polytypo 1.2.0 → 1.3.0

package/dist/{chunk-6NTDU63P.js → chunk-PNX4XW5I.js}

@@ -1,14 +1,18 @@
 import {
   htmlFragmentSpans,
   isSkippedElement,
-  runOverSpans,
   wrapParserErrors
-} from "./chunk-VARU43JT.js";
+} from "./chunk-NT2RADO3.js";
+import {
+  analyzeOverSpans,
+  runOverSpans
+} from "./chunk-PLWW5KPC.js";
 import {
   PolytypoError,
   getLocaleData,
-  planRules
-} from "./chunk-OHF7OFWY.js";
+  planRules,
+  resolveNarrowTarget
+} from "./chunk-GLQQOA6T.js";
 
 // src/modes/markdown.ts
 import { parse, postprocess, preprocess } from "micromark";
@@ -35,9 +39,11 @@ var SKIPPED_TOKEN_TYPES = /* @__PURE__ */ new Set([
   "mdxFlowExpression",
   "mdxTextExpression",
   "mdxjsEsm",
-  // Not in modes.md 3.7, and it has to be: without it the second `---` of a YAML block reads as
-  // a setext underline, `title: Une note` becomes a paragraph, and `fr` puts a narrow no-break
-  // space in front of the colon of a machine-read metadata field. Reported as a spec gap.
+  // modes.md 3.7.3, and it matters: without it the second `---` of a YAML block reads as a setext
+  // underline, `title: Une note` becomes a paragraph, and `fr` puts a narrow no-break space in
+  // front of the colon of a machine-read metadata field. Covered by conformance fixtures
+  // en-us-markdown-{commonmark,mdx}-frontmatter, fr-markdown-{commonmark,mdx}-frontmatter-nbsp,
+  // en-us-markdown-commonmark-frontmatter-toml and -frontmatter-unterminated.
   "frontmatter"
 ]);
 var RAW_TAG_TOKEN_TYPES = /* @__PURE__ */ new Set([
@@ -104,8 +110,10 @@ function resolveDialect(dialect) {
     `Unknown dialect "${String(dialect)}". Expected "commonmark" or "mdx".`
   );
 }
+var FRONTMATTER_MATTERS = ["yaml", "toml"];
 function extensionsFor(dialect) {
-  return dialect === "mdx" ? [frontmatter(), gfm(), mdxjs()] : [frontmatter(), gfm()];
+  const fm = frontmatter([...FRONTMATTER_MATTERS]);
+  return dialect === "mdx" ? [fm, gfm(), mdxjs()] : [fm, gfm()];
 }
 function markdownSpans(source, dialect) {
   const events = wrapParserErrors(dialect, () => tokenize(source, extensionsFor(dialect)));
@@ -150,14 +158,30 @@ function markdownSpans(source, dialect) {
 
 // src/engine/markdown-pipeline.ts
 function runMarkdownPipeline(input, options) {
+  const narrowTarget = resolveNarrowTarget(options.narrowNbsp);
   const planned = planRules(options.rules);
   const locale = getLocaleData(options.locale);
   const dialect = resolveDialect(options.dialect);
   const spans = markdownSpans(input, dialect);
-  return runOverSpans(input, spans, planned, locale, "markdown");
+  return runOverSpans(input, spans, planned, locale, "markdown", narrowTarget);
+}
+function analyzeMarkdownPipeline(input, options) {
+  const narrowTarget = resolveNarrowTarget(options.narrowNbsp);
+  const planned = planRules(options.rules);
+  const locale = getLocaleData(options.locale);
+  const dialect = resolveDialect(options.dialect);
+  return analyzeOverSpans(
+    input,
+    markdownSpans(input, dialect),
+    planned,
+    locale,
+    "markdown",
+    narrowTarget
+  );
 }
 
 export {
-  runMarkdownPipeline
+  runMarkdownPipeline,
+  analyzeMarkdownPipeline
 };
-//# sourceMappingURL=chunk-6NTDU63P.js.map
+//# sourceMappingURL=chunk-PNX4XW5I.js.map

