polytypo 1.5.0 → 1.6.1

package/dist/{chunk-77CWF35X.cjs → chunk-PGHLZCGV.cjs}

@@ -2,17 +2,17 @@
 
 
 
-var _chunkJ2JVIZ6Ccjs = require('./chunk-J2JVIZ6C.cjs');
+var _chunkV6SSSZLPcjs = require('./chunk-V6SSSZLP.cjs');
 
 
 
-var _chunkEJ3BHHOQcjs = require('./chunk-EJ3BHHOQ.cjs');
+var _chunkDKUR763Gcjs = require('./chunk-DKUR763G.cjs');
 
 
 
 
 
-var _chunkSD4SLSARcjs = require('./chunk-SD4SLSAR.cjs');
+var _chunkVCD45VZHcjs = require('./chunk-VCD45VZH.cjs');
 
 // src/modes/markdown.ts
 var _micromark = require('micromark');
@@ -90,7 +90,7 @@ function updateElementStack(stack, tag, caseSensitive) {
     if (stack[stack.length - 1] === name) stack.pop();
     return;
   }
-  if (!parsed.selfClosing && _chunkJ2JVIZ6Ccjs.isSkippedElement.call(void 0, name)) stack.push(name);
+  if (!parsed.selfClosing && _chunkV6SSSZLPcjs.isSkippedElement.call(void 0, name)) stack.push(name);
 }
 function tokenize(source, extensions) {
   return _micromark.postprocess.call(void 0, 
@@ -100,12 +100,12 @@ function tokenize(source, extensions) {
 function resolveDialect(dialect) {
   if (dialect === "commonmark" || dialect === "mdx") return dialect;
   if (dialect === void 0) {
-    throw new (0, _chunkSD4SLSARcjs.PolytypoError)(
+    throw new (0, _chunkVCD45VZHcjs.PolytypoError)(
       "POLYTYPO_INVALID_DIALECT",
       'Mode "markdown" requires a dialect. Expected "commonmark" or "mdx"; there is no default.'
     );
   }
-  throw new (0, _chunkSD4SLSARcjs.PolytypoError)(
+  throw new (0, _chunkVCD45VZHcjs.PolytypoError)(
     "POLYTYPO_INVALID_DIALECT",
     `Unknown dialect "${String(dialect)}". Expected "commonmark" or "mdx".`
   );
@@ -116,7 +116,7 @@ function extensionsFor(dialect) {
   return dialect === "mdx" ? [fm, _micromarkextensiongfm.gfm.call(void 0, ), _micromarkextensionmdxjs.mdxjs.call(void 0, )] : [fm, _micromarkextensiongfm.gfm.call(void 0, )];
 }
 function markdownSpans(source, dialect) {
-  const events = _chunkJ2JVIZ6Ccjs.wrapParserErrors.call(void 0, dialect, () => tokenize(source, extensionsFor(dialect)));
+  const events = _chunkV6SSSZLPcjs.wrapParserErrors.call(void 0, dialect, () => tokenize(source, extensionsFor(dialect)));
   const spans = [];
   const elementStack = [];
   let skipDepth = 0;
@@ -139,7 +139,7 @@ function markdownSpans(source, dialect) {
     }
     if (type === "htmlFlow") {
       if (kind === "enter" && skipDepth === 0 && elementStack.length === 0) {
-        for (const span of _chunkJ2JVIZ6Ccjs.htmlFragmentSpans.call(void 0, 
+        for (const span of _chunkV6SSSZLPcjs.htmlFragmentSpans.call(void 0, 
           source.slice(token.start.offset, token.end.offset),
           token.start.offset
         )) {
@@ -158,19 +158,19 @@ function markdownSpans(source, dialect) {
 
 // src/engine/markdown-pipeline.ts
 function runMarkdownPipeline(input, options) {
-  const narrowTarget = _chunkSD4SLSARcjs.resolveNarrowTarget.call(void 0, options.narrowNbsp);
-  const planned = _chunkSD4SLSARcjs.planRules.call(void 0, options.rules);
-  const locale = _chunkSD4SLSARcjs.getLocaleData.call(void 0, options.locale);
+  const narrowTarget = _chunkVCD45VZHcjs.resolveNarrowTarget.call(void 0, options.narrowNbsp);
+  const planned = _chunkVCD45VZHcjs.planRules.call(void 0, options.rules);
+  const locale = _chunkVCD45VZHcjs.getLocaleData.call(void 0, options.locale);
   const dialect = resolveDialect(options.dialect);
   const spans = markdownSpans(input, dialect);
-  return _chunkEJ3BHHOQcjs.runOverSpans.call(void 0, input, spans, planned, locale, "markdown", narrowTarget);
+  return _chunkDKUR763Gcjs.runOverSpans.call(void 0, input, spans, planned, locale, "markdown", narrowTarget);
 }
 function analyzeMarkdownPipeline(input, options) {
-  const narrowTarget = _chunkSD4SLSARcjs.resolveNarrowTarget.call(void 0, options.narrowNbsp);
-  const planned = _chunkSD4SLSARcjs.planRules.call(void 0, options.rules);
-  const locale = _chunkSD4SLSARcjs.getLocaleData.call(void 0, options.locale);
+  const narrowTarget = _chunkVCD45VZHcjs.resolveNarrowTarget.call(void 0, options.narrowNbsp);
+  const planned = _chunkVCD45VZHcjs.planRules.call(void 0, options.rules);
+  const locale = _chunkVCD45VZHcjs.getLocaleData.call(void 0, options.locale);
   const dialect = resolveDialect(options.dialect);
-  return _chunkEJ3BHHOQcjs.analyzeOverSpans.call(void 0, 
+  return _chunkDKUR763Gcjs.analyzeOverSpans.call(void 0, 
     input,
     markdownSpans(input, dialect),
     planned,
@@ -184,4 +184,4 @@ function analyzeMarkdownPipeline(input, options) {
 
 
 exports.runMarkdownPipeline = runMarkdownPipeline; exports.analyzeMarkdownPipeline = analyzeMarkdownPipeline;
-//# sourceMappingURL=chunk-77CWF35X.cjs.map
+//# sourceMappingURL=chunk-PGHLZCGV.cjs.map

