST = /<(ul|ol)\b[^>]*>(?:(?!<[uo]l\b)[\s\S])*?<\/\1\s*>/giu;
+const wordlessLists = (html) => [...html.matchAll(LIST)].map((list) => list[0]).filter((list) => isWordlessList(list.replace(/^<[^>]*>/u, "")));
+/** What a block holds beside its lists of icons. */
+const besideIcons = (range) => wordlessLists(range.inner).reduce((text, list) => text.replace(list, " "), range.inner);
+/**
+ * Beside its wordless lists, one line, no sentence and no heading: the label of a row of icons ("Share & print",
+ * "Follow us"). A heading beside share buttons is the page's title.
+ */
+const isIconMenu = (range) => {
+    const beside = besideIcons(range);
+    return textLines(beside).length === 1 && !hasSentence(beside) && !/<h[1-6]\b/iu.test(beside);
+};
+/**
+ * The innermost block holding a list of icons (share buttons, social links) and anything else goes whole when that is
+ * only a label; a wrapper around the list alone is passed over for the block that holds its label.
+ */
+const withoutIconMenus = (html) => {
+    const menus = ICON_MENU_BLOCKS.flatMap((tag) => elementRanges(html, tag)).filter((range) => wordlessLists(range.inner).length > 0 && textLines(besideIcons(range)).length > 0);
+    const innermost = menus.filter((range) => !menus.some((inner) => isInside(range, inner)));
+    return withoutRanges(html, innermost.filter(isIconMenu).toSorted((left, right) => left.start - right.start));
+};
+const IMAGE = /<img\b[^>]*>/giu;
+const ALT = /\salt\s*=\s*(?:"([^"]*)"|'([^']*)'|([^\s"'>]+))/iu;
+const altText = (image) => {
+    const alt = ALT.exec(image);
+    return plainText(alt?.[1] ?? alt?.[2] ?? alt?.[3] ?? "");
+};
+/** The page's <title>, which names the page wherever it is shown (a tab, a bookmark, a search result). */
+const documentTitle = (html) => plainText(/<title\b[^>]*>([\s\S]*?)<\/title\s*>/iu.exec(html)?.[1] ?? "");
+/**
+ * A heading holding images and nothing else (a logo, a banner, a title set as a picture) carries the images' alt text
+ * only when that text is the page's title, which the reader sees drawn there; otherwise it goes whole, since the
+ * converter keeps no image's alt text anywhere else either.
+ */
+const withImageHeadingsRead = (html, title) => html.replace(/<h([1-6])\b([^>]*)>([\s\S]*?)<\/h\1\s*>/giu, (whole, level, attributes, inside) => {
+    const images = inside.match(IMAGE) ?? [];
+    if (images.length === 0 || plainText(inside) !== "")
+        return whole;
+    const alt = images.map(altText).join(" ").trim();
+    return alt !== "" && alt === title ? `<h${level}${attributes}>${asMarkup(alt)}</h${level}>` : " ";
+});
+// An in-page link is wrapped in these marks so that one standing alone on its line ("Jump to main text") can be told
+// from one inside a sentence ("see Table 1"); the first is dropped, the second keeps its text.
+const LINK_START = "\u0001";
+const LINK_END = "\u0002";
+const markChromeLinks = (html) => html.replace(ANY_LINK, (link) => (isChromeLink(link) ? `${LINK_START}${stripTags(link)}${LINK_END}` : link));
+const MARKED_LINK = new RegExp(`${LINK_START}[^${LINK_END}]*${LINK_END}`, "gu");
+/**
+ * A line of in-page or paging links and at most a mark such as "▲" or "|" between them. Never a heading: a heading
+ * wrapped in a link back to the table of contents is still the section's title.
+ */
+const isChromeLinkLine = (line) => headingLevel(line) === 0 && line.includes(LINK_START) && hasNoWords(line.replace(MARKED_LINK, ""));
+// Without a year only the sign marks a notice: "(c)" alone opens an enumerated paragraph, "Copyright" alone a sentence.
+const COPYRIGHT_NOTICE = /^(?:(?:copyright\s*)?(?:©|\(c\)|copyright)\s*\d{4}\b|copyright\s*(?:©|\(c\))|©)/iu;
+/** A copyright notice at the very end of the page, where a site puts it when it has no <footer>. */
+const withoutClosingCopyright = (lines) => {
+    const last = lines.findLastIndex((line) => line !== "");
+    return last >= 0 && COPYRIGHT_NOTICE.test(lines[last] ?? "") ? withoutClosingCopyright(lines.slice(0, last)) : [...lines];
+};
+const headingLevel = (line) => /^#{1,6} /u.exec(line)?.[0].length ?? 0;
+/** A rule drawn with characters ("_____", "-----"), which Markdown reads as a thematic break: a line, not text. */
+const isDrawnRule = (line) => /^(?:([-_*=])\s*)(?:\1\s*){2,}$/u.test(line);
+/**
+ * Headings with nothing under them before the next heading of the same or a higher level: what dropped navigation left.
+ * A drawn rule under a heading is not something under it.
+ */
+const withoutEmptySections = (lines) => lines.reduceRight((state, line) => {
+    if (line === "" || isDrawnRule(line))
+        return { kept: [line, ...state.kept], nextLevel: state.nextLevel };
+    const level = headingLevel(line) - 1;
+    if (level < 0)
+        return { kept: [line, ...state.kept], nextLevel: Number.POSITIVE_INFINITY };
+    if (state.nextLevel <= level)
+        return state;
+    return { kept: [line, ...state.kept], nextLevel: level };
+}, { kept: [], nextLevel: 0 }).kept;
+const BLOCK_TAGS = new Set([
+    "p",
+    "div",
+    "ul",
+    "ol",
+    "hr",
+    "blockquote",
+    "section",
+    "article",
+    "header",
+    "footer",
+    "main",
+    "aside",
+    "dl",
+    "dt",
+    "dd",
+    "figure",
+    "figcaption",
+    "pre",
+    "body",
+    "html",
+]);
+const heading = (_whole, level, inside) => `\n\n${"#".repeat(Number(level))} ${stripTags(inside).trim()}\n\n`;
+const asLines = (html) => html
+    .replace(/<h([1-6])\b[^>]*>([\s\S]*?)<\/h\1\s*>/giu, heading)
+    .replace(/<li\b[^>]*>/giu, "\n- ")
+    .replace(/<\/li\s*>/giu, "")
+    .replace(/<br\s*\/?>/giu, "\n")
+    .replace(/<\/?([a-z][a-z0-9]*)\b[^>]*>/giu, (whole, tag) => (BLOCK_TAGS.has(tag.toLowerCase()) ? "\n\n" : whole));
+const textLines = (html) => decodeEntities(stripTags(asLines(html)))
+    .split("\n")
+    .map((line) => line.trim())
+    .filter((line) => line !== "");
+const ADDRESS = /<address\b[^>]*>[\s\S]*?<\/address\s*>/giu;
+const isOnlyNotices = (html) => {
+    const lines = textLines(html);
+    return lines.length > 0 && lines.every((line) => COPYRIGHT_NOTICE.test(line));
+};
+const ADDRESS_BLOCKS = ["div", "section"];
+/** What the innermost block around the address holds besides it, when it holds anything else. */
+const besideAddress = (html, address) => {
+    const end = address.index + address[0].length;
+    const around = ADDRESS_BLOCKS.flatMap((tag) => elementRanges(html, tag))
+        .filter((range) => range.start < address.index && end <= range.end)
+        .toSorted((left, right) => right.start - left.start)
+        .map((range) => range.inner.replace(address[0], " "));
+    return around.find((beside) => textLines(beside).length > 0);
+};
+/**
+ * An <address> that shares its block with nothing but copyright notices, and has nothing but them after it, closes
+ * the page: the site's own contact line, in a footer written without <footer>. An address beside any other text (a
+ * label such as "Send comments to:"), with anything else after it, or with nothing after it, is kept.
+ */
+const withoutClosingAddress = (html) => {
+    const last = [...html.matchAll(ADDRESS)].at(-1);
+    if (last === undefined)
+        return html;
+    const end = last.index + last[0].length;
+    const beside = besideAddress(html, last);
+    const isFooter = beside !== undefined && isOnlyNotices(beside) && isOnlyNotices(html.slice(end));
+    return isFooter ? `${html.slice(0, last.index)} ${html.slice(end)}` : html;
+};
+const CLOSING_BLOCKS = ["div", "section", "p"];
+const isClosingMenu = (text, range) => isLinksOnly(range) && !plainLinks(range.inner).some(isFileLink) && textLines(text.slice(range.end)).length === 0;
+/** Blocks of links to pages with nothing after them ("一覧に戻る"): a way around the site, as a menu is above the title. */
+const withoutClosingLinks = (html) => untilStable(html, (text) => {
+    const closing = CLOSING_BLOCKS.flatMap((tag) => elementRanges(text, tag)).filter((range) => isClosingMenu(text, range));
+    return withoutRanges(text, closing.toSorted((left, right) => left.start - right.start));
+});
+// Markup's own spacing, collapsed as HTML does; U+3000 is a character Japanese text writes (a clause number and its text), so it stays.
+const MARKUP_SPACE = /[^\S\u3000]+/gu;
+// A zero-width space or joiner draws nothing, so a line of nothing else is empty (a page-number anchor).
+const INVISIBLE_ONLY = /^[\s\p{Cf}]*$/u;
+// An <h2> with no text in it (a spacer, an emptied widget) becomes "##": a heading with nothing to read.
+const UNTITLED_HEADING = /^#{1,6}$/u;
+export const htmlToMarkdown = (html) => {
+    const uncommented = withLostParagraphTagsRestored(withAttributeMarkupEscaped(html).replace(/<!--[\s\S]*?-->|<\?[\s\S]*?\?>/gu, ""));
+    const preformatted = withPreformattedStashed(withTablesRead(DROPPED.reduce(withoutElement, withoutRubyText(uncommented))));
+    const kept = withoutFootnoteMarks(withoutHeadingSelfLinks(withoutWidgets(withoutButtons(withoutReaderOnlyText(withoutHiddenElements(mainContent(preformatted.html))))))).replace(MARKUP_SPACE, " ");
+    const content = withoutClosingLinks(withoutLinkGroups(withoutNavigation(withoutIconMenus(withoutSiteHeader(withoutNavigationLandmarks(withoutClosingAddress(kept)))))));
+    const text = decodeEntities(stripTags(asLines(markChromeLinks(withImageHeadingsRead(withRunInHeadingsRead(content), documentTitle(uncommented))))));
+    const lines = text
+        .split("\n")
+        .map((line) => (INVISIBLE_ONLY.test(line) ? "" : line.trim()))
+        .filter((line) => line !== "-" && !isChromeLinkLine(line))
+        .map((line) => line.replaceAll(LINK_START, "").replaceAll(LINK_END, ""))
+        .filter((line) => !UNTITLED_HEADING.test(line));
+    return tidyLines(withPreformattedRestored(withoutEmptySections(withoutClosingCopyright(withoutEmptySections(lines))), preformatted.blocks));
+};
+//# sourceMappingURL=html-markdown.js.map