package/dist/chunk-PNX4XW5I.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/modes/markdown.ts","../src/engine/markdown-pipeline.ts"],"sourcesContent":["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"],"mappings":";;;;;;;;;;;;;;;;;AAAA,SAAS,OAAO,aAAa,kBAAkB;AAC/C,SAAS,mBAAmB;AAC5B,SAAS,WAAW;AACpB,SAAS,aAAa;AAgBtB,IAAM,sBAA2C,oBAAI,IAAI;AAAA;AAAA,EAEvD;AAAA,EACA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA,EACA;AAAA;AAAA;AAAA,EAGA;AAAA,EACA;AAAA,EACA;AAAA;AAAA;AAAA;AAAA,EAIA;AAAA,EACA;AAAA,EACA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA;AACF,CAAC;AAGD,IAAM,sBAA2C,oBAAI,IAAI;AAAA,EACvD;AAAA,EACA;AAAA,EACA;AACF,CAAC;AAED,IAAM,YAAY;AAClB,IAAM,eAAe;AACrB,IAAM,QAAQ;AACd,IAAM,cAAc;AACpB,IAAM,WAAW;AAEjB,SAAS,cAAc,MAAuB;AAC5C,SAAQ,QAAQ,MAAQ,QAAQ,OAAU,QAAQ,MAAQ,QAAQ;AACpE;AAEA,SAAS,cAAc,MAAuB;AAI5C,SACE,SAAS,gBACT,SAAS,SACT,SAAS,MACT,SAAS,KACT,SAAS,MACT,SAAS;AAEb;AAGA,SAAS,WAAW,OAAuB;AACzC,MAAI,MAAM;AACV,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK,GAAG;AACxC,UAAM,OAAO,MAAM,WAAW,CAAC;AAC/B,WAAO,QAAQ,MAAQ,QAAQ,KAAO,OAAO,aAAa,OAAO,EAAI,IAAI,MAAM,CAAC;AAAA,EAClF;AACA,SAAO;AACT;AAQA,SAAS,QAAQ,KAA4B;AAC3C,MAAI,IAAI,WAAW,CAAC,MAAM,UAAW,QAAO;AAC5C,MAAI,IAAI;AACR,QAAM,UAAU,IAAI,WAAW,CAAC,MAAM;AACtC,MAAI,QAAS,MAAK;AAElB,MAAI,IAAI,WAAW,CAAC,MAAM,eAAe,IAAI,WAAW,CAAC,MAAM,SAAU,QAAO;AAChF,MAAI,CAAC,cAAc,IAAI,WAAW,CAAC,CAAC,EAAG,QAAO;AAC9C,QAAM,SAAS;AACf,SAAO,IAAI,IAAI,UAAU,cAAc,IAAI,WAAW,CAAC,CAAC,EAAG,MAAK;AAChE,QAAM,cAAc,IAAI,WAAW,IAAI,SAAS,CAAC,MAAM;AACvD,SAAO,EAAE,MAAM,IAAI,MAAM,QAAQ,CAAC,GAAG,SAAS,YAAY;AAC5D;AAQA,SAAS,mBAAmB,OAAiB,KAAa,eAA8B;AACtF,QAAM,SAAS,QAAQ,GAAG;AAC1B,MAAI,WAAW,KAAM;AACrB,QAAM,OAAO,gBAAgB,OAAO,OAAO,WAAW,OAAO,IAAI;AACjE,MAAI,OAAO,SAAS;AAClB,QAAI,MAAM,MAAM,SAAS,CAAC,MAAM,KAAM,OAAM,IAAI;AAChD;AAAA,EACF;AACA,MAAI,CAAC,OAAO,eAAe,iBAAiB,IAAI,EAAG,OAAM,KAAK,IAAI;AACpE;AAEA,SAAS,SAAS,QAAgB,YAA2C;AAC3E,SAAO;AAAA,IACL,MAAM,EAAE,