package/dist/{chunk-77CWF35X.cjs.map → chunk-PGHLZCGV.cjs.map}

@@ -1 +1 @@
-{"version":3,"sources":["/home/runner/work/polytypo-js/polytypo-js/dist/chunk-77CWF35X.cjs","../src/modes/markdown.ts","../src/engine/markdown-pipeline.ts"],"names":[],"mappings":"AAAA;AACE;AACA;AACA;AACF,wDAA6B;AAC7B;AACE;AACA;AACF,wDAA6B;AAC7B;AACE;AACA;AACA;AACA;AACF,wDAA6B;AAC7B;AACA;AChBA,sCAA+C;AAC/C,gFAA4B;AAC5B,gEAAoB;AACpB,oEAAsB;AAgBtB,IAAM,oBAAA,kBAA2C,IAAI,GAAA,CAAI;AAAA;AAAA,EAEvD,YAAA;AAAA,EACA,cAAA;AAAA;AAAA,EAEA,UAAA;AAAA;AAAA,EAEA,UAAA;AAAA,EACA,iBAAA;AAAA;AAAA;AAAA,EAGA,UAAA;AAAA,EACA,WAAA;AAAA,EACA,YAAA;AAAA;AAAA;AAAA;AAAA,EAIA,mBAAA;AAAA,EACA,mBAAA;AAAA,EACA,UAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA;AACF,CAAC,CAAA;AAGD,IAAM,oBAAA,kBAA2C,IAAI,GAAA,CAAI;AAAA,EACvD,UAAA;AAAA,EACA,eAAA;AAAA,EACA;AACF,CAAC,CAAA;AAED,IAAM,UAAA,EAAY,EAAA;AAClB,IAAM,aAAA,EAAe,EAAA;AACrB,IAAM,MAAA,EAAQ,EAAA;AACd,IAAM,YAAA,EAAc,EAAA;AACpB,IAAM,SAAA,EAAW,EAAA;AAEjB,SAAS,aAAA,CAAc,IAAA,EAAuB;AAC5C,EAAA,OAAQ,KAAA,GAAQ,GAAA,GAAQ,KAAA,GAAQ,IAAA,GAAU,KAAA,GAAQ,GAAA,GAAQ,KAAA,GAAQ,EAAA;AACpE;AAEA,SAAS,aAAA,CAAc,IAAA,EAAuB;AAI5C,EAAA,OACE,KAAA,IAAS,aAAA,GACT,KAAA,IAAS,MAAA,GACT,KAAA,IAAS,GAAA,GACT,KAAA,IAAS,EAAA,GACT,KAAA,IAAS,GAAA,GACT,KAAA,IAAS,EAAA;AAEb;AAGA,SAAS,UAAA,CAAW,KAAA,EAAuB;AACzC,EAAA,IAAI,IAAA,EAAM,EAAA;AACV,EAAA,IAAA,CAAA,IAAS,EAAA,EAAI,CAAA,EAAG,EAAA,EAAI,KAAA,CAAM,MAAA,EAAQ,EAAA,GAAK,CAAA,EAAG;AACxC,IAAA,MAAM,KAAA,EAAO,KAAA,CAAM,UAAA,CAAW,CAAC,CAAA;AAC/B,IAAA,IAAA,GAAO,KAAA,GAAQ,GAAA,GAAQ,KAAA,GAAQ,GAAA,EAAO,MAAA,CAAO,YAAA,CAAa,KAAA,EAAO,EAAI,EAAA,EAAI,KAAA,CAAM,CAAC,CAAA;AAAA,EAClF;AACA,EAAA,OAAO,GAAA;AACT;AAQA,SAAS,OAAA,CAAQ,GAAA,EAA4B;AAC3C,EAAA,GAAA,CAAI,GAAA,CAAI,UAAA,CAAW,CAAC,EAAA,IAAM,SAAA,EAAW,OAAO,IAAA;AAC5C,EAAA,IAAI,EAAA,EAAI,CAAA;AACR,EAAA,MAAM,QAAA,EAAU,GAAA,CAAI,UAAA,CAAW,CAAC,EAAA,IAAM,KAAA;AACtC,EAAA,GAAA,CAAI,OAAA,EAAS,EAAA,GAAK,CAAA;AAElB,EAAA,GAAA,CAAI,GAAA,CAAI,UAAA,CAAW,CAAC,EAAA,IAAM,YAAA,GAAe,GAAA,CAAI,UAAA,CAAW,CAAC,EAAA,IAAM,QAAA,EAAU,OAAO,IAAA;AAChF,EAAA,GAAA,CAAI,CAAC,aAAA,CAAc,GAAA,CAAI,UAAA,CAAW,CAAC,CAAC,CAAA,EAAG,OAAO,IAAA;AAC9C,EAAA,MAAM,OAAA,EAAS,CAAA;AACf,EAAA,MAAA,CAAO,EAAA,EAAI,GAAA,CAAI,OAAA,GAAU,aAAA,CAAc,GAAA,CAAI,UAAA,CAAW,CAAC,CAAC,CAAA,EAAG,EAAA,GAAK,CAAA;AAChE,EAAA,MAAM,YAAA,EAAc,GAAA,CAAI,UAAA,CAAW,GAAA,CAAI,OAAA,EAAS,CAAC,EAAA,IAAM,KAAA;AACvD,EAAA,OAAO,EAAE,IAAA,EAAM,GAAA,CAAI,KAAA,CAAM,MAAA,EAAQ,CAAC,CAAA,EAAG,OAAA,EAAS,YAAY,CAAA;AAC5D;AAQA,SAAS,kBAAA,CAAmB,KAAA,EAAiB,GAAA,EAAa,aAAA,EAA8B;AACtF,EAAA,MAAM,OAAA,EAAS,OAAA,CAAQ,GAAG,CAAA;AAC1B,EAAA,GAAA,CAAI,OAAA,IAAW,IAAA,EAAM,MAAA;AACrB,EAAA,MAAM,KAAA,EAAO,cAAA,EAAgB,MAAA,CAAO,KAAA,EAAO,UAAA,CAAW,MAAA,CAAO,IAAI,CAAA;AACjE,EAAA,GAAA,CAAI,MAAA,CAAO,OAAA,EAAS;AAClB,IAAA,GAAA,CAAI,KAAA,CAAM,KAAA,CAAM,OAAA,EAAS,CAAC,EAAA,IAAM,IAAA,EAAM,KAAA,CAAM,GAAA,CAAI,CAAA;AAChD,IAAA,MAAA;AAAA,EACF;AACA,EAAA,GAAA,CAAI,CAAC,MAAA,CAAO,YAAA,GAAe,gDAAA,IAAqB,CAAA,EAAG,KAAA,CAAM,IAAA,CAAK,IAAI,CAAA;AACpE;AAEA,SAAS,QAAA,CAAS,MAAA,EAAgB,UAAA,EAA2C;AAC3E,EAAA,OAAO,oCAAA;AAAA,IACL,8BAAA,EAAQ,UAAA,EAAY,CAAC,GAAG,UAAU,EAAE,CAAC,CAAA,CAClC,QAAA,CAAS,CAAA,CACT,KAAA,CAAM,mCAAA,CAAW,CAAE,MAAA,EAAQ,IAAA,EAAM,IAAI,CAAC;AAAA,EAC3C,CAAA;AACF;AAQO,SAAS,cAAA,CAAe,OAAA,EAAuC;AACpE,EAAA,GAAA,CAAI,QAAA,IAAY,aAAA,GAAgB,QAAA,IAAY,KAAA,EAAO,OAAO,OAAA;AAC1D,EAAA,GAAA,CAAI,QAAA,IAAY,KAAA,CAAA,EAAW;AACzB,IAAA,MAAM,IAAI,oCAAA;AAAA,MACR,0BAAA;AAAA,MACA;AAAA,IACF,CAAA;AAAA,EACF;AACA,EAAA,MAAM,IAAI,oCAAA;AAAA,IACR,0BAAA;AAAA,IACA,CAAA,iBAAA,EAAoB,MAAA,CAAO,OAAO,CAAC,CAAA,kCAAA;AAAA,EACrC,CAAA;AACF;AAKA,IAAM,oBAAA,EAAsB,CAAC,MAAA,EAAQ,MAAM,CAAA;AAE3C,SAAS,aAAA,CAAc,OAAA,EAA+B;AACpD,EAAA,MAAM,GAAA,EAAK,wDAAA,CAAa,GAAG,mBAAmB,CAAC,CAAA;AAC/C,EAAA,OAAO,QAAA,IAAY,MAAA,EAAQ,CAAC,EAAA,EAAI,wCAAA,CAAI,EAAG,4CAAA,CAAO,EAAA,EAAI,CAAC,EAAA,EAAI,wCAAA,CAAK,CAAA;AAC9D;AAEO,SAAS,aAAA,CAAc,MAAA,EAAgB,OAAA,EAA0B;AACtE,EAAA,MAAM,OAAA,EAAS,gDAAA,OAAiB,EAAS,CAAA,EAAA,GAAM,QAAA,CAAS,MAAA,EAAQ,aAAA,CAAc,OAAO,CAAC,CAAC,CAAA;AACvF,EAAA,MAAM,MAAA,EAAgB,CAAC,CAAA;AACvB,EAAA,MAAM,aAAA,EAAyB,CAAC,CAAA;AAChC,EAAA,IAAI,UAAA,EAAY,CAAA;AAEhB,EAAA,IAAA,CAAA,MAAW,CAAC,IAAA,EAAM,KAAK,EAAA,GAAK,MAAA,EAAQ;AAClC,IAAA,MAAM,KAAA,EAAO,KAAA,CAAM,IAAA;AAEnB,IAAA,GAAA,CAAI,mBAAA,CAAoB,GAAA,CAAI,IAAI,CAAA,EAAG;AACjC,MAAA,UAAA,GAAa,KAAA,IAAS,QAAA,EAAU,EAAA,EAAI,CAAA,CAAA;AACpC,MAAA,QAAA;AAAA,IACF;AAEA,IAAA,GAAA,CAAI,mBAAA,CAAoB,GAAA,CAAI,IAAI,CAAA,EAAG;AACjC,MAAA,GAAA,CAAI,KAAA,IAAS,OAAA,EAAS;AACpB,QAAA,kBAAA;AAAA,UACE,YAAA;AAAA,UACA,MAAA,CAAO,KAAA,CAAM,KAAA,CAAM,KAAA,CAAM,MAAA,EAAQ,KAAA,CAAM,GAAA,CAAI,MAAM,CAAA;AAAA,UACjD,KAAA,IAAS;AAAA,QACX,CAAA;AAAA,MACF;AACA,MAAA,UAAA,GAAa,KAAA,IAAS,QAAA,EAAU,EAAA,EAAI,CAAA,CAAA;AACpC,MAAA,QAAA;AAAA,IACF;AAIA,IAAA,GAAA,CAAI,KAAA,IAAS,UAAA,EAAY;AACvB,MAAA,GAAA,CAAI,KAAA,IAAS,QAAA,GAAW,UAAA,IAAc,EAAA,GAAK,YAAA,CAAa,OAAA,IAAW,CAAA,EAAG;AACpE,QAAA,IAAA,CAAA,MAAW,KAAA,GAAQ,iDAAA;AAAA,UACjB,MAAA,CAAO,KAAA,CAAM,KAAA,CAAM,KAAA,CAAM,MAAA,EAAQ,KAAA,CAAM,GAAA,CAAI,MAAM,CAAA;AAAA,UACjD,KAAA,CAAM,KAAA,CAAM;AAAA,QACd,CAAA,EAAG;AACD,UAAA,KAAA,CAAM,IAAA,CAAK,IAAI,CAAA;AAAA,QACjB;AAAA,MACF;AACA,MAAA,UAAA,GAAa,KAAA,IAAS,QAAA,EAAU,EAAA,EAAI,CAAA,CAAA;AACpC,MAAA,QAAA;AAAA,IACF;AAEA,IAAA,GAAA,CAAI,KAAA,IAAS,QAAA,GAAW,KAAA,IAAS,OAAA,GAAU,UAAA,IAAc,EAAA,GAAK,YAAA,CAAa,OAAA,IAAW,CAAA,EAAG;AACvF,MAAA,KAAA,CAAM,IAAA,CAAK,EAAE,KAAA,EAAO,KAAA,CAAM,KAAA,CAAM,MAAA,EAAQ,GAAA,EAAK,KAAA,CAAM,GAAA,CAAI,OAAO,CAAC,CAAA;AAAA,IACjE;AAAA,EACF;AAEA,EAAA,OAAO,KAAA;AACT;ADrDA;AACA;AE5IO,SAAS,mBAAA,CAAoB,KAAA,EAAe,OAAA,EAAmC;AACpF,EAAA,MAAM,aAAA,EAAe,mDAAA,OAAoB,CAAQ,UAAU,CAAA;AAC3D,EAAA,MAAM,QAAA,EAAU,yCAAA,OAAU,CAAQ,KAAK,CAAA;AACvC,EAAA,MAAM,OAAA,EAAS,6CAAA,OAAc,CAAQ,MAAM,CAAA;AAC3C,EAAA,MAAM,QAAA,EAAU,cAAA,CAAe,OAAA,CAAQ,OAAO,CAAA;AAC9C,EAAA,MAAM,MAAA,EAAQ,aAAA,CAAc,KAAA,EAAO,OAAO,CAAA;AAC1C,EAAA,OAAO,4CAAA,KAAa,EAAO,KAAA,EAAO,OAAA,EAAS,MAAA,EAAQ,UAAA,EAAY,YAAY,CAAA;AAC7E;AAKO,SAAS,uBAAA,CAAwB,KAAA,EAAe,OAAA,EAAqC;AAC1F,EAAA,MAAM,aAAA,EAAe,mDAAA,OAAoB,CAAQ,UAAU,CAAA;AAC3D,EAAA,MAAM,QAAA,EAAU,yCAAA,OAAU,CAAQ,KAAK,CAAA;AACvC,EAAA,MAAM,OAAA,EAAS,6CAAA,OAAc,CAAQ,MAAM,CAAA;AAC3C,EAAA,MAAM,QAAA,EAAU,cAAA,CAAe,OAAA,CAAQ,OAAO,CAAA;AAC9C,EAAA,OAAO,gDAAA;AAAA,IACL,KAAA;AAAA,IACA,aAAA,CAAc,KAAA,EAAO,OAAO,CAAA;AAAA,IAC5B,OAAA;AAAA,IACA,MAAA;AAAA,IACA,UAAA;AAAA,IACA;AAAA,EACF,CAAA;AACF;AF0IA;AACA;AACE;AACA;AACF,6GAAC","file":"/home/runner/work/polytypo-js/polytypo-js/dist/chunk-77CWF35X.cjs","sourcesContent":[null,"import { parse, postprocess, preprocess } from \"micromark\";\nimport { frontmatter } from \"micromark-extension-frontmatter\";\nimport { gfm } from \"micromark-extension-gfm\";\nimport { mdxjs } from \"micromark-extension-mdxjs\";\nimport type { Event, Extension } from \"micromark-util-types\";\nimport { PolytypoError } from \"../errors.js\";\nimport type { Dialect } from \"../types.js\";\nimport { htmlFragmentSpans, isSkippedElement } from \"./html.js\";\nimport { wrapParserErrors } from \"./parse-error.js\";\nimport type { Span } from \"./spans.js\";\n\n/**\n * spec/rules/modes.md 3.7. Each entry is skipped **whole**, including anything that looks\n * processable inside it. The complement is not enumerated: a span is emitted only for a\n * micromark `data` token, which is the tokenizer's own name for \"literal content\", so list\n * markers, emphasis delimiters, heading sequences, table padding, line endings, hard-break\n * spaces, character escapes and character references are outside every span by construction\n * rather than by a second list that could drift from the first.\n */\nconst SKIPPED_TOKEN_TYPES: ReadonlySet<string> = new Set([\n  // fenced code blocks (including the info string and the fences) and indented code blocks\n  \"codeFenced\",\n  \"codeIndented\",\n  // inline code spans, including the backticks\n  \"codeText\",\n  // autolinks `<https://…>`, and GFM's bare-URL form\n  \"autolink\",\n  \"literalAutolink\",\n  // link and image destinations and titles; the definition line of a reference link. The link\n  // *text* is a `labelText`, which is not skipped.\n  \"resource\",\n  \"reference\",\n  \"definition\",\n  // MDX: expression containers in full, and every JSX attribute — the whole tag is skipped, so\n  // attributes never surface. JSX element *children* sit outside the tag tokens and are\n  // processable.\n  \"mdxFlowExpression\",\n  \"mdxTextExpression\",\n  \"mdxjsEsm\",\n  // modes.md 3.7.3, and it matters: without it the second `---` of a YAML block reads as a setext\n  // underline, `title: Une note` becomes a paragraph, and `fr` puts a narrow no-break space in\n  // front of the colon of a machine-read metadata field. Covered by conformance fixtures\n  // en-us-markdown-{commonmark,mdx}-frontmatter, fr-markdown-{commonmark,mdx}-frontmatter-nbsp,\n  // en-us-markdown-commonmark-frontmatter-toml and -frontmatter-unterminated.