@pie-players/pie-assessment-toolkit 0.3.44 → 0.3.45

package/dist/services/tts/math-alignment/speech-tokenizer.js

@@ -0,0 +1,143 @@
+import { extractSpokenText } from "../ssml/spoken-text.js";
+import { createSpeechAlignmentTokenPattern } from "../text-processing.js";
+// Shared Unicode-aware tokenizer, identical to the catalog span aligner's, so
+// spoken text tokenizes the same way on both paths (see
+// `createSpeechAlignmentTokenPattern`).
+const TOKEN_PATTERN = createSpeechAlignmentTokenPattern();
+const NUMERIC_WORDS = new Map([
+    ["zero", "0"],
+    ["one", "1"],
+    ["two", "2"],
+    ["three", "3"],
+    ["four", "4"],
+    ["five", "5"],
+    ["six", "6"],
+    ["seven", "7"],
+    ["eight", "8"],
+    ["nine", "9"],
+    ["ten", "10"],
+]);
+const normalizedTokenValue = (value) => {
+    const normalized = value.toLowerCase();
+    return NUMERIC_WORDS.get(normalized) || normalized;
+};
+const tokenizeSpokenText = (spokenText) => {
+    const tokens = [];
+    for (const match of spokenText.matchAll(TOKEN_PATTERN)) {
+        const text = match[0];
+        const start = match.index ?? 0;
+        tokens.push({
+            id: `speech-token-${tokens.length}`,
+            text,
+            normalized: normalizedTokenValue(text),
+            start,
+            end: start + text.length,
+            sourceStart: start,
+            sourceEnd: start + text.length,
+            coordinateSystem: "normalized-speech",
+        });
+    }
+    return tokens;
+};
+export const tokenizeSpeechSource = (args) => {
+    const extracted = extractSpokenText(args.speechText);
+    return {
+        speechText: args.speechText,
+        spokenText: extracted.spokenText,
+        tokens: tokenizeSpokenText(extracted.spokenText),
+        rawToSpokenOffsetMap: extracted.rawToSpokenOffsetMap,
+        boundaryOffsetSpace: extracted.unsupportedSemantic
+            ? "unsupported"
+            : extracted.hasMarkup
+                ? "raw-ssml"
+                : "plain-spoken-text",
+        unsupportedSemantic: extracted.unsupportedSemantic,
+        hasMarkup: extracted.hasMarkup,
+    };
+};
+const normalizeBoundaryWord = (word) => {
+    if (!word)
+        return null;
+    if (/^<[^>]+>$/.test(word.trim()))
+        return null;
+    const match = word.match(TOKEN_PATTERN);
+    if (!match?.[0])
+        return null;
+    return normalizedTokenValue(match[0]);
+};
+const findTokenAtOffset = (tokens, offset) => tokens.find((token) => token.start <= offset && offset < token.end) || null;
+const mapRawOffsetToSpokenOffset = (tokenization, position) => {
+    const direct = tokenization.rawToSpokenOffsetMap.get(position);
+    if (direct !== undefined)
+        return direct;
+    for (let offset = position; offset < tokenization.speechText.length; offset++) {
+        const mapped = tokenization.rawToSpokenOffsetMap.get(offset);
+        if (mapped !== undefined)
+            return mapped;
+        if (tokenization.speechText[offset] === ">")
+            break;
+    }
+    return null;
+};