YAAY,CAAC,GAAG,UAAU,EAAE,CAAC,EAClC,SAAS,EACT,MAAM,WAAW,EAAE,QAAQ,MAAM,IAAI,CAAC;AAAA,EAC3C;AACF;AAQO,SAAS,eAAe,SAAuC;AACpE,MAAI,YAAY,gBAAgB,YAAY,MAAO,QAAO;AAC1D,MAAI,YAAY,QAAW;AACzB,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,QAAM,IAAI;AAAA,IACR;AAAA,IACA,oBAAoB,OAAO,OAAO,CAAC;AAAA,EACrC;AACF;AAKA,IAAM,sBAAsB,CAAC,QAAQ,MAAM;AAE3C,SAAS,cAAc,SAA+B;AACpD,QAAM,KAAK,YAAY,CAAC,GAAG,mBAAmB,CAAC;AAC/C,SAAO,YAAY,QAAQ,CAAC,IAAI,IAAI,GAAG,MAAM,CAAC,IAAI,CAAC,IAAI,IAAI,CAAC;AAC9D;AAEO,SAAS,cAAc,QAAgB,SAA0B;AACtE,QAAM,SAAS,iBAAiB,SAAS,MAAM,SAAS,QAAQ,cAAc,OAAO,CAAC,CAAC;AACvF,QAAM,QAAgB,CAAC;AACvB,QAAM,eAAyB,CAAC;AAChC,MAAI,YAAY;AAEhB,aAAW,CAAC,MAAM,KAAK,KAAK,QAAQ;AAClC,UAAM,OAAO,MAAM;AAEnB,QAAI,oBAAoB,IAAI,IAAI,GAAG;AACjC,mBAAa,SAAS,UAAU,IAAI;AACpC;AAAA,IACF;AAEA,QAAI,oBAAoB,IAAI,IAAI,GAAG;AACjC,UAAI,SAAS,SAAS;AACpB;AAAA,UACE;AAAA,UACA,OAAO,MAAM,MAAM,MAAM,QAAQ,MAAM,IAAI,MAAM;AAAA,UACjD,SAAS;AAAA,QACX;AAAA,MACF;AACA,mBAAa,SAAS,UAAU,IAAI;AACpC;AAAA,IACF;AAIA,QAAI,SAAS,YAAY;AACvB,UAAI,SAAS,WAAW,cAAc,KAAK,aAAa,WAAW,GAAG;AACpE,mBAAW,QAAQ;AAAA,UACjB,OAAO,MAAM,MAAM,MAAM,QAAQ,MAAM,IAAI,MAAM;AAAA,UACjD,MAAM,MAAM;AAAA,QACd,GAAG;AACD,gBAAM,KAAK,IAAI;AAAA,QACjB;AAAA,MACF;AACA,mBAAa,SAAS,UAAU,IAAI;AACpC;AAAA,IACF;AAEA,QAAI,SAAS,WAAW,SAAS,UAAU,cAAc,KAAK,aAAa,WAAW,GAAG;AACvF,YAAM,KAAK,EAAE,OAAO,MAAM,MAAM,QAAQ,KAAK,MAAM,IAAI,OAAO,CAAC;AAAA,IACjE;AAAA,EACF;AAEA,SAAO;AACT;;;AChMO,SAAS,oBAAoB,OAAe,SAAmC;AACpF,QAAM,eAAe,oBAAoB,QAAQ,UAAU;AAC3D,QAAM,UAAU,UAAU,QAAQ,KAAK;AACvC,QAAM,SAAS,cAAc,QAAQ,MAAM;AAC3C,QAAM,UAAU,eAAe,QAAQ,OAAO;AAC9C,QAAM,QAAQ,cAAc,OAAO,OAAO;AAC1C,SAAO,aAAa,OAAO,OAAO,SAAS,QAAQ,YAAY,YAAY;AAC7E;AAKO,SAAS,wBAAwB,OAAe,SAAqC;AAC1F,QAAM,eAAe,oBAAoB,QAAQ,UAAU;AAC3D,QAAM,UAAU,UAAU,QAAQ,KAAK;AACvC,QAAM,SAAS,cAAc,QAAQ,MAAM;AAC3C,QAAM,UAAU,eAAe,QAAQ,OAAO;AAC9C,SAAO;AAAA,IACL;AAAA,IACA,cAAc,OAAO,OAAO;AAAA,IAC5B;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;","names":[]}