\n  \"frontmatter\",\n]);\n\n/** Token types whose enter/exit also maintains the skipped-element stack. */\nconst RAW_TAG_TOKEN_TYPES: ReadonlySet<string> = new Set([\n  \"htmlText\",\n  \"mdxJsxTextTag\",\n  \"mdxJsxFlowTag\",\n]);\n\nconst LESS_THAN = 0x3c;\nconst GREATER_THAN = 0x3e;\nconst SLASH = 0x2f;\nconst EXCLAMATION = 0x21;\nconst QUESTION = 0x3f;\n\nfunction isAsciiLetter(unit: number): boolean {\n  return (unit >= 0x61 && unit <= 0x7a) || (unit >= 0x41 && unit <= 0x5a);\n}\n\nfunction isTagNameUnit(unit: number): boolean {\n  // Deliberately permissive: anything that is not a delimiter belongs to the name. Custom\n  // elements (`my-callout`) and member expressions (`Foo.Bar`) must come out whole so that they\n  // fail the skip-list membership test rather than being truncated into something that passes.\n  return (\n    unit !== GREATER_THAN &&\n    unit !== SLASH &&\n    unit !== 0x20 &&\n    unit !== 0x09 &&\n    unit !== 0x0a &&\n    unit !== 0x0d\n  );\n}\n\n/** ASCII-only, because HTML tag names are ASCII case-insensitive and `toLowerCase()` is not. */\nfunction asciiLower(value: string): string {\n  let out = \"\";\n  for (let i = 0; i < value.length; i += 1) {\n    const unit = value.charCodeAt(i);\n    out += unit >= 0x41 && unit <= 0x5a ? String.fromCharCode(unit + 0x20) : value[i];\n  }\n  return out;\n}\n\ninterface RawTag {\n  readonly name: string;\n  readonly closing: boolean;\n  readonly selfClosing: boolean;\n}\n\nfunction readTag(tag: string): RawTag | null {\n  if (tag.charCodeAt(0) !== LESS_THAN) return null;\n  let i = 1;\n  const closing = tag.charCodeAt(i) === SLASH;\n  if (closing) i += 1;\n  // Comments, declarations and processing instructions carry no element name.\n  if (tag.charCodeAt(i) === EXCLAMATION || tag.charCodeAt(i) === QUESTION) return null;\n  if (!isAsciiLetter(tag.charCodeAt(i))) return null;\n  const nameAt = i;\n  while (i < tag.length && isTagNameUnit(tag.charCodeAt(i))) i += 1;\n  const selfClosing = tag.charCodeAt(tag.length - 2) === SLASH;\n  return { name: tag.slice(nameAt, i), closing, selfClosing };\n}\n\n/**\n * modes.md 3.7: raw HTML is \"handed to the html skip list of 3.6\". Inline raw HTML reaches the\n * tokenizer as isolated tags with ordinary markdown content between them, so honouring the\n * subtree rule means tracking which skipped element is currently open. JSX names are compared\n * case-sensitively (`<Code>` is a component, `<code>` is an element); HTML names are not.\n */\nfunction updateElementStack(stack: string[], tag: string, caseSensitive: boolean): void {\n  const parsed = readTag(tag);\n  if (parsed === null) return;\n  const name = caseSensitive ? parsed.name : asciiLower(parsed.name);\n  if (parsed.closing) {\n    if (stack[stack.length - 1] === name) stack.pop();\n    return;\n  }\n  if (!parsed.selfClosing && isSkippedElement(name)) stack.push(name);\n}\n\nfunction tokenize(source: string, extensions: readonly Extension[]): Event[] {\n  return postprocess(\n    parse({ extensions: [...extensions] })\n      .document()\n      .write(preprocess()(source, null, true)),\n  );\n}\n\n/**\n * modes.md 3.7.1. The dialect is the caller's, never detected: `\"commonmark\"` is CommonMark 0.31\n * plus GFM, `\"mdx\"` is the same minus indented code blocks and `<…>` autolinks, plus JSX and\n * `{…}` expression containers. Both enable frontmatter and GFM, which 3.7.2 makes normative —\n * §4's promise that table alignment rows survive is empty unless tables are recognised at all.\n */\nexport function resolveDialect(dialect: Dialect | undefined): Dialect {\n  if (dialect === \"commonmark\" || dialect === \"mdx\") return dialect;\n  if (dialect === undefined) {\n    throw new PolytypoError(\n      \"POLYTYPO_INVALID_DIALECT\",\n      'Mode \"markdown\" requires a dialect. Expected \"commonmark\" or \"mdx\"; there is no default.',\n    );\n  }\n  throw new PolytypoError(\n    \"POLYTYPO_INVALID_DIALECT\",\n    `Unknown dialect \"${String(dialect)}\". Expected \"commonmark\" or \"mdx\".`,\n  );\n}\n\n// modes.md 3.7.3 names both frontmatter delimiters, `---` (YAML) and `+++` (TOML), and skips\n// either whole. micromark's default matter is YAML alone, so both must be asked for by name: with\n// the default, a `+++` block parses as prose and the machine-read fields inside it get typeset.\nconst FRONTMATTER_MATTERS = [\"yaml\", \"toml\"] as const;\n\nfunction extensionsFor(dialect: Dialect): Extension[] {\n  const fm = frontmatter([...FRONTMATTER_MATTERS]);\n  return dialect === \"mdx\" ? [fm, gfm(), mdxjs()] : [fm, gfm()];\n}\n\nexport function markdownSpans(source: string, dialect: Dialect): Span[] {\n  const events = wrapParserErrors(dialect, () => tokenize(source, extensionsFor(dialect)));\n  const spans: Span[] = [];\n  const elementStack: string[] = [];\n  let skipDepth = 0;\n\n  for (const [kind, token] of events) {\n    const type = token.type;\n\n    if (SKIPPED_TOKEN_TYPES.has(type)) {\n      skipDepth += kind === \"enter\" ? 1 : -1;\n      continue;\n    }\n\n    if (RAW_TAG_TOKEN_TYPES.has(type)) {\n      if (kind === \"enter\") {\n        updateElementStack(\n          elementStack,\n          source.slice(token.start.offset, token.end.offset),\n          type !== \"htmlText\",\n        );\n      }\n      skipDepth += kind === \"enter\" ? 1 : -1;\n      continue;\n    }\n\n    // An HTML block is handed to the `html` extractor rather than skipped whole, so that the\n    // prose inside `<div>…</div>` is typeset while the markup is not.\n    if (type === \"htmlFlow\") {\n      if (kind === \"enter\" && skipDepth === 0 && elementStack.length === 0) {\n        for (const span of htmlFragmentSpans(\n          source.slice(token.start.offset, token.end.offset),\n          token.start.offset,\n        )) {\n          spans.push(span);\n        }\n      }\n      skipDepth += kind === \"enter\" ? 1 : -1;\n      continue;\n    }\n\n    if (kind === \"enter\" && type === \"data\" && skipDepth === 0 && elementStack.length === 0) {\n      spans.push({ start: token.start.offset, end: token.end.offset });\n    }\n  }\n\n  return spans;\n}\n","import { markdownSpans, resolveDialect } from \"../modes/markdown.js\";\nimport type { Options } from \"../types.js\";\nimport { getLocaleData } from \"./locale.js\";\nimport { resolveNarrowTarget } from \"./narrow-target.js\";\nimport { planRules } from \"./rule-runner.js\";\nimport { analyzeOverSpans, runOverSpans } from \"./span-runner.js\";\nimport type { Change } from \"./origin.js\";\n\n/**\n * `markdown` mode only. Imports the Micromark/MDX stack and `parse5` (via `../modes/markdown.js`\n * importing `../modes/html.js` for normative embedded-HTML handling, modes.md 3.7) — both are\n * legitimately reachable from `polytypo/markdown` (AUDIT_REMEDIATION_AND_RELEASE_PLAN.md 5.1).\n *\n * Validation order — `planRules`, then `getLocaleData`, then `resolveDialect`/parsing — mirrors\n * the pre-Stage-5 aggregate `runPipeline` exactly: an unknown-rule error wins over an\n * unknown-locale error, which wins over a missing/invalid dialect, which wins over a parse\n * failure. All four are public, tested behaviour this refactor was not authorised to change.\n */\nexport function runMarkdownPipeline(input: string, options: Partial<Options>): string {\n  const narrowTarget = resolveNarrowTarget(options.narrowNbsp);\n  const planned = planRules(options.rules);\n  const locale = getLocaleData(options.locale);\n  const dialect = resolveDialect(options.dialect);\n  const spans = markdownSpans(input, dialect);\n  return runOverSpans(input, spans, planned, locale, \"markdown\", narrowTarget);\n}\n\n/** analyze.md §1, `markdown` mode. Dialect validation happens here exactly as it does for\n * `runMarkdownPipeline`, so an absent dialect throws POLYTYPO_INVALID_DIALECT from `analyze`\n * too (analyze.md §4 A1). */\nexport function analyzeMarkdownPipeline(input: string, options: Partial<Options>): Change[] {\n  const narrowTarget = resolveNarrowTarget(options.narrowNbsp);\n  const planned = planRules(options.rules);\n  const locale = getLocaleData(options.locale);\n  const dialect = resolveDialect(options.dialect);\n  return analyzeOverSpans(\n    input,\n    markdownSpans(input, dialect),\n    planned,\n    locale,\n    \"markdown\",\n    narrowTarget,\n  );\n}\n"]}