+const candidateForOffset = (tokenization, offset, length) => {
+    if (offset === null)
+        return null;
+    const token = findTokenAtOffset(tokenization.tokens, offset);
+    if (!token)
+        return null;
+    return {
+        token,
+        start: offset,
+        length: Math.max(1, Math.min(length, token.end - offset)),
+        confidence: 0.75,
+    };
+};
+const candidateMatchesWord = (candidate, boundaryWord) => {
+    const normalizedBoundary = normalizeBoundaryWord(boundaryWord);
+    return Boolean(candidate &&
+        normalizedBoundary &&
+        candidate.token.normalized === normalizedBoundary);
+};
+export const resolveBoundaryToSpeechToken = (args) => {
+    if (args.tokenization.boundaryOffsetSpace === "unsupported" ||
+        !Number.isFinite(args.position) ||
+        (args.boundaryWord && /^<[^>]+>$/.test(args.boundaryWord.trim()))) {
+        return null;
+    }
+    const safeLength = Math.max(1, Number.isFinite(args.length) ? args.length : 1);
+    const rawCandidate = candidateForOffset(args.tokenization, mapRawOffsetToSpokenOffset(args.tokenization, args.position), safeLength);
+    const plainCandidate = candidateForOffset(args.tokenization, args.position >= 0 && args.position < args.tokenization.spokenText.length
+        ? args.position
+        : null, safeLength);
+    if (candidateMatchesWord(rawCandidate, args.boundaryWord)) {
+        return { ...rawCandidate, confidence: 1 };
+    }
+    if (candidateMatchesWord(plainCandidate, args.boundaryWord)) {
+        return { ...plainCandidate, confidence: 1 };
+    }
+    if (args.boundaryWord && normalizeBoundaryWord(args.boundaryWord)) {
+        return null;
+    }
+    return args.tokenization.boundaryOffsetSpace === "raw-ssml"
+        ? rawCandidate || plainCandidate
+        : plainCandidate || rawCandidate;
+};
+export const resolveSpokenOffsetToSpeechToken = (args) => {
+    if (args.tokenization.boundaryOffsetSpace === "unsupported" ||
+        !Number.isFinite(args.position) ||
+        (args.boundaryWord && /^<[^>]+>$/.test(args.boundaryWord.trim()))) {
+        return null;
+    }
+    const safeLength = Math.max(1, Number.isFinite(args.length) ? args.length : 1);
+    const spokenCandidate = candidateForOffset(args.tokenization, args.position >= 0 && args.position < args.tokenization.spokenText.length
+        ? args.position
+        : null, safeLength);
+    if (candidateMatchesWord(spokenCandidate, args.boundaryWord)) {
+        return { ...spokenCandidate, confidence: 1 };
+    }
+    if (args.boundaryWord && normalizeBoundaryWord(args.boundaryWord)) {
+        return null;
+    }
+    return spokenCandidate;
+};
+//# sourceMappingURL=speech-tokenizer.js.map