package/dist/chunk-QLV5EWGF.cjs

@@ -0,0 +1,33 @@
+"use strict";Object.defineProperty(exports, "__esModule", {value: true});
+
+var _chunk72GQS2N7cjs = require('./chunk-72GQS2N7.cjs');
+
+
+
+var _chunk7S5PZ7T5cjs = require('./chunk-7S5PZ7T5.cjs');
+
+
+
+
+var _chunkYX6QTWOVcjs = require('./chunk-YX6QTWOV.cjs');
+
+// src/engine/html-pipeline.ts
+function runHtmlPipeline(input, options) {
+  const narrowTarget = _chunkYX6QTWOVcjs.resolveNarrowTarget.call(void 0, options.narrowNbsp);
+  const planned = _chunkYX6QTWOVcjs.planRules.call(void 0, options.rules);
+  const locale = _chunkYX6QTWOVcjs.getLocaleData.call(void 0, options.locale);
+  const spans = _chunk72GQS2N7cjs.htmlSpans.call(void 0, input);
+  return _chunk7S5PZ7T5cjs.runOverSpans.call(void 0, input, spans, planned, locale, "html", narrowTarget);
+}
+function analyzeHtmlPipeline(input, options) {
+  const narrowTarget = _chunkYX6QTWOVcjs.resolveNarrowTarget.call(void 0, options.narrowNbsp);
+  const planned = _chunkYX6QTWOVcjs.planRules.call(void 0, options.rules);
+  const locale = _chunkYX6QTWOVcjs.getLocaleData.call(void 0, options.locale);
+  return _chunk7S5PZ7T5cjs.analyzeOverSpans.call(void 0, input, _chunk72GQS2N7cjs.htmlSpans.call(void 0, input), planned, locale, "html", narrowTarget);
+}
+
+
+
+
+exports.runHtmlPipeline = runHtmlPipeline; exports.analyzeHtmlPipeline = analyzeHtmlPipeline;
+//# sourceMappingURL=chunk-QLV5EWGF.cjs.map

package/dist/chunk-QLV5EWGF.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["/home/runner/work/polytypo-js/polytypo-js/dist/chunk-QLV5EWGF.cjs","../src/engine/html-pipeline.ts"],"names":[],"mappings":"AAAA;AACE;AACF,wDAA6B;AAC7B;AACE;AACA;AACF,wDAA6B;AAC7B;AACE;AACA;AACA;AACF,wDAA6B;AAC7B;AACA;ACIO,SAAS,eAAA,CAAgB,KAAA,EAAe,OAAA,EAAmC;AAChF,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,MAAA,EAAQ,yCAAA,KAAe,CAAA;AAC7B,EAAA,OAAO,4CAAA,KAAa,EAAO,KAAA,EAAO,OAAA,EAAS,MAAA,EAAQ,MAAA,EAAQ,YAAY,CAAA;AACzE;AAGO,SAAS,mBAAA,CAAoB,KAAA,EAAe,OAAA,EAAqC;AACtF,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,OAAO,gDAAA,KAAiB,EAAO,yCAAA,KAAe,CAAA,EAAG,OAAA,EAAS,MAAA,EAAQ,MAAA,EAAQ,YAAY,CAAA;AACxF;ADJA;AACA;AACE;AACA;AACF,6FAAC","file":"/home/runner/work/polytypo-js/polytypo-js/dist/chunk-QLV5EWGF.cjs","sourcesContent":[null,"import { htmlSpans } from \"../modes/html.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 * `html` mode only. Imports `parse5` (via `../modes/html.js`) and nothing from\n * `../modes/markdown.js` or the Micromark/MDX stack — this is what makes `polytypo/html`'s\n * module graph exclude the Markdown parser (AUDIT_REMEDIATION_AND_RELEASE_PLAN.md 5.1).\n *\n * Validation order — `planRules` before `getLocaleData` before parsing — mirrors the pre-Stage-5\n * aggregate `runPipeline` exactly (an unknown-rule error must win over an unknown-locale error,\n * and both must win over a parse failure, since that is public, tested behaviour).\n */\nexport function runHtmlPipeline(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 spans = htmlSpans(input);\n  return runOverSpans(input, spans, planned, locale, \"html\", narrowTarget);\n}\n\n/** analyze.md §1, `html` mode: offsets are into the document, not into a span (analyze.md §6). */\nexport function analyzeHtmlPipeline(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  return analyzeOverSpans(input, htmlSpans(input), planned, locale, \"html\", narrowTarget);\n}\n"]}