+{"version":3,"sources":["/home/runner/work/polytypo-js/polytypo-js/dist/chunk-PGHLZCGV.cjs","../src/modes/markdown.ts","../src/engine/markdown-pipeline.ts"],"names":[],"mappings":"AAAA;AACE;AACA;AACA;AACF,wDAA6B;AAC7B;AACE;AACA;AACF,wDAA6B;AAC7B;AACE;AACA;AACA;AACA;AACF,wDAA6B;AAC7B;AACA;AChBA,sCAA+C;AAC/C,gFAA4B;AAC5B,gEAAoB;AACpB,oEAAsB;AAgBtB,IAAM,oBAAA,kBAA2C,IAAI,GAAA,CAAI;AAAA;AAAA,EAEvD,YAAA;AAAA,EACA,cAAA;AAAA;AAAA,EAEA,UAAA;AAAA;AAAA,EAEA,UAAA;AAAA,EACA,iBAAA;AAAA;AAAA;AAAA,EAGA,UAAA;AAAA,EACA,WAAA;AAAA,EACA,YAAA;AAAA;AAAA;AAAA;AAAA,EAIA,mBAAA;AAAA,EACA,mBAAA;AAAA,EACA,UAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA;AACF,CAAC,CAAA;AAGD,IAAM,oBAAA,kBAA2C,IAAI,GAAA,CAAI;AAAA,EACvD,UAAA;AAAA,EACA,eAAA;AAAA,EACA;AACF,CAAC,CAAA;AAED,IAAM,UAAA,EAAY,EAAA;AAClB,IAAM,aAAA,EAAe,EAAA;AACrB,IAAM,MAAA,EAAQ,EAAA;AACd,IAAM,YAAA,EAAc,EAAA;AACpB,IAAM,SAAA,EAAW,EAAA;AAEjB,SAAS,aAAA,CAAc,IAAA,EAAuB;AAC5C,EAAA,OAAQ,KAAA,GAAQ,GAAA,GAAQ,KAAA,GAAQ,IAAA,GAAU,KAAA,GAAQ,GAAA,GAAQ,KAAA,GAAQ,EAAA;AACpE;AAEA,SAAS,aAAA,CAAc,IAAA,EAAuB;AAI5C,EAAA,OACE,KAAA,IAAS,aAAA,GACT,KAAA,IAAS,MAAA,GACT,KAAA,IAAS,GAAA,GACT,KAAA,IAAS,EAAA,GACT,KAAA,IAAS,GAAA,GACT,KAAA,IAAS,EAAA;AAEb;AAGA,SAAS,UAAA,CAAW,KAAA,EAAuB;AACzC,EAAA,IAAI,IAAA,EAAM,EAAA;AACV,EAAA,IAAA,CAAA,IAAS,EAAA,EAAI,CAAA,EAAG,EAAA,EAAI,KAAA,CAAM,MAAA,EAAQ,EAAA,GAAK,CAAA,EAAG;AACxC,IAAA,MAAM,KAAA,EAAO,KAAA,CAAM,UAAA,CAAW,CAAC,CAAA;AAC/B,IAAA,IAAA,GAAO,KAAA,GAAQ,GAAA,GAAQ,KAAA,GAAQ,GAAA,EAAO,MAAA,CAAO,YAAA,CAAa,KAAA,EAAO,EAAI,EAAA,EAAI,KAAA,CAAM,CAAC,CAAA;AAAA,EAClF;AACA,EAAA,OAAO,GAAA;AACT;AAQA,SAAS,OAAA,CAAQ,GAAA,EAA4B;AAC3C,EAAA,GAAA,CAAI,GAAA,CAAI,UAAA,CAAW,CAAC,EAAA,IAAM,SAAA,EAAW,OAAO,IAAA;AAC5C,EAAA,IAAI,EAAA,EAAI,CAAA;AACR,EAAA,MAAM,QAAA,EAAU,GAAA,CAAI,UAAA,CAAW,CAAC,EAAA,IAAM,KAAA;AACtC,EAAA,GAAA,CAAI,OAAA,EAAS,EAAA,GAAK,CAAA;AAElB,EAAA,GAAA,CAAI,GAAA,CAAI,UAAA,CAAW,CAAC,EAAA,IAAM,YAAA,GAAe,GAAA,CAAI,UAAA,CAAW,CAAC,EAAA,IAAM,QAAA,EAAU,OAAO,IAAA;AAChF,EAAA,GAAA,CAAI,CAAC,aAAA,CAAc,GAAA,CAAI,UAAA,CAAW,CAAC,CAAC,CAAA,EAAG,OAAO,IAAA;AAC9C,EAAA,MAAM,OAAA,EAAS,CAAA;AACf,EAAA,MAAA,CAAO,EAAA,EAAI,GAAA,CAAI,OAAA,GAAU,aAAA,CAAc,GAAA,CAAI,UAAA,CAAW,CAAC,CAAC,CAAA,EAAG,EAAA,GAAK,CAAA;AAChE,EAAA,MAAM,YAAA,EAAc,GAAA,CAAI,UAAA,CAAW,GAAA,CAAI,OAAA,EAAS,CAAC,EAAA,IAAM,KAAA;AACvD,EAAA,OAAO,EAAE,IAAA,EAAM,GAAA,CAAI,KAAA,CAAM,MAAA,EAAQ,CAAC,CAAA,EAAG,OAAA,EAAS,YAAY,CAAA;AAC5D;AAQA,SAAS,kBAAA,CAAmB,KAAA,EAAiB,GAAA,EAAa,aAAA,EAA8B;AACtF,EAAA,MAAM,OAAA,EAAS,OAAA,CAAQ,GAAG,CAAA;AAC1B,EAAA,GAAA,CAAI,OAAA,IAAW,IAAA,EAAM,MAAA;AACrB,EAAA,MAAM,KAAA,EAAO,cAAA,EAAgB,MAAA,CAAO,KAAA,EAAO,UAAA,CAAW,MAAA,CAAO,IAAI,CAAA;AACjE,EAAA,GAAA,CAAI,MAAA,CAAO,OAAA,EAAS;AAClB,IAAA,GAAA,CAAI,KAAA,CAAM,KAAA,CAAM,OAAA,EAAS,CAAC,EAAA,IAAM,IAAA,EAAM,KAAA,CAAM,GAAA,CAAI,CAAA;AAChD,IAAA,MAAA;AAAA,EACF;AACA,EAAA,GAAA,CAAI,CAAC,MAAA,CAAO,YAAA,GAAe,gDAAA,IAAqB,CAAA,EAAG,KAAA,CAAM,IAAA,CAAK,IAAI,CAAA;AACpE;AAEA,SAAS,QAAA,CAAS,MAAA,EAAgB,UAAA,EAA2C;AAC3E,EAAA,OAAO,oCAAA;AAAA,IACL,8BAAA,EAAQ,UAAA,EAAY,CAAC,GAAG,UAAU,EAAE,CAAC,CAAA,CAClC,QAAA,CAAS,CAAA,CACT,KAAA,CAAM,mCAAA,CAAW,CAAE,MAAA,EAAQ,IAAA,EAAM,IAAI,CAAC;AAAA,EAC3C,CAAA;AACF;AAQO,SAAS,cAAA,CAAe,OAAA,EAAuC;AACpE,EAAA,GAAA,CAAI,QAAA,IAAY,aAAA,GAAgB,QAAA,IAAY,KAAA,EAAO,OAAO,OAAA;AAC1D,EAAA,GAAA,CAAI,QAAA,IAAY,KAAA,CAAA,EAAW;AACzB,IAAA,MAAM,IAAI,oCAAA;AAAA,MACR,0BAAA;AAAA,MACA;AAAA,IACF,CAAA;AAAA,EACF;AACA,EAAA,MAAM,IAAI,oCAAA;AAAA,IACR,0BAAA;AAAA,IACA,CAAA,iBAAA,EAAoB,MAAA,CAAO,OAAO,CAAC,CAAA,kCAAA;AAAA,EACrC,CAAA;AACF;AAKA,IAAM,oBAAA,EAAsB,CAAC,MAAA,EAAQ,MAAM,CAAA;AAE3C,SAAS,aAAA,CAAc,OAAA,EAA+B;AACpD,EAAA,MAAM,GAAA,EAAK,wDAAA,CAAa,GAAG,mBAAmB,CAAC,CAAA;AAC/C,EAAA,OAAO,QAAA,IAAY,MAAA,EAAQ,CAAC,EAAA,EAAI,wCAAA,CAAI,EAAG,4CAAA,CAAO,EAAA,EAAI,CAAC,EAAA,EAAI,wCAAA,CAAK,CAAA;AAC9D;AAEO,SAAS,aAAA,CAAc,MAAA,EAAgB,OAAA,EAA0B;AACtE,EAAA,MAAM,OAAA,EAAS,gDAAA,OAAiB,EAAS,CAAA,EAAA,GAAM,QAAA,CAAS,MAAA,EAAQ,aAAA,CAAc,OAAO,CAAC,CAAC,CAAA;AACvF,EAAA,MAAM,MAAA,EAAgB,CAAC,CAAA;AACvB,EAAA,MAAM,aAAA,EAAyB,CAAC,CAAA;AAChC,EAAA,IAAI,UAAA,EAAY,CAAA;AAEhB,EAAA,IAAA,CAAA,MAAW,CAAC,IAAA,EAAM,KAAK,EAAA,GAAK,MAAA,EAAQ;AAClC,IAAA,MAAM,KAAA,EAAO,KAAA,CAAM,IAAA;AAEnB,IAAA,GAAA,CAAI,mBAAA,CAAoB,GAAA,CAAI,IAAI,CAAA,EAAG;AACjC,MAAA,UAAA,GAAa,KAAA,IAAS,QAAA,EAAU,EAAA,EAAI,CAAA,CAAA;AACpC,MAAA,QAAA;AAAA,IACF;AAEA,IAAA,GAAA,CAAI,mBAAA,CAAoB,GAAA,CAAI,IAAI,CAAA,EAAG;AACjC,MAAA,GAAA,CAAI,KAAA,IAAS,OAAA,EAAS;AACpB,QAAA,kBAAA;AAAA,UACE,YAAA;AAAA,UACA,MAAA,CAAO,KAAA,CAAM,KAAA,CAAM,KAAA,CAAM,MAAA,EAAQ,KAAA,CAAM,GAAA,CAAI,MAAM,CAAA;AAAA,UACjD,KAAA,IAAS;AAAA,QACX,CAAA;AAAA,MACF;AACA,MAAA,UAAA,GAAa,KAAA,IAAS,QAAA,EAAU,EAAA,EAAI,CAAA,CAAA;AACpC,MAAA,QAAA;AAAA,IACF;AAIA,IAAA,GAAA,CAAI,KAAA,IAAS,UAAA,EAAY;AACvB,MAAA,GAAA,CAAI,KAAA,IAAS,QAAA,GAAW,UAAA,IAAc,EAAA,GAAK,YAAA,CAAa,OAAA,IAAW,CAAA,EAAG;AACpE,QAAA,IAAA,CAAA,MAAW,KAAA,GAAQ,iDAAA;AAAA,UACjB,MAAA,CAAO,KAAA,CAAM,KAAA,CAAM,KAAA,CAAM,MAAA,EAAQ,KAAA,CAAM,GAAA,CAAI,MAAM,CAAA;AAAA,UACjD,KAAA,CAAM,KAAA,CAAM;AAAA,QACd,CAAA,EAAG;AACD,UAAA,KAAA,CAAM,IAAA,CAAK,IAAI,CAAA;AAAA,QACjB;AAAA,MACF;AACA,MAAA,UAAA,GAAa,KAAA,IAAS,QAAA,EAAU,EAAA,EAAI,CAAA,CAAA;AACpC,MAAA,QAAA;AAAA,IACF;AAEA,IAAA,GAAA,CAAI,KAAA,IAAS,QAAA,GAAW,KAAA,IAAS,OAAA,GAAU,UAAA,IAAc,EAAA,GAAK,YAAA,CAAa,OAAA,IAAW,CAAA,EAAG;AACvF,MAAA,KAAA,CAAM,IAAA,CAAK,EAAE,KAAA,EAAO,KAAA,CAAM,KAAA,CAAM,MAAA,EAAQ,GAAA,EAAK,KAAA,CAAM,GAAA,CAAI,OAAO,CAAC,CAAA;AAAA,IACjE;AAAA,EACF;AAEA,EAAA,OAAO,KAAA;AACT;ADrDA;AACA;AE5IO,SAAS,mBAAA,CAAoB,KAAA,EAAe,OAAA,EAAmC;AACpF,EAAA,MAAM,aAAA,EAAe,mDAAA,OAAoB,CAAQ,UAAU,CAAA;AAC3D,EAAA,MAAM,QAAA,EAAU,yCAAA,OAAU,CAAQ,KAAK,CAAA;AACvC,EAAA,MAAM,OAAA,EAAS,6CAAA,OAAc,CAAQ,MAAM,CAAA;AAC3C,EAAA,MAAM,QAAA,EAAU,cAAA,CAAe,OAAA,CAAQ,OAAO,CAAA;AAC9C,EAAA,MAAM,MAAA,EAAQ,aAAA,CAAc,KAAA,EAAO,OAAO,CAAA;AAC1C,EAAA,OAAO,4CAAA,KAAa,EAAO,KAAA,EAAO,OAAA,EAAS,MAAA,EAAQ,UAAA,EAAY,YAAY,CAAA;AAC7E;AAKO,SAAS,uBAAA,CAAwB,KAAA,EAAe,OAAA,EAAqC;AAC1F,EAAA,MAAM,aAAA,EAAe,mDAAA,OAAoB,CAAQ,UAAU,CAAA;AAC3D,EAAA,MAAM,QAAA,EAAU,yCAAA,OAAU,CAAQ,KAAK,CAAA;AACvC,EAAA,MAAM,OAAA,EAAS,6CAAA,OAAc,CAAQ,MAAM,CAAA;AAC3C,EAAA,MAAM,QAAA,EAAU,cAAA,CAAe,OAAA,CAAQ,OAAO,CAAA;AAC9C,EAAA,OAAO,gDAAA;AAAA,IACL,KAAA;AAAA,IACA,aAAA,CAAc,KAAA,EAAO,OAAO,CAAA;AAAA,IAC5B,OAAA;AAAA,IACA,MAAA;AAAA,IACA,UAAA;AAAA,IACA;AAAA,EACF,CAAA;AACF;AF0IA;AACA;AACE;AACA;AACF,6GAAC","file":"/home/runner/work/polytypo-js/polytypo-js/dist/chunk-PGHLZCGV.cjs","sourcesContent":[null,"import { parse, postprocess, preprocess } from \"micromark\";\nimport { frontmatter } from \"micromark-extension-frontmatter\";\nimport { gfm } from \"micromark-extension-gfm\";\nimport { mdxjs } from \"micromark-extension-mdxjs\";\nimport type { Event, Extension } from \"micromark-util-types\";\nimport { PolytypoError } from \"../errors.js\";\nimport type { Dialect } from \"../types.js\";\nimport { htmlFragmentSpans, isSkippedElement } from \"./html.js\";\nimport { wrapParserErrors } from \"./parse-error.js\";\nimport type { Span } from \"./spans.js\";\n\n/**\n * spec/rules/modes.md 3.7. Each entry is skipped **whole**, including anything that looks\n * processable inside it. The complement is not enumerated: a span is emitted only for a\n * micromark `data` token, which is the tokenizer's own name for \"literal content\", so list\n * markers, emphasis delimiters, heading sequences, table padding, line endings, hard-break\n * spaces, character escapes and character references are outside every span by construction\n * rather than by a second list that could drift from the first.\n */\nconst SKIPPED_TOKEN_TYPES: ReadonlySet<string> = new Set([\n  // fenced code blocks (including the info string and the fences) and indented code blocks\n  \"codeFenced\",\n  \"codeIndented\",\n  // inline code spans, including the backticks\n  \"codeText\",\n  // autolinks `<https://…>`, and GFM's bare-URL form\n  \"autolink\",\n  \"literalAutolink\",\n  // link and image destinations and titles; the definition line of a reference link. The link\n  // *text* is a `labelText`, which is not skipped.\n  \"resource\",\n  \"reference\",\n  \"definition\",\n  // MDX: expression containers in full, and every JSX attribute — the whole tag is skipped, so\n  // attributes never surface. JSX element *children* sit outside the tag tokens and are\n  // processable.\n  \"mdxFlowExpression\",\n  \"mdxTextExpression\",\n  \"mdxjsEsm\",\n  // modes.md 3.7.3, and it matters: without it the second `---` of a YAML block reads as a setext\n  // underline, `title: Une note` becomes a paragraph, and `fr` puts a narrow no-break space in\n  // front of the colon of a machine-read metadata field. Covered by conformance fixtures\n  // en-us-markdown-{commonmark,mdx}-frontmatter, fr-markdown-{commonmark,mdx}-frontmatter-nbsp,\n  // en-us-markdown-commonmark-frontmatter-toml and -frontmatter-unterminated.