package/dist/services/tts/math-alignment/speech-tokenizer.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"speech-tokenizer.js","sourceRoot":"","sources":["../../../../src/services/tts/math-alignment/speech-tokenizer.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,iBAAiB,EAAE,MAAM,wBAAwB,CAAC;AAK3D,OAAO,EAAE,iCAAiC,EAAE,MAAM,uBAAuB,CAAC;AAmB1E,8EAA8E;AAC9E,wDAAwD;AACxD,wCAAwC;AACxC,MAAM,aAAa,GAAG,iCAAiC,EAAE,CAAC;AAE1D,MAAM,aAAa,GAAG,IAAI,GAAG,CAAiB;IAC7C,CAAC,MAAM,EAAE,GAAG,CAAC;IACb,CAAC,KAAK,EAAE,GAAG,CAAC;IACZ,CAAC,KAAK,EAAE,GAAG,CAAC;IACZ,CAAC,OAAO,EAAE,GAAG,CAAC;IACd,CAAC,MAAM,EAAE,GAAG,CAAC;IACb,CAAC,MAAM,EAAE,GAAG,CAAC;IACb,CAAC,KAAK,EAAE,GAAG,CAAC;IACZ,CAAC,OAAO,EAAE,GAAG,CAAC;IACd,CAAC,OAAO,EAAE,GAAG,CAAC;IACd,CAAC,MAAM,EAAE,GAAG,CAAC;IACb,CAAC,KAAK,EAAE,IAAI,CAAC;CACb,CAAC,CAAC;AAEH,MAAM,oBAAoB,GAAG,CAAC,KAAa,EAAU,EAAE;IACtD,MAAM,UAAU,GAAG,KAAK,CAAC,WAAW,EAAE,CAAC;IACvC,OAAO,aAAa,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,UAAU,CAAC;AACpD,CAAC,CAAC;AAEF,MAAM,kBAAkB,GAAG,CAAC,UAAkB,EAA0B,EAAE;IACzE,MAAM,MAAM,GAA2B,EAAE,CAAC;IAC1C,KAAK,MAAM,KAAK,IAAI,UAAU,CAAC,QAAQ,CAAC,aAAa,CAAC,EAAE,CAAC;QACxD,MAAM,IAAI,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC;QACtB,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,IAAI,CAAC,CAAC;QAC/B,MAAM,CAAC,IAAI,CAAC;YACX,EAAE,EAAE,gBAAgB,MAAM,CAAC,MAAM,EAAE;YACnC,IAAI;YACJ,UAAU,EAAE,oBAAoB,CAAC,IAAI,CAAC;YACtC,KAAK;YACL,GAAG,EAAE,KAAK,GAAG,IAAI,CAAC,MAAM;YACxB,WAAW,EAAE,KAAK;YAClB,SAAS,EAAE,KAAK,GAAG,IAAI,CAAC,MAAM;YAC9B,gBAAgB,EAAE,mBAAmB;SACrC,CAAC,CAAC;IACJ,CAAC;IACD,OAAO,MAAM,CAAC;AACf,CAAC,CAAC;AAEF,MAAM,CAAC,MAAM,oBAAoB,GAAG,CAAC,IAEpC,EAA4B,EAAE;IAC9B,MAAM,SAAS,GAAG,iBAAiB,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;IACrD,OAAO;QACN,UAAU,EAAE,IAAI,CAAC,UAAU;QAC3B,UAAU,EAAE,SAAS,CAAC,UAAU;QAChC,MAAM,EAAE,kBAAkB,CAAC,SAAS,CAAC,UAAU,CAAC;QAChD,oBAAoB,EAAE,SAAS,CAAC,oBAAoB;QACpD,mBAAmB,EAAE,SAAS,CAAC,mBAAmB;YACjD,CAAC,CAAC,aAAa;YACf,CAAC,CAAC,SAAS,CAAC,SAAS;gBACpB,CAAC,CAAC,UAAU;gBACZ,CAAC,CAAC,mBAAmB;QACvB,mBAAmB,EAAE,SAAS,CAAC,mBAAmB;QAClD,SAAS,EAAE,SAAS,CAAC,SAAS;KAC9B,CAAC;AACH,CAAC,CAAC;AAEF,MAAM,qBAAqB,GAAG,CAAC,IAAa,EAAiB,EAAE;IAC9D,IAAI,CAAC,IAAI;QAAE,OAAO,IAAI,CAAC;IACvB,IAAI,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC;QAAE,OAAO,IAAI,CAAC;IAC/C,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,aAAa,CAAC,CAAC;IACxC,IAAI,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC;QAAE,OAAO,IAAI,CAAC;IAC7B,OAAO,oBAAoB,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;AACvC,CAAC,CAAC;AAEF,MAAM,iBAAiB,GAAG,CACzB,MAA8B,EAC9B,MAAc,EACgB,EAAE,CAChC,MAAM,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,KAAK,IAAI,MAAM,IAAI,MAAM,GAAG,KAAK,CAAC,GAAG,CAAC,IAAI,IAAI,CAAC;AAE7E,MAAM,0BAA0B,GAAG,CAClC,YAAsC,EACtC,QAAgB,EACA,EAAE;IAClB,MAAM,MAAM,GAAG,YAAY,CAAC,oBAAoB,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;IAC/D,IAAI,MAAM,KAAK,SAAS;QAAE,OAAO,MAAM,CAAC;IACxC,KAAK,IAAI,MAAM,GAAG,QAAQ,EAAE,MAAM,GAAG,YAAY,CAAC,UAAU,CAAC,MAAM,EAAE,MAAM,EAAE,EAAE,CAAC;QAC/E,MAAM,MAAM,GAAG,YAAY,CAAC,oBAAoB,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;QAC7D,IAAI,MAAM,KAAK,SAAS;YAAE,OAAO,MAAM,CAAC;QACxC,IAAI,YAAY,CAAC,UAAU,CAAC,MAAM,CAAC,KAAK,GAAG;YAAE,MAAM;IACpD,CAAC;IACD,OAAO,IAAI,CAAC;AACb,CAAC,CAAC;AAEF,MAAM,kBAAkB,GAAG,CAC1B,YAAsC,EACtC,MAAqB,EACrB,MAAc,EACkB,EAAE;IAClC,IAAI,MAAM,KAAK,IAAI;QAAE,OAAO,IAAI,CAAC;IACjC,MAAM,KAAK,GAAG,iBAAiB,CAAC,YAAY,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAC7D,IAAI,CAAC,KAAK;