package/dist/chunk-W3XMTX7Q.js

@@ -0,0 +1,31 @@
+import {
+  fromCodePoints,
+  getLocaleData,
+  planRules,
+  resolveNarrowTarget,
+  runRules,
+  runRulesRecording,
+  toCodePoints
+} from "./chunk-GLQQOA6T.js";
+
+// src/engine/text-pipeline.ts
+function runTextPipeline(input, options) {
+  const narrowTarget = resolveNarrowTarget(options.narrowNbsp);
+  const planned = planRules(options.rules);
+  const locale = getLocaleData(options.locale);
+  return fromCodePoints(runRules(toCodePoints(input), planned, locale, "text", narrowTarget));
+}
+function analyzeTextPipeline(input, options) {
+  const narrowTarget = resolveNarrowTarget(options.narrowNbsp);
+  const planned = planRules(options.rules);
+  const locale = getLocaleData(options.locale);
+  const cp = toCodePoints(input);
+  const origin = cp.map((_value, index) => index);
+  return runRulesRecording(cp, planned, locale, "text", narrowTarget, origin, cp.length);
+}
+
+export {
+  runTextPipeline,
+  analyzeTextPipeline
+};
+//# sourceMappingURL=chunk-W3XMTX7Q.js.map

package/dist/chunk-W3XMTX7Q.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/engine/text-pipeline.ts"],"sourcesContent":["import type { Options } from \"../types.js\";\nimport { fromCodePoints, toCodePoints } from \"./codepoints.js\";\nimport type { Change } from \"./origin.js\";\nimport { getLocaleData } from \"./locale.js\";\nimport { resolveNarrowTarget } from \"./narrow-target.js\";\nimport { planRules, runRules, runRulesRecording } from \"./rule-runner.js\";\n\n/**\n * `text` mode only. Deliberately imports nothing from `../modes/html.js` or\n * `../modes/markdown.js` (or their parser dependencies) — this is what makes `polytypo/text`'s\n * module graph exclude `parse5` and the Micromark stack (AUDIT_REMEDIATION_AND_RELEASE_PLAN.md\n * 5.1).\n *\n * Validation order — `planRules` before `getLocaleData` — mirrors the pre-Stage-5 aggregate\n * `runPipeline` exactly (an unknown-rule error must win over an unknown-locale error when both\n * are present, since that is public, tested behaviour, not an implementation detail this\n * refactor was authorised to change).\n */\nexport function runTextPipeline(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  return fromCodePoints(runRules(toCodePoints(input), planned, locale, \"text\", narrowTarget));\n}\n\n/** analyze.md §1: the same pipeline as `runTextPipeline`, reporting instead of applying. */\nexport function analyzeTextPipeline(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 cp = toCodePoints(input);\n  const origin = cp.map((_value, index) => index);\n  return runRulesRecording(cp, planned, locale, \"text\", narrowTarget, origin, cp.length);\n}\n"],"mappings":";;;;;;;;;;;AAkBO,SAAS,gBAAgB,OAAe,SAAmC;AAChF,QAAM,eAAe,oBAAoB,QAAQ,UAAU;AAC3D,QAAM,UAAU,UAAU,QAAQ,KAAK;AACvC,QAAM,SAAS,cAAc,QAAQ,MAAM;AAC3C,SAAO,eAAe,SAAS,aAAa,KAAK,GAAG,SAAS,QAAQ,QAAQ,YAAY,CAAC;AAC5F;AAGO,SAAS,oBAAoB,OAAe,SAAqC;AACtF,QAAM,eAAe,oBAAoB,QAAQ,UAAU;AAC3D,QAAM,UAAU,UAAU,QAAQ,KAAK;AACvC,QAAM,SAAS,cAAc,QAAQ,MAAM;AAC3C,QAAM,KAAK,aAAa,KAAK;AAC7B,QAAM,SAAS,GAAG,IAAI,CAAC,QAAQ,UAAU,KAAK;AAC9C,SAAO,kBAAkB,IAAI,SAAS,QAAQ,QAAQ,cAAc,QAAQ,GAAG,MAAM;AACvF;","names":[]}