\n  \"frontmatter\",\n]);\n\n/** Token types whose enter/exit also maintains the skipped-element stack. */\nconst RAW_TAG_TOKEN_TYPES: ReadonlySet<string> = new Set([\n  \"htmlText\",\n  \"mdxJsxTextTag\",\n  \"mdxJsxFlowTag\",\n]);\n\nconst LESS_THAN = 0x3c;\nconst GREATER_THAN = 0x3e;\nconst SLASH = 0x2f;\nconst EXCLAMATION = 0x21;\nconst QUESTION = 0x3f;\n\nfunction isAsciiLetter(unit: number): boolean {\n  return (unit >= 0x61 && unit <= 0x7a) || (unit >= 0x41 && unit <= 0x5a);\n}\n\nfunction isTagNameUnit(unit: number): boolean {\n  // Deliberately permissive: anything that is not a delimiter belongs to the name. Custom\n  // elements (`my-callout`) and member expressions (`Foo.Bar`) must come out whole so that they\n  // fail the skip-list membership test rather than being truncated into something that passes.\n  return (\n    unit !== GREATER_THAN &&\n    unit !== SLASH &&\n    unit !== 0x20 &&\n    unit !== 0x09 &&\n    unit !== 0x0a &&\n    unit !== 0x0d\n  );\n}\n\n/** ASCII-only, because HTML tag names are ASCII case-insensitive and `toLowerCase()` is not. */\nfunction asciiLower(value: string): string {\n  let out = \"\";\n  for (let i = 0; i < value.length; i += 1) {\n    const unit = value.charCodeAt(i);\n    out += unit >= 0x41 && unit <= 0x5a ? String.fromCharCode(unit + 0x20) : value[i];\n  }\n  return out;\n}\n\ninterface RawTag {\n  readonly name: string;\n  readonly closing: boolean;\n  readonly selfClosing: boolean;\n}\n\nfunction readTag(tag: string): RawTag | null {\n  if (tag.charCodeAt(0) !== LESS_THAN) return null;\n  let i = 1;\n  const closing = tag.charCodeAt(i) === SLASH;\n  if (closing) i += 1;\n  // Comments, declarations and processing instructions carry no element name.\n  if (tag.charCodeAt(i) === EXCLAMATION || tag.charCodeAt(i) === QUESTION) return null;\n  if (!isAsciiLetter(tag.charCodeAt(i))) return null;\n  const nameAt = i;\n  while (i < tag.length && isTagNameUnit(tag.charCodeAt(i))) i += 1;\n  const selfClosing = tag.charCodeAt(tag.length - 2) === SLASH;\n  return { name: tag.slice(nameAt, i), closing, selfClosing };\n}\n\n/**\n * modes.md 3.7: raw HTML is \"handed to the html skip list of 3.6\". Inline raw HTML reaches the\n * tokenizer as isolated tags with ordinary markdown content between them, so honouring the\n * subtree rule means tracking which skipped element is currently open. JSX names are compared\n * case-sensitively (`<Code>` is a component, `<code>` is an element); HTML names are not.\n */\nfunction updateElementStack(stack: string[], tag: string, caseSensitive: boolean): void {\n  const parsed = readTag(tag);\n  if (parsed === null) return;\n  const name = caseSensitive ? parsed.name : asciiLower(parsed.name);\n  if (parsed.closing) {\n    if (stack[stack.length - 1] === name) stack.pop();\n    return;\n  }\n  if (!parsed.selfClosing && isSkippedElement(name)) stack.push(name);\n}\n\nfunction tokenize(source: string, extensions: readonly Extension[]): Event[] {\n  return postprocess(\n    parse({ extensions: [...extensions] })\n      .document()\n      .write(preprocess()(source, null, true)),\n  );\n}\n\n/**\n * modes.md 3.7.1. The dialect is the caller's, never detected: `\"commonmark\"` is CommonMark 0.31\n * plus GFM, `\"mdx\"` is the same minus indented code blocks and `<…>` autolinks, plus JSX and\n * `{…}` expression containers. Both enable frontmatter and GFM, which 3.7.2 makes normative —\n * §4's promise that table alignment rows survive is empty unless tables are recognised at all.\n */\nexport function resolveDialect(dialect: Dialect | undefined): Dialect {\n  if (dialect === \"commonmark\" || dialect === \"mdx\") return dialect;\n  if (dialect === undefined) {\n    throw new PolytypoError(\n      \"POLYTYPO_INVALID_DIALECT\",\n      'Mode \"markdown\" requires a dialect. Expected \"commonmark\" or \"mdx\"; there is no default.',\n    );\n  }\n  throw new PolytypoError(\n    \"POLYTYPO_INVALID_DIALECT\",\n    `Unknown dialect \"${String(dialect)}\". Expected \"commonmark\" or \"mdx\".`,\n  );\n}\n\n// modes.md 3.7.3 names both frontmatter delimiters, `---` (YAML) and `+++` (TOML), and skips\n// either whole. micromark's default matter is YAML alone, so both must be asked for by name: with\n// the default, a `+++` block parses as prose and the machine-read fields inside it get typeset.\nconst FRONTMATTER_MATTERS = [\"yaml\", \"toml\"] as const;\n\nfunction extensionsFor(dialect: Dialect): Extension[] {\n  const fm = frontmatter([...FRONTMATTER_MATTERS]);\n  return dialect === \"mdx\" ? [fm, gfm(), mdxjs()] : [fm, gfm()];\n}\n\nexport function markdownSpans(source: string, dialect: Dialect): Span[] {\n  const events = wrapParserErrors(dialect, () => tokenize(source, extensionsFor(dialect)));\n  const spans: Span[] = [];\n  const elementStack: string[] = [];\n  let skipDepth = 0;\n\n  for (const [kind, token] of events) {\n    const type = token.type;\n\n    if (SKIPPED_TOKEN_TYPES.has(type)) {\n      skipDepth += kind === \"enter\" ? 1 : -1;\n      continue;\n    }\n\n    if (RAW_TAG_TOKEN_TYPES.has(type)) {\n      if (kind === \"enter\") {\n        updateElementStack(\n          elementStack,\n          source.slice(token.start.offset, token.end.offset),\n          type !== \"htmlText\",\n        );\n      }\n      skipDepth += kind === \"enter\" ? 1 : -1;\n      continue;\n    }\n\n    // An HTML block is handed to the `html` extractor rather than skipped whole, so that the\n    // prose inside `<div>…</div>` is typeset while the markup is not.\n    if (type === \"htmlFlow\") {\n      if (kind === \"enter\" && skipDepth === 0 && elementStack.length === 0) {\n        for (const span of htmlFragmentSpans(\n          source.slice(token.start.offset, token.end.offset),\n          token.start.offset,\n        )) {\n          spans.push(span);\n        }\n      }\n      skipDepth += kind === \"enter\" ? 1 : -1;\n      continue;\n    }\n\n    if (kind === \"enter\" && type === \"data\" && skipDepth === 0 && elementStack.length === 0) {\n      spans.push({ start: token.start.offset, end: token.end.offset });\n    }\n  }\n\n  return spans;\n}\n","import { markdownSpans, resolveDialect } from \"../modes/markdown.js\";\nimport type { Options } from \"../types.js\";\nimport { getLocaleData } from \"./locale.js\";\nimport { resolveNarrowTarget } from \"./narrow-target.js\";\nimport { planRules } from \"./rule-runner.js\";\nimport { analyzeOverSpans, runOverSpans } from \"./span-runner.js\";\nimport type { Change } from \"./origin.js\";\n\n/**\n * `markdown` mode only. Imports the Micromark/MDX stack and `parse5` (via `../modes/markdown.js`\n * importing `../modes/html.js` for normative embedded-HTML handling, modes.md 3.7) — both are\n * legitimately reachable from `polytypo/markdown` (AUDIT_REMEDIATION_AND_RELEASE_PLAN.md 5.1).\n *\n * Validation order — `planRules`, then `getLocaleData`, then `resolveDialect`/parsing — mirrors\n * the pre-Stage-5 aggregate `runPipeline` exactly: an unknown-rule error wins over an\n * unknown-locale error, which wins over a missing/invalid dialect, which wins over a parse\n * failure. All four are public, tested behaviour this refactor was not authorised to change.\n */\nexport function runMarkdownPipeline(input: string, options: Partial<Options>): string {\n  const narrowTarget = resolveNarrowTarget(options.narrowNbsp);\n  const planned = planRules(options.rules);\n  const locale = getLocaleData(options.locale);\n  const dialect = resolveDialect(options.dialect);\n  const spans = markdownSpans(input, dialect);\n  return runOverSpans(input, spans, planned, locale, \"markdown\", narrowTarget);\n}\n\n/** analyze.md §1, `markdown` mode. Dialect validation happens here exactly as it does for\n * `runMarkdownPipeline`, so an absent dialect throws POLYTYPO_INVALID_DIALECT from `analyze`\n * too (analyze.md §4 A1). */\nexport function analyzeMarkdownPipeline(input: string, options: Partial<Options>): Change[] {\n  const narrowTarget = resolveNarrowTarget(options.narrowNbsp);\n  const planned = planRules(options.rules);\n  const locale = getLocaleData(options.locale);\n  const dialect = resolveDialect(options.dialect);\n  return analyzeOverSpans(\n    input,\n    markdownSpans(input, dialect),\n    planned,\n    locale,\n    \"markdown\",\n    narrowTarget,\n  );\n}\n"]}