QAAE,OAAO,IAAI,CAAC;IACxB,OAAO;QACN,KAAK;QACL,KAAK,EAAE,MAAM;QACb,MAAM,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,MAAM,EAAE,KAAK,CAAC,GAAG,GAAG,MAAM,CAAC,CAAC;QACzD,UAAU,EAAE,IAAI;KAChB,CAAC;AACH,CAAC,CAAC;AAEF,MAAM,oBAAoB,GAAG,CAC5B,SAAwC,EACxC,YAAqB,EACX,EAAE;IACZ,MAAM,kBAAkB,GAAG,qBAAqB,CAAC,YAAY,CAAC,CAAC;IAC/D,OAAO,OAAO,CACb,SAAS;QACR,kBAAkB;QAClB,SAAS,CAAC,KAAK,CAAC,UAAU,KAAK,kBAAkB,CAClD,CAAC;AACH,CAAC,CAAC;AAEF,MAAM,CAAC,MAAM,4BAA4B,GAAG,CAAC,IAK5C,EAAiC,EAAE;IACnC,IACC,IAAI,CAAC,YAAY,CAAC,mBAAmB,KAAK,aAAa;QACvD,CAAC,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC;QAC/B,CAAC,IAAI,CAAC,YAAY,IAAI,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,YAAY,CAAC,IAAI,EAAE,CAAC,CAAC,EAChE,CAAC;QACF,OAAO,IAAI,CAAC;IACb,CAAC;IACD,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,MAAO,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IAChF,MAAM,YAAY,GAAG,kBAAkB,CACtC,IAAI,CAAC,YAAY,EACjB,0BAA0B,CAAC,IAAI,CAAC,YAAY,EAAE,IAAI,CAAC,QAAQ,CAAC,EAC5D,UAAU,CACV,CAAC;IACF,MAAM,cAAc,GAAG,kBAAkB,CACxC,IAAI,CAAC,YAAY,EACjB,IAAI,CAAC,QAAQ,IAAI,CAAC,IAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,YAAY,CAAC,UAAU,CAAC,MAAM;QACxE,CAAC,CAAC,IAAI,CAAC,QAAQ;QACf,CAAC,CAAC,IAAI,EACP,UAAU,CACV,CAAC;IAEF,IAAI,oBAAoB,CAAC,YAAY,EAAE,IAAI,CAAC,YAAY,CAAC,EAAE,CAAC;QAC3D,OAAO,EAAE,GAAG,YAAa,EAAE,UAAU,EAAE,CAAC,EAAE,CAAC;IAC5C,CAAC;IACD,IAAI,oBAAoB,CAAC,cAAc,EAAE,IAAI,CAAC,YAAY,CAAC,EAAE,CAAC;QAC7D,OAAO,EAAE,GAAG,cAAe,EAAE,UAAU,EAAE,CAAC,EAAE,CAAC;IAC9C,CAAC;IACD,IAAI,IAAI,CAAC,YAAY,IAAI,qBAAqB,CAAC,IAAI,CAAC,YAAY,CAAC,EAAE,CAAC;QACnE,OAAO,IAAI,CAAC;IACb,CAAC;IAED,OAAO,IAAI,CAAC,YAAY,CAAC,mBAAmB,KAAK,UAAU;QAC1D,CAAC,CAAC,YAAY,IAAI,cAAc;QAChC,CAAC,CAAC,cAAc,IAAI,YAAY,CAAC;AACnC,CAAC,CAAC;AAEF,MAAM,CAAC,MAAM,gCAAgC,GAAG,CAAC,IAKhD,EAAiC,EAAE;IACnC,IACC,IAAI,CAAC,YAAY,CAAC,mBAAmB,KAAK,aAAa;QACvD,CAAC,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC;QAC/B,CAAC,IAAI,CAAC,YAAY,IAAI,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,YAAY,CAAC,IAAI,EAAE,CAAC,CAAC,EAChE,CAAC;QACF,OAAO,IAAI,CAAC;IACb,CAAC;IACD,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,MAAO,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IAChF,MAAM,eAAe,GAAG,kBAAkB,CACzC,IAAI,CAAC,YAAY,EACjB,IAAI,CAAC,QAAQ,IAAI,CAAC,IAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,YAAY,CAAC,UAAU,CAAC,MAAM;QACxE,CAAC,CAAC,IAAI,CAAC,QAAQ;QACf,CAAC,CAAC,IAAI,EACP,UAAU,CACV,CAAC;IACF,IAAI,oBAAoB,CAAC,eAAe,EAAE,IAAI,CAAC,YAAY,CAAC,EAAE,CAAC;QAC9D,OAAO,EAAE,GAAG,eAAgB,EAAE,UAAU,EAAE,CAAC,EAAE,CAAC;IAC/C,CAAC;IACD,IAAI,IAAI,CAAC,YAAY,IAAI,qBAAqB,CAAC,IAAI,CAAC,YAAY,CAAC,EAAE,CAAC;QACnE,OAAO,IAAI,CAAC;IACb,CAAC;IACD,OAAO,eAAe,CAAC;AACxB,CAAC,CAAC","sourcesContent":["import { extractSpokenText } from \"../ssml/spoken-text.js\";\nimport type {\n\tBoundaryOffsetSpace,\n\tSpeechAlignmentToken,\n} from \"./types.js\";\nimport { createSpeechAlignmentTokenPattern } from \"../text-processing.js\";\n\nexport interface SpeechSourceTokenization {\n\tspeechText: string;\n\tspokenText: string;\n\ttokens: SpeechAlignmentToken[];\n\trawToSpokenOffsetMap: Map<number, number>;\n\tboundaryOffsetSpace: BoundaryOffsetSpace;\n\tunsupportedSemantic: boolean;\n\thasMarkup: boolean;\n}\n\nexport interface ResolvedSpeechBoundary {\n\ttoken: SpeechAlignmentToken;\n\tstart: number;\n\tlength: number;\n\tconfidence: number;\n}\n\n// Shared Unicode-aware tokenizer, identical to the catalog span aligner's, so\n// spoken text tokenizes the same way on both paths (see\n// `createSpeechAlignmentTokenPattern`).\nconst TOKEN_PATTERN = createSpeechAlignmentTokenPattern();\n\nconst NUMERIC_WORDS = new Map<string, string>([\n\t[\"zero\", \"0\"],\n\t[\"one\", \"1\"],\n\t[\"two\", \"2\"],\n\t[\"three\", \"3\"],\n\t[\"four\", \"4\"],\n\t[\"five\", \"5\"],\n\t[\"six\", \"6\"],\n\t[\"seven\", \"7\"],\n\t[\"eight\", \"8\"],\n\t[\"nine\", \"9\"],\n\t[\"ten\", \"10\"],\n]);\n\nconst normalizedTokenValue = (value: string): string => {\n\tconst normalized = value.toLowerCase();\n\treturn NUMERIC_WORDS.get(normalized) || normalized;\n};\n\nconst tokenizeSpokenText = (spokenText: string): SpeechAlignmentToken[] => {\n\tconst tokens: SpeechAlignmentToken[] = [];\n\tfor (const match of spokenText.matchAll(TOKEN_PATTERN)) {\n\t\tconst text = match[0];\n\t\tconst start = match.index ?? 0;\n\t\ttokens.push({\n\t\t\tid: `speech-token-${tokens.length}`,\n\t\t\ttext,\n\t\t\tnormalized: normalizedTokenValue(text),\n\t\t\tstart,\n\t\t\tend: start + text.length,\n\t\t\tsourceStart: start,\n\t\t\tsourceEnd: start + text.length,\n\t\t\tcoordinateSystem: \"normalized-speech\",\n\t\t});\n\t}\n\treturn tokens;\n};\n\nexport const tokenizeSpeechSource = (args: {\n\tspeechText: string;\n}): SpeechSourceTokenization => {\n\tconst extracted = extractSpokenText(args.speechText);\n\treturn {\n\t\tspeechText: args.speechText,\n\t\tspokenText: extracted.spokenText,\n\t\ttokens: tokenizeSpokenText(extracted.spokenText),\n\t\trawToSpokenOffsetMap: extracted.rawToSpokenOffsetMap,\n\t\tboundaryOffsetSpace: extracted.unsupportedSemantic\n\t\t\t? \"unsupported\"\n\t\t\t: extracted.hasMarkup\n\t\t\t\t? \"raw-ssml\"\n\t\t\t\t: \"plain-spoken-text\",\n\t\tunsupportedSemantic: extracted.unsupportedSemantic,\n\t\thasMarkup: extracted.hasMarkup,\n\t};\n};\n\nconst normalizeBoundaryWord = (word?: string): string | null => {\n\tif (!word) return null;\n\tif (/^<[^>]+>$/.test(word.trim())) return null;\n\tconst match = word.match(TOKEN_PATTERN);\n\tif (!match?.[0]) return null;\n\treturn normalizedTokenValue(match[0]);\n};\n\nconst findTokenAtOffset = (\n\ttokens: SpeechAlignmentToken[],\n\toffset: number,\n): SpeechAlignmentToken | null =>\n\ttokens.find((token) => token.start <= offset && offset < token.end) || null;\n\nconst mapRawOffsetToSpokenOffset = (\n\ttokenization: SpeechSourceTokenization,\n\tposition: number,\n): number | null => {\n\tconst direct = tokenization.rawToSpokenOffsetMap.get(position);\n\tif (direct !== undefined) return direct;\n\tfor (let offset = position; offset < tokenization.speechText.length; offset++) {\n\t\tconst mapped = tokenization.rawToSpokenOffsetMap.get(offset);\n\t\tif (mapped !