package/dist/{chunk-ZZ3OK5SL.js → chunk-URE2U73R.js}

@@ -1,13 +1,13 @@
 import {
   analyzeOverSpans,
   runOverSpans
-} from "./chunk-OBE7MMRU.js";
+} from "./chunk-LDXWXCGL.js";
 import {
   PolytypoError,
   getLocaleData,
   planRules,
   resolveNarrowTarget
-} from "./chunk-ETKYEVE6.js";
+} from "./chunk-W77K6K6G.js";
 
 // src/modes/yaml.ts
 var TAB = 9;
@@ -261,4 +261,4 @@ export {
   runYamlPipeline,
   analyzeYamlPipeline
 };
-//# sourceMappingURL=chunk-ZZ3OK5SL.js.map
+//# sourceMappingURL=chunk-URE2U73R.js.map

package/dist/{chunk-J2JVIZ6C.cjs → chunk-V6SSSZLP.cjs}

@@ -1,6 +1,6 @@
 "use strict";Object.defineProperty(exports, "__esModule", {value: true});
 
-var _chunkSD4SLSARcjs = require('./chunk-SD4SLSAR.cjs');
+var _chunkVCD45VZHcjs = require('./chunk-VCD45VZH.cjs');
 
 // src/modes/html.ts
 var _parse5 = require('parse5');
@@ -20,8 +20,8 @@ function wrapParserErrors(what, run) {
   try {
     return run();
   } catch (error) {
-    if (error instanceof _chunkSD4SLSARcjs.PolytypoError) throw error;
-    throw new (0, _chunkSD4SLSARcjs.PolytypoError)(
+    if (error instanceof _chunkVCD45VZHcjs.PolytypoError) throw error;
+    throw new (0, _chunkVCD45VZHcjs.PolytypoError)(
       "POLYTYPO_MALFORMED_INPUT",
       `Input does not parse as ${what}: ${describe(error)}`
     );
@@ -144,4 +144,4 @@ function htmlFragmentSpans(source, offset) {
 
 
 exports.wrapParserErrors = wrapParserErrors; exports.isSkippedElement = isSkippedElement; exports.htmlSpans = htmlSpans; exports.htmlFragmentSpans = htmlFragmentSpans;
-//# sourceMappingURL=chunk-J2JVIZ6C.cjs.map
+//# sourceMappingURL=chunk-V6SSSZLP.cjs.map

package/dist/{chunk-J2JVIZ6C.cjs.map → chunk-V6SSSZLP.cjs.map}

@@ -1 +1 @@
-{"version":3,"sources":["/home/runner/work/polytypo-js/polytypo-js/dist/chunk-J2JVIZ6C.cjs","../src/modes/html.ts","../src/modes/parse-error.ts"],"names":[],"mappings":"AAAA;AACE;AACF,wDAA6B;AAC7B;AACA;ACJA,gCAAqC;ADMrC;AACA;AEAA,SAAS,QAAA,CAAS,KAAA,EAAwB;AACxC,EAAA,GAAA,CAAI,OAAO,MAAA,IAAU,SAAA,GAAY,MAAA,IAAU,IAAA,EAAM,OAAO,MAAA,CAAO,KAAK,CAAA;AACpE,EAAA,MAAM,WAAA,EAAa,KAAA;AACnB,EAAA,MAAM,OAAA,EACJ,OAAO,UAAA,CAAW,OAAA,IAAW,SAAA,EAAW,UAAA,CAAW,OAAA,EAAU,KAAA,CAAgB,OAAA;AAC/E,EAAA,MAAM,MAAA,EAAQ,UAAA,CAAW,KAAA;AACzB,EAAA,GAAA,CACE,MAAA,IAAU,KAAA,GACV,MAAA,IAAU,KAAA,EAAA,GACV,OAAO,KAAA,CAAM,KAAA,IAAS,SAAA,GACtB,OAAO,KAAA,CAAM,OAAA,IAAW,QAAA,EACxB;AACA,IAAA,OAAO,CAAA,EAAA;AACT,EAAA;AACO,EAAA;AACT;AAcgB;AACV,EAAA;AACK,IAAA;AACA,EAAA;AACH,IAAA;AACE,IAAA;AACJ,MAAA;AACA,MAAA;AACF,IAAA;AACF,EAAA;AACF;AFjBY;AACA;ACfN;AACJ,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACD;AAEe;AACP,EAAA;AACT;AAEM;AACA;AACO;AACP;AACA;AAGA;AAEG;AACA,EAAA;AACT;AAES;AACA,EAAA;AACT;AAES;AACA,EAAA;AACT;AAQS;AACH,EAAA;AACI,EAAA;AACC,EAAA;AAEL,EAAA;AACG,IAAA;AACC,IAAA;AACA,IAAA;AACG,IAAA;AACH,IAAA;AAEJ,IAAA;AAGK,MAAA;AACP,IAAA;AACI,IAAA;AACG,IAAA;AACT,EAAA;AAEM,EAAA;AAEJ,EAAA;AAIK,IAAA;AACP,EAAA;AACU,EAAA;AACH,EAAA;AACT;AAOgB;AACR,EAAA;AACF,EAAA;AACI,EAAA;AACD,EAAA;AACD,IAAA;AACI,MAAA;AACF,MAAA;AACE,QAAA;AACJ,QAAA;AACI,QAAA;AACJ,QAAA;AACF,MAAA;AACF,IAAA;AACK,IAAA;AACP,EAAA;AACU,EAAA;AACH,EAAA;AACT;AAES;AACA,EAAA;AACT;AAES;AACE,EAAA;AACD,IAAA;AACF,IAAA;AACJ,IAAA;AACQ,MAAA;AACR,IAAA;AACA,IAAA;AACF,EAAA;AAGS,EAAA;AAEL,EAAA;AAGK,EAAA;AACC,IAAA;AACR,IAAA;AACF,EAAA;AAEK,EAAA;AACL,EAAA;AACF;AAQgB;AACR,EAAA;AACA,EAAA;AACE,EAAA;AACD,EAAA;AACT;AAGgB;AACR,EAAA;AAA4B,IAAA;AAChC,IAAA;AACF,EAAA;AACM,EAAA;AACE,EAAA;AACD,EAAA;AAGT;AD9BY;AACA;AACA;AACA;AACA;AACA;AACA","file":"/home/runner/work/polytypo-js/polytypo-js/dist/chunk-J2JVIZ6C.cjs","sourcesContent":[null,"import { parse, parseFragment } from \"parse5\";\nimport type { DefaultTreeAdapterMap } from \"parse5\";\nimport { wrapParserErrors } from \"./parse-error.js\";\nimport type { Span } from \"./spans.js\";\n\ntype Node = DefaultTreeAdapterMap[\"node\"];\ntype ParentNode = DefaultTreeAdapterMap[\"parentNode\"];\n\n/**\n * spec/rules/modes.md 3.6, exhaustive and **closed**: extending it is a spec change, not an\n * implementation decision. `svg` and `math` are here because in MathML a quotation mark, a\n * hyphen and a prime are operators and identifiers — substituting a curly glyph changes what\n * the expression means. Everything not listed is processable, **including unknown and custom\n * elements**: guessing from a tag name is exactly the heuristic that diverges across runtimes.\n */\nconst SKIPPED_ELEMENTS: ReadonlySet<string> = new Set([\n  \"code\",\n  \"pre\",\n  \"kbd\",\n  \"samp\",\n  \"var\",\n  \"script\",\n  \"style\",\n  \"textarea\",\n  \"svg\",\n  \"math\",\n]);\n\nexport function isSkippedElement(tagName: string): boolean {\n  return SKIPPED_ELEMENTS.has(tagName);\n}\n\nconst AMPERSAND = 0x26;\nconst SEMICOLON = 0x3b;\nconst HASH = 0x23;\nconst LOWER_X = 0x78;\nconst UPPER_X = 0x58;\n\n/** Longest named reference in the WHATWG table is 32 characters; the guard is a bound, not a rule. */\nconst MAX_REFERENCE_BODY = 34;\n\nfunction isAsciiDigit(unit: number): boolean {\n  return unit >= 0x30 && unit <= 0x39;\n}\n\nfunction isAsciiHexDigit(unit: number): boolean {\n  return isAsciiDigit(unit) || (unit >= 0x61 && unit <= 0x66) || (unit >= 0x41 && unit <= 0x46);\n}\n\nfunction isAsciiAlphanumeric(unit: number): boolean {\n  return isAsciiDigit(unit) || (unit >= 0x61 && unit <= 0x7a) || (unit >= 0x41 && unit <= 0x5a);\n}\n\n/**\n * The end offset of a well-formed character reference starting at `at`, or -1. A bare `&` that a\n * parser would repair is deliberately **not** treated as a reference: it stays inside the span,\n * where no rule can emit or delete it, rather than manufacturing a boundary that suppresses\n * conversions around it.\n */\nfunction characterReferenceEnd(source: string, at: number, limit: number): number {\n  if (source.charCodeAt(at) !== AMPERSAND) return -1;\n  let i = at + 1;\n  if (i >= limit) return -1;\n\n  if (source.charCodeAt(i) === HASH) {\n    i += 1;\n    const unit = i < limit ? source.charCodeAt(i) : -1;\n    const hex = unit === LOWER_X || unit === UPPER_X;\n    if (hex) i += 1;\n    const digitsAt = i;\n    while (\n      i < limit &&\n      (hex ? isAsciiHexDigit(source.charCodeAt(i)) : isAsciiDigit(source.charCodeAt(i)))\n    ) {\n      i += 1;\n    }\n    if (i === digitsAt) return -1;\n    return i < limit && source.charCodeAt(i) === SEMICOLON ? i + 1 : -1;\n  }\n\n  const nameAt = i;\n  while (\n    i < limit &&\n    i - nameAt < MAX_REFERENCE_BODY &&\n    isAsciiAlphanumeric(source.charCodeAt(i))\n  ) {\n    i += 1;\n  }\n  if (i === nameAt) return -1;\n  return i < limit && source.charCodeAt(i) === SEMICOLON ? i + 1 : -1;\n}\n\n/**\n * modes.md 3.6: every character reference is skipped as an **opaque unit**, and a text node\n * containing one is split into spans around it. That is the only way `&nbsp;` survives as\n * `&nbsp;` instead of becoming a literal U+00A0 on the way out.\n */\nexport function splitCharacterReferences(source: string, start: number, end: number): Span[] {\n  const spans: Span[] = [];\n  let cursor = start;\n  let i = start;\n  while (i < end) {\n    if (source.charCodeAt(i) === AMPERSAND) {\n      const stop = characterReferenceEnd(source, i, end);\n      if (stop > 0) {\n        if (i > cursor) spans.push({ start: cursor, end: i });\n        cursor = stop;\n        i = stop;\n        continue;\n      }\n    }\n    i += 1;\n  }\n  if (end > cursor) spans.push({ start: cursor, end });\n  return spans;\n}\n\nfunction isParent(node: Node): node is ParentNode {\n  return \"childNodes\" in node;\n}\n\nfunction collect(node: Node, source: string, spans: Span[]): void {\n  if (node.nodeName === \"#text\") {\n    const location = node.sourceCodeLocation;\n    if (location === undefined || location === null) return;\n    for (const span of splitCharacterReferences(source, location.startOffset, location.endOffset)) {\n      spans.push(span);\n    }\n    return;\n  }\n\n  // Comments, doctype and processing instructions never contribute a span (modes.md 3.6).\n  if (node.nodeName === \"#comment\" || node.nodeName === \"#documentType\") return;\n\n  if (\"tagName\" in node && isSkippedElement(node.tagName)) return;\n\n  // A `template`'s children live in a separate document fragment; `template` is not skipped.\n  if (node.nodeName === \"template\" && \"content\" in node) {\n    collect(node.content, source, spans);\n    return;\n  }\n\n  if (!isParent(node)) return;\n  for (const child of node.childNodes) collect(child, source, spans);\n}\n\n/**\n * Locate the processable spans of an HTML document. The tree is used only to find offsets and is\n * then discarded — modes.md 4 forbids reserialisation, which is what makes attribute quoting,\n * self-closing forms, entity spelling, tag case and malformed-input recovery preserved by\n * construction rather than by effort.\n */\nexport function htmlSpans(source: string): Span[] {\n  const document = wrapParserErrors(\"HTML\", () => parse(source, { sourceCodeLocationInfo: true }));\n  const spans: Span[] = [];\n  collect(document, source, spans);\n  return spans;\n}\n\n/** The same, for a fragment: used for HTML blocks embedded in markdown (modes.md 3.7). */\nexport function htmlFragmentSpans(source: string, offset: number): Span[] {\n  const fragment = wrapParserErrors(\"HTML\", () =>\n    parseFragment(source, { sourceCodeLocationInfo: true }),\n  );\n  const spans: Span[] = [];\n  collect(fragment, source, spans);\n  return offset === 0\n    ? spans\n    : spans.map((span) => ({ start: span.start + offset, end: span.end + offset }));\n}\n","import { PolytypoError } from \"../errors.js\";\n\ninterface PositionedError {\n  readonly reason?: unknown;\n  readonly place?: { readonly line?: unknown; readonly column?: unknown } | null;\n}\n\nfunction describe(error: unknown): string {\n  if (typeof error !== \"object\" || error === null) return String(error);\n  const positioned = error as PositionedError;\n  const reason =\n    typeof positioned.reason === \"string\" ? positioned.reason : (error as Error).message;\n  const place = positioned.place;\n  if (\n    place !== null &&\n    place !== undefined &&\n    typeof place.line === \"number\" &&\n    typeof place.column === \"number\"\n  ) {\n    return `${reason} (line ${place.line}, column ${place.column})`;\n  }\n  return reason;\n}\n\n/**\n * spec/rules/modes.md 3.7.2. **The parser's own error type must never escape.** A\n * `VFileMessage`, a `Nokogiri::SyntaxError` or a Python exception on the public surface puts a\n * dependency's type into the contract and is unreproducible in the other four runtimes; only the\n * code is contractual (ARCHITECTURE.md 4.6), while the message and the source position are\n * useful and are not.\n *\n * Reachable in exactly one place: `dialect: \"mdx\"`, because MDX embeds JavaScript. HTML parsing\n * is specified with total error recovery and every byte sequence is valid CommonMark, so neither\n * of the other two languages can fail. The wrapper is applied to all of them anyway — the\n * guarantee is that no parser type escapes, not that this particular parser is trusted.\n */\nexport function wrapParserErrors<T>(what: string, run: () => T): T {\n  try {\n    return run();\n  } catch (error) {\n    if (error instanceof PolytypoError) throw error;\n    throw new PolytypoError(\n      \"POLYTYPO_MALFORMED_INPUT\",\n      `Input does not parse as ${what}: ${describe(error)}`,\n    );\n  }\n}\n"]}
+{"version":3,"sources":["/home/runner/work/polytypo-js/polytypo-js/dist/chunk-V6SSSZLP.cjs","../src/modes/html.ts","../src/modes/parse-error.ts"],"names":[],"mappings":"AAAA;AACE;AACF,wDAA6B;AAC7B;AACA;ACJA,gCAAqC;ADMrC;AACA;AEAA,SAAS,QAAA,CAAS,KAAA,EAAwB;AACxC,EAAA,GAAA,CAAI,OAAO,MAAA,IAAU,SAAA,GAAY,MAAA,IAAU,IAAA,EAAM,OAAO,MAAA,CAAO,KAAK,CAAA;AACpE,EAAA,MAAM,WAAA,EAAa,KAAA;AACnB,EAAA,MAAM,OAAA,EACJ,OAAO,UAAA,CAAW,OAAA,IAAW,SAAA,EAAW,UAAA,CAAW,OAAA,EAAU,KAAA,CAAgB,OAAA;AAC/E,EAAA,MAAM,MAAA,EAAQ,UAAA,CAAW,KAAA;AACzB,EAAA,GAAA,CACE,MAAA,IAAU,KAAA,GACV,MAAA,IAAU,KAAA,EAAA,GACV,OAAO,KAAA,CAAM,KAAA,IAAS,SAAA,GACtB,OAAO,KAAA,CAAM,OAAA,IAAW,QAAA,EACxB;AACA,IAAA,OAAO,CAAA,EAAA;AACT,EAAA;AACO,EAAA;AACT;AAcgB;AACV,EAAA;AACK,IAAA;AACA,EAAA;AACH,IAAA;AACE,IAAA;AACJ,MAAA;AACA,MAAA;AACF,IAAA;AACF,EAAA;AACF;AFjBY;AACA;ACfN;AACJ,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACD;AAEe;AACP,EAAA;AACT;AAEM;AACA;AACO;AACP;AACA;AAGA;AAEG;AACA,EAAA;AACT;AAES;AACA,EAAA;AACT;AAES;AACA,EAAA;AACT;AAQS;AACH,EAAA;AACI,EAAA;AACC,EAAA;AAEL,EAAA;AACG,IAAA;AACC,IAAA;AACA,IAAA;AACG,IAAA;AACH,IAAA;AAEJ,IAAA;AAGK,MAAA;AACP,IAAA;AACI,IAAA;AACG,IAAA;AACT,EAAA;AAEM,EAAA;AAEJ,EAAA;AAIK,IAAA;AACP,EAAA;AACU,EAAA;AACH,EAAA;AACT;AAOgB;AACR,EAAA;AACF,EAAA;AACI,EAAA;AACD,EAAA;AACD,IAAA;AACI,MAAA;AACF,MAAA;AACE,QAAA;AACJ,QAAA;AACI,QAAA;AACJ,QAAA;AACF,MAAA;AACF,IAAA;AACK,IAAA;AACP,EAAA;AACU,EAAA;AACH,EAAA;AACT;AAES;AACA,EAAA;AACT;AAES;AACE,EAAA;AACD,IAAA;AACF,IAAA;AACJ,IAAA;AACQ,MAAA;AACR,IAAA;AACA,IAAA;AACF,EAAA;AAGS,EAAA;AAEL,EAAA;AAGK,EAAA;AACC,IAAA;AACR,IAAA;AACF,EAAA;AAEK,EAAA;AACL,EAAA;AACF;AAQgB;AACR,EAAA;AACA,EAAA;AACE,EAAA;AACD,EAAA;AACT;AAGgB;AACR,EAAA;AAA4B,IAAA;AAChC,IAAA;AACF,EAAA;AACM,EAAA;AACE,EAAA;AACD,EAAA;AAGT;AD9BY;AACA;AACA;AACA;AACA;AACA;AACA","file":"/home/runner/work/polytypo-js/polytypo-js/dist/chunk-V6SSSZLP.cjs","sourcesContent":[null,"import { parse, parseFragment } from \"parse5\";\nimport type { DefaultTreeAdapterMap } from \"parse5\";\nimport { wrapParserErrors } from \"./parse-error.js\";\nimport type { Span } from \"./spans.js\";\n\ntype Node = DefaultTreeAdapterMap[\"node\"];\ntype ParentNode = DefaultTreeAdapterMap[\"parentNode\"];\n\n/**\n * spec/rules/modes.md 3.6, exhaustive and **closed**: extending it is a spec change, not an\n * implementation decision. `svg` and `math` are here because in MathML a quotation mark, a\n * hyphen and a prime are operators and identifiers — substituting a curly glyph changes what\n * the expression means. Everything not listed is processable, **including unknown and custom\n * elements**: guessing from a tag name is exactly the heuristic that diverges across runtimes.\n */\nconst SKIPPED_ELEMENTS: ReadonlySet<string> = new Set([\n  \"code\",\n  \"pre\",\n  \"kbd\",\n  \"samp\",\n  \"var\",\n  \"script\",\n  \"style\",\n  \"textarea\",\n  \"svg\",\n  \"math\",\n]);\n\nexport function isSkippedElement(tagName: string): boolean {\n  return SKIPPED_ELEMENTS.has(tagName);\n}\n\nconst AMPERSAND = 0x26;\nconst SEMICOLON = 0x3b;\nconst HASH = 0x23;\nconst LOWER_X = 0x78;\nconst UPPER_X = 0x58;\n\n/** Longest named reference in the WHATWG table is 32 characters; the guard is a bound, not a rule. */\nconst MAX_REFERENCE_BODY = 34;\n\nfunction isAsciiDigit(unit: number): boolean {\n  return unit >= 0x30 && unit <= 0x39;\n}\n\nfunction isAsciiHexDigit(unit: number): boolean {\n  return isAsciiDigit(unit) || (unit >= 0x61 && unit <= 0x66) || (unit >= 0x41 && unit <= 0x46);\n}\n\nfunction isAsciiAlphanumeric(unit: number): boolean {\n  return isAsciiDigit(unit) || (unit >= 0x61 && unit <= 0x7a) || (unit >= 0x41 && unit <= 0x5a);\n}\n\n/**\n * The end offset of a well-formed character reference starting at `at`, or -1. A bare `&` that a\n * parser would repair is deliberately **not** treated as a reference: it stays inside the span,\n * where no rule can emit or delete it, rather than manufacturing a boundary that suppresses\n * conversions around it.\n */\nfunction characterReferenceEnd(source: string, at: number, limit: number): number {\n  if (source.charCodeAt(at) !== AMPERSAND) return -1;\n  let i = at + 1;\n  if (i >= limit) return -1;\n\n  if (source.charCodeAt(i) === HASH) {\n    i += 1;\n    const unit = i < limit ? source.charCodeAt(i) : -1;\n    const hex = unit === LOWER_X || unit === UPPER_X;\n    if (hex) i += 1;\n    const digitsAt = i;\n    while (\n      i < limit &&\n      (hex ? isAsciiHexDigit(source.charCodeAt(i)) : isAsciiDigit(source.charCodeAt(i)))\n    ) {\n      i += 1;\n    }\n    if (i === digitsAt) return -1;\n    return i < limit && source.charCodeAt(i) === SEMICOLON ? i + 1 : -1;\n  }\n\n  const nameAt = i;\n  while (\n    i < limit &&\n    i - nameAt < MAX_REFERENCE_BODY &&\n    isAsciiAlphanumeric(source.charCodeAt(i))\n  ) {\n    i += 1;\n  }\n  if (i === nameAt) return -1;\n  return i < limit && source.charCodeAt(i) === SEMICOLON ? i + 1 : -1;\n}\n\n/**\n * modes.md 3.6: every character reference is skipped as an **opaque unit**, and a text node\n * containing one is split into spans around it. That is the only way `&nbsp;` survives as\n * `&nbsp;` instead of becoming a literal U+00A0 on the way out.\n */\nexport function splitCharacterReferences(source: string, start: number, end: number): Span[] {\n  const spans: Span[] = [];\n  let cursor = start;\n  let i = start;\n  while (i < end) {\n    if (source.charCodeAt(i) === AMPERSAND) {\n      const stop = characterReferenceEnd(source, i, end);\n      if (stop > 0) {\n        if (i > cursor) spans.push({ start: cursor, end: i });\n        cursor = stop;\n        i = stop;\n        continue;\n      }\n    }\n    i += 1;\n  }\n  if (end > cursor) spans.push({ start: cursor, end });\n  return spans;\n}\n\nfunction isParent(node: Node): node is ParentNode {\n  return \"childNodes\" in node;\n}\n\nfunction collect(node: Node, source: string, spans: Span[]): void {\n  if (node.nodeName === \"#text\") {\n    const location = node.sourceCodeLocation;\n    if (location === undefined || location === null) return;\n    for (const span of splitCharacterReferences(source, location.startOffset, location.endOffset)) {\n      spans.push(span);\n    }\n    return;\n  }\n\n  // Comments, doctype and processing instructions never contribute a span (modes.md 3.6).\n  if (node.nodeName === \"#comment\" || node.nodeName === \"#documentType\") return;\n\n  if (\"tagName\" in node && isSkippedElement(node.tagName)) return;\n\n  // A `template`'s children live in a separate document fragment; `template` is not skipped.\n  if (node.nodeName === \"template\" && \"content\" in node) {\n    collect(node.content, source, spans);\n    return;\n  }\n\n  if (!isParent(node)) return;\n  for (const child of node.childNodes) collect(child, source, spans);\n}\n\n/**\n * Locate the processable spans of an HTML document. The tree is used only to find offsets and is\n * then discarded — modes.md 4 forbids reserialisation, which is what makes attribute quoting,\n * self-closing forms, entity spelling, tag case and malformed-input recovery preserved by\n * construction rather than by effort.\n */\nexport function htmlSpans(source: string): Span[] {\n  const document = wrapParserErrors(\"HTML\", () => parse(source, { sourceCodeLocationInfo: true }));\n  const spans: Span[] = [];\n  collect(document, source, spans);\n  return spans;\n}\n\n/** The same, for a fragment: used for HTML blocks embedded in markdown (modes.md 3.7). */\nexport function htmlFragmentSpans(source: string, offset: number): Span[] {\n  const fragment = wrapParserErrors(\"HTML\", () =>\n    parseFragment(source, { sourceCodeLocationInfo: true }),\n  );\n  const spans: Span[] = [];\n  collect(fragment, source, spans);\n  return offset === 0\n    ? spans\n    : spans.map((span) => ({ start: span.start + offset, end: span.end + offset }));\n}\n","import { PolytypoError } from \"../errors.js\";\n\ninterface PositionedError {\n  readonly reason?: unknown;\n  readonly place?: { readonly line?: unknown; readonly column?: unknown } | null;\n}\n\nfunction describe(error: unknown): string {\n  if (typeof error !== \"object\" || error === null) return String(error);\n  const positioned = error as PositionedError;\n  const reason =\n    typeof positioned.reason === \"string\" ? positioned.reason : (error as Error).message;\n  const place = positioned.place;\n  if (\n    place !== null &&\n    place !== undefined &&\n    typeof place.line === \"number\" &&\n    typeof place.column === \"number\"\n  ) {\n    return `${reason} (line ${place.line}, column ${place.column})`;\n  }\n  return reason;\n}\n\n/**\n * spec/rules/modes.md 3.7.2. **The parser's own error type must never escape.** A\n * `VFileMessage`, a `Nokogiri::SyntaxError` or a Python exception on the public surface puts a\n * dependency's type into the contract and is unreproducible in the other four runtimes; only the\n * code is contractual (ARCHITECTURE.md 4.6), while the message and the source position are\n * useful and are not.\n *\n * Reachable in exactly one place: `dialect: \"mdx\"`, because MDX embeds JavaScript. HTML parsing\n * is specified with total error recovery and every byte sequence is valid CommonMark, so neither\n * of the other two languages can fail. The wrapper is applied to all of them anyway — the\n * guarantee is that no parser type escapes, not that this particular parser is trusted.\n */\nexport function wrapParserErrors<T>(what: string, run: () => T): T {\n  try {\n    return run();\n  } catch (error) {\n    if (error instanceof PolytypoError) throw error;\n    throw new PolytypoError(\n      \"POLYTYPO_MALFORMED_INPUT\",\n      `Input does not parse as ${what}: ${describe(error)}`,\n    );\n  }\n}\n"]}

package/dist/{chunk-QBPOFGSD.js → chunk-VBZAV63L.js}

@@ -6,7 +6,7 @@ import {
   runRules,
   runRulesRecording,
   toCodePoints
-} from "./chunk-ETKYEVE6.js";
+} from "./chunk-W77K6K6G.js";
 
 // src/engine/text-pipeline.ts
 function runTextPipeline(input, options) {
@@ -28,4 +28,4 @@ export {
   runTextPipeline,
   analyzeTextPipeline
 };
-//# sourceMappingURL=chunk-QBPOFGSD.js.map
+//# sourceMappingURL=chunk-VBZAV63L.js.map

package/dist/{chunk-SD4SLSAR.cjs → chunk-VCD45VZH.cjs}

@@ -61,6 +61,7 @@ var KNOWN_LOCALES = [
   "pt-PT",
   "ru",
   "sv",
+  "tr",
   "uk"
 ];
 var ALIASES = {
@@ -86,6 +87,7 @@ var LOCALES = {
   "pt-PT": { locale: "pt-PT", name: "Portuguese (Portugal)", quotes: { primary: { open: "\xAB", close: "\xBB", innerSpace: "none" }, secondary: { open: "\u201C", close: "\u201D", innerSpace: "none" }, elisionIdioms: [], elisionClitics: { before: ["d", "n", "pel", "m", "t", "lh", "sant"], after: [] } }, dash: { parenthetical: "em-spaced", range: "none" }, ellipsis: { abbreviatedAfterTerminal: false }, hyphen: { prefixes: [], suffixes: [], compounds: [] }, nbsp: { beforePunctuation: [], narrowBeforePunctuation: [], afterShortWords: [], abbreviations: ["p. ex."], beforeUnits: ["%", "\u2030", "\xB0C", "km", "cm", "mm", "kg", "kW", "kWh", "Hz"], beforeNumber: ["p."], beforeWord: [], afterSymbols: [], initialBinding: "none" } },
   ru: { locale: "ru", name: "Russian", quotes: { primary: { open: "\xAB", close: "\xBB", innerSpace: "none" }, secondary: { open: "\u201E", close: "\u201C", innerSpace: "none" }, elisionIdioms: [], elisionClitics: { before: [], after: [] } }, dash: { parenthetical: "em-spaced", range: "em-tight" }, ellipsis: { abbreviatedAfterTerminal: true }, hyphen: { prefixes: ["\u043A\u043E\u0435-", "\u043A\u043E\u0439-"], suffixes: ["-\u0442\u043E", "-\u043B\u0438\u0431\u043E", "-\u043D\u0438\u0431\u0443\u0434\u044C", "-\u0442\u0430\u043A\u0438", "-\u043A\u0430"], compounds: ["\u0438\u0437-\u0437\u0430", "\u0438\u0437-\u043F\u043E\u0434"] }, nbsp: { beforePunctuation: [], narrowBeforePunctuation: [], afterShortWords: ["\u0430", "\u0432", "\u0438", "\u043A", "\u043E", "\u0441", "\u0443", "\u0432\u043E", "\u0434\u043E", "\u0437\u0430", "\u0438\u0437", "\u043A\u043E", "\u043D\u0430", "\u043E\u0431", "\u043E\u0442", "\u043F\u043E", "\u0441\u043E"], abbreviations: ["\u0438 \u0442. \u0434.", "\u0438 \u0442. \u043F.", "\u0442. \u0435.", "\u0438 \u0434\u0440."], beforeUnits: ["%", "\u2030", "\u20BD", "\xB0C", "\u043A\u043C", "\u0441\u043C", "\u043C\u043C", "\u043A\u0433", "\u0433.", "\u0433\u0433.", "\u0432.", "\u0432\u0432.", "\u0442\u044B\u0441.", "\u043C\u043B\u043D", "\u043C\u043B\u0440\u0434"], beforeNumber: [], beforeWord: ["\u0443\u043B.", "\u043F\u043B."], afterSymbols: ["\u2116", "\xA7"], initialBinding: "chain" } },
   sv: { locale: "sv", name: "Swedish", quotes: { primary: { open: "\u201D", close: "\u201D", innerSpace: "none" }, secondary: { open: "\u2019", close: "\u2019", innerSpace: "none" }, elisionIdioms: [], elisionClitics: { before: [], after: [] } }, dash: { parenthetical: "en-spaced", range: "en-tight" }, ellipsis: { abbreviatedAfterTerminal: false }, hyphen: { prefixes: [], suffixes: [], compounds: [] }, nbsp: { beforePunctuation: [], narrowBeforePunctuation: [], afterShortWords: [], abbreviations: [], beforeUnits: ["%", "\u2030", "kr", "\u20AC", "kg", "g", "mg", "km", "m", "cm", "mm", "l", "dl", "cl", "ml", "kW", "kWh", "min", "sek", "tim", "h", "\xB0C"], beforeNumber: [], beforeWord: [], afterSymbols: ["\xA7"], initialBinding: "none" } },
+  tr: { locale: "tr", name: "Turkish", quotes: { primary: { open: "\u201C", close: "\u201D", innerSpace: "none" }, secondary: { open: "\u2018", close: "\u2019", innerSpace: "none" }, elisionIdioms: [], elisionClitics: { before: [], after: ["nin", "n\u0131n", "de", "da", "te", "ye", "yle", "n\u0131", "dan", "den", "lik", "nci", "\xFCm"] } }, dash: { parenthetical: "none", range: "none" }, ellipsis: { abbreviatedAfterTerminal: true }, hyphen: { prefixes: [], suffixes: [], compounds: [] }, nbsp: { beforePunctuation: [], narrowBeforePunctuation: [], afterShortWords: [], abbreviations: ["Kur. B\u015Fk.", "N\xF6. Sb."], beforeUnits: ["mm", "cm", "km", "kg", "mg", "hl", "m\xB2", "cm\xB2", "\xB0C", "ton"], beforeNumber: [], beforeWord: [], afterSymbols: [], initialBinding: "none" } },
   uk: { locale: "uk", name: "Ukrainian", quotes: { primary: { open: "\xAB", close: "\xBB", innerSpace: "none" }, secondary: { open: "\u201C", close: "\u201D", innerSpace: "none" }, elisionIdioms: [], elisionClitics: { before: [], after: [] } }, dash: { parenthetical: "em-spaced", range: "em-tight" }, ellipsis: { abbreviatedAfterTerminal: true }, hyphen: { prefixes: ["\u0431\u0443\u0434\u044C-", "\u043A\u0430\u0437\u043D\u0430-", "\u0445\u0442\u043E\u0437\u043D\u0430-", "\u0431\u043E\u0437\u043D\u0430-"], suffixes: ["-\u0431\u043E", "-\u043D\u043E", "-\u043E\u0442", "-\u0442\u043E", "-\u0442\u0430\u043A\u0438", "-\u0431\u0443\u0434\u044C", "-\u043D\u0435\u0431\u0443\u0434\u044C"], compounds: ["\u0432\u0438\u0434-\u0432\u043E", "\u0433\u0440-\u043D", "\u0456\u043D-\u0442", "\u0440-\u043D", "\u0443\u043D-\u0442", "\u0444-\u043A\u0430", "\u0442-\u0432\u043E", "\u0437-\u0437\u0430", "\u0456\u0437-\u0437\u0430", "\u0437-\u043D\u0430\u0434", "\u0437-\u043F\u0435\u0440\u0435\u0434", "\u0437-\u043F\u0456\u0434", "\u0456\u0437-\u043F\u0456\u0434", "\u0437-\u043F\u043E\u0437\u0430", "\u0437-\u043F\u043E\u043C\u0456\u0436", "\u0437-\u043F\u043E\u043D\u0430\u0434", "\u0437-\u043F\u043E\u043F\u0456\u0434", "\u0437-\u043F\u043E\u0441\u0435\u0440\u0435\u0434", "\u0437-\u043F\u0440\u043E\u043C\u0456\u0436"] }, nbsp: { beforePunctuation: [], narrowBeforePunctuation: [], afterShortWords: [], abbreviations: ["\u0456 \u0442. \u0434.", "\u0456 \u0442. \u0456\u043D.", "\u0442\u0430 \u0456\u043D.", "\u043A\u0443\u0431. \u0441\u043C"], beforeUnits: ["%", "\u0433\u0430", "\u0433\u043E\u0434", "\u0434\u043C", "\u043A\u0411", "\u043A\u0412\u0442", "\u043A\u0433", "\u043A\u043C", "\u043C\u043C", "\u0441\u043C", "\u0445\u0432", "\u0412\u0442", "\u043A\u043C/\u0433\u043E\u0434", "\u043C/\u0441", "\u043C\u043B\u043D", "\u043C\u043B\u0440\u0434", "\u0442\u0440\u043B\u043D", "\u0433\u0440\u043D", "\u0440."], beforeNumber: [], beforeWord: ["\u0430\u043A\u0430\u0434.", "\u0434\u043E\u0446.", "\u043F\u0440\u043E\u0444."], afterSymbols: [], initialBinding: "chain" } }
 };
 
@@ -5835,4 +5837,4 @@ function runRulesRecording(cp, planned, locale, mode, narrowTarget, origin, inpu
 
 
 exports.PolytypoError = PolytypoError; exports.toCodePoints = toCodePoints; exports.fromCodePoints = fromCodePoints; exports.getLocaleData = getLocaleData; exports.resolveNarrowTarget = resolveNarrowTarget; exports.MARKER = MARKER; exports.LINE_MARKER = LINE_MARKER; exports.isMarker = isMarker; exports.RULES = RULES; exports.applyEdits = applyEdits; exports.NO_ORIGIN = NO_ORIGIN; exports.planRules = planRules; exports.runRules = runRules; exports.runRulesRecording = runRulesRecording;
-//# sourceMappingURL=chunk-SD4SLSAR.cjs.map
+//# sourceMappingURL=chunk-VCD45VZH.cjs.map