== undefined) return mapped;\n\t\tif (tokenization.speechText[offset] === \">\") break;\n\t}\n\treturn null;\n};\n\nconst candidateForOffset = (\n\ttokenization: SpeechSourceTokenization,\n\toffset: number | null,\n\tlength: number,\n): ResolvedSpeechBoundary | null => {\n\tif (offset === null) return null;\n\tconst token = findTokenAtOffset(tokenization.tokens, offset);\n\tif (!token) return null;\n\treturn {\n\t\ttoken,\n\t\tstart: offset,\n\t\tlength: Math.max(1, Math.min(length, token.end - offset)),\n\t\tconfidence: 0.75,\n\t};\n};\n\nconst candidateMatchesWord = (\n\tcandidate: ResolvedSpeechBoundary | null,\n\tboundaryWord?: string,\n): boolean => {\n\tconst normalizedBoundary = normalizeBoundaryWord(boundaryWord);\n\treturn Boolean(\n\t\tcandidate &&\n\t\t\tnormalizedBoundary &&\n\t\t\tcandidate.token.normalized === normalizedBoundary,\n\t);\n};\n\nexport const resolveBoundaryToSpeechToken = (args: {\n\ttokenization: SpeechSourceTokenization;\n\tposition: number;\n\tlength?: number;\n\tboundaryWord?: string;\n}): ResolvedSpeechBoundary | null => {\n\tif (\n\t\targs.tokenization.boundaryOffsetSpace === \"unsupported\" ||\n\t\t!Number.isFinite(args.position) ||\n\t\t(args.boundaryWord && /^<[^>]+>$/.test(args.boundaryWord.trim()))\n\t) {\n\t\treturn null;\n\t}\n\tconst safeLength = Math.max(1, Number.isFinite(args.length) ? args.length! : 1);\n\tconst rawCandidate = candidateForOffset(\n\t\targs.tokenization,\n\t\tmapRawOffsetToSpokenOffset(args.tokenization, args.position),\n\t\tsafeLength,\n\t);\n\tconst plainCandidate = candidateForOffset(\n\t\targs.tokenization,\n\t\targs.position >= 0 && args.position < args.tokenization.spokenText.length\n\t\t\t? args.position\n\t\t\t: null,\n\t\tsafeLength,\n\t);\n\n\tif (candidateMatchesWord(rawCandidate, args.boundaryWord)) {\n\t\treturn { ...rawCandidate!, confidence: 1 };\n\t}\n\tif (candidateMatchesWord(plainCandidate, args.boundaryWord)) {\n\t\treturn { ...plainCandidate!, confidence: 1 };\n\t}\n\tif (args.boundaryWord && normalizeBoundaryWord(args.boundaryWord)) {\n\t\treturn null;\n\t}\n\n\treturn args.tokenization.boundaryOffsetSpace === \"raw-ssml\"\n\t\t? rawCandidate || plainCandidate\n\t\t: plainCandidate || rawCandidate;\n};\n\nexport const resolveSpokenOffsetToSpeechToken = (args: {\n\ttokenization: SpeechSourceTokenization;\n\tposition: number;\n\tlength?: number;\n\tboundaryWord?: string;\n}): ResolvedSpeechBoundary | null => {\n\tif (\n\t\targs.tokenization.boundaryOffsetSpace === \"unsupported\" ||\n\t\t!Number.isFinite(args.position) ||\n\t\t(args.boundaryWord && /^<[^>]+>$/.test(args.boundaryWord.trim()))\n\t) {\n\t\treturn null;\n\t}\n\tconst safeLength = Math.max(1, Number.isFinite(args.length) ? args.length! : 1);\n\tconst spokenCandidate = candidateForOffset(\n\t\targs.tokenization,\n\t\targs.position >= 0 && args.position < args.tokenization.spokenText.length\n\t\t\t? args.position\n\t\t\t: null,\n\t\tsafeLength,\n\t);\n\tif (candidateMatchesWord(spokenCandidate, args.boundaryWord)) {\n\t\treturn { ...spokenCandidate!, confidence: 1 };\n\t}\n\tif (args.boundaryWord && normalizeBoundaryWord(args.boundaryWord)) {\n\t\treturn null;\n\t}\n\treturn spokenCandidate;\n};\n"]}

package/dist/services/tts/math-alignment/types.d.ts

@@ -0,0 +1,58 @@
+export type HighlightTargetQuality = "exact-word" | "element-range" | "region-fallback";
+export type HighlightTarget = {
+    type: "text-range";
+    quality: "exact-word";
+    node: Text;
+    startOffset: number;
+    endOffset: number;
+} | {
+    type: "element-range";
+    quality: Exclude<HighlightTargetQuality, "exact-word">;
+    element: Element;
+} | {
+    type: "region";
+    quality: "region-fallback";
+    element: Element;
+};
+export type MathAlignmentTokenKind = "identifier" | "number" | "operator" | "text" | "space" | "structure" | "layout" | "unknown";
+export interface MathAlignmentToken {
+    id: string;
+    kind: MathAlignmentTokenKind;
+    role: string;
+    text: string;
+    normalized: string;
+    spokenAliases: string[];
+    sourceElement: Element;
+    path: string;
+    target: HighlightTarget;
+    optional?: boolean;
+}
+export interface MathMLTokenizationResult {
+    tokens: MathAlignmentToken[];
+    layoutTargets: MathAlignmentToken[];
+    expressionTarget: HighlightTarget;
+}
+export type BoundaryCoordinateSystem = "normalized-speech" | "raw-ssml" | "provider-speech-marks";
+export type BoundaryOffsetSpace = "plain-spoken-text" | "raw-ssml" | "unsupported";
+export interface SpeechAlignmentToken {
+    id: string;
+    text: string;
+    normalized: string;
+    start: number;
+    end: number;
+    sourceStart: number;
+    sourceEnd: number;
+    coordinateSystem: BoundaryCoordinateSystem;
+}
+export interface AlignmentSegment {
+    speechTokenIds: string[];
+    mathTokenIds: string[];
+    score: number;
+    confidence: number;
+    target: HighlightTarget;
+}
+export interface AlignmentResult {
+    segments: AlignmentSegment[];
+    confidence: number;
+    fallbackTarget: HighlightTarget;
+}

package/dist/services/tts/math-alignment/types.js

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=types.js.map

package/dist/services/tts/math-alignment/types.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"types.js","sourceRoot":"","sources":["../../../../src/services/tts/math-alignment/types.ts"],"names":[],"mappings":"","sourcesContent":["export type HighlightTargetQuality =\n\t| \"exact-word\"\n\t| \"element-range\"\n\t| \"region-fallback\";\n\nexport type HighlightTarget =\n\t| {\n\t\t\ttype: \"text-range\";\n\t\t\tquality: \"exact-word\";\n\t\t\tnode: Text;\n\t\t\tstartOffset: number;\n\t\t\tendOffset: number;\n\t  }\n\t| {\n\t\t\ttype: \"element-range\";\n\t\t\tquality: Exclude<HighlightTargetQuality, \"exact-word\">;\n\t\t\telement: Element;\n\t  }\n\t| {\n\t\t\ttype: \"region\";\n\t\t\tquality: \"region-fallback\";\n\t\t\telement: Element;\n\t  };\n\nexport type MathAlignmentTokenKind =\n\t| \"identifier\"\n\t| \"number\"\n\t| \"operator\"\n\t| \"text\"\n\t| \"space\"\n\t| \"structure\"\n\t| \"layout\"\n\t| \"unknown\";\n\nexport interface MathAlignmentToken {\n\tid: string;\n\tkind: MathAlignmentTokenKind;\n\trole: string;\n\ttext: string;\n\tnormalized: string;\n\tspokenAliases: string[];\n\tsourceElement: Element;\n\tpath: string;\n\ttarget: HighlightTarget;\n\toptional?: boolean;\n}\n\nexport interface MathMLTokenizationResult {\n\ttokens: MathAlignmentToken[];\n\tlayoutTargets: MathAlignmentToken[];\n\texpressionTarget: HighlightTarget;\n}\n\nexport type BoundaryCoordinateSystem =\n\t| \"normalized-speech\"\n\t| \"raw-ssml\"\n\t| \"provider-speech-marks\";\n\nexport type BoundaryOffsetSpace =\n\t| \"plain-spoken-text\"\n\t| \"raw-ssml\"\n\t| \"unsupported\";\n\nexport interface SpeechAlignmentToken {\n\tid: string;\n\ttext: string;\n\tnormalized: string;\n\tstart: number;\n\tend: number;\n\tsourceStart: number;\n\tsourceEnd: number;\n\tcoordinateSystem: BoundaryCoordinateSystem;\n}\n\nexport interface AlignmentSegment {\n\tspeechTokenIds: string[];\n\tmathTokenIds: string[];\n\tscore: number;\n\tconfidence: number;\n\ttarget: HighlightTarget;\n}\n\nexport interface AlignmentResult {\n\tsegments: AlignmentSegment[];\n\tconfidence: number;\n\tfallbackTarget: HighlightTarget;\n}\n"]}

package/dist/services/tts/math-aware-text-processing.d.ts

@@ -0,0 +1,18 @@
+import { type NormalizedTextMap, type TextProcessingOptions } from "./text-processing.js";
+export type MathAwareSpeechChunk = {
+    type: "text";
+    text: string;
+    sourceElement?: Element;
+} | {
+    type: "math";
+    mathml: string;
+    fallbackText: string;
+    sourceElement?: Element;
+};
+export interface MathAwareTextResult {
+    visibleText: string;
+    map: NormalizedTextMap;
+    chunks: MathAwareSpeechChunk[];
+    containsMathMarkup: boolean;
+}
+export declare const collectMathAwareTextAndMap: (element: Element, options?: TextProcessingOptions) => MathAwareTextResult;

package/dist/services/tts/math-aware-text-processing.js

@@ -0,0 +1,348 @@
+import { canonicalizeMathML } from "./mathml-sanitization.js";
+import { collectVisibleTextAndMap, isNodeHiddenForTTS, normalizeTextForSpeech, shouldInsertWordBoundarySpace, } from "./text-processing.js";
+const createAccumulator = (options) => ({
+    chars: [],
+    map: new Map(),
+    position: 0,
+    inLeadingWhitespace: true,
+    lastCharWasWhitespace: false,
+    lastMapped: null,
+    options,
+});
+const isMathJaxElement = (element) => element.classList?.contains("MathJax") ||
+    element.tagName?.toLowerCase() === "mjx-container";
+const isAssistiveMathElement = (element) => element.tagName?.toLowerCase() === "mjx-assistive-mml";
+const isAnnotationElement = (element) => {
+    const tagName = element.tagName?.toLowerCase();
+    return tagName === "annotation" || tagName === "annotation-xml";
+};
+const getDataMathML = (element) => {
+    const value = element.getAttribute("data-mathml");
+    return value && value.trim() ? value : null;
+};
+const findAssistiveMathML = (element) => {
+    const assistive = element.querySelector("mjx-assistive-mml math");
+    return assistive?.outerHTML || null;
+};
+const findCanonicalMathML = (element) => {
+    if (element.tagName.toLowerCase() === "math") {
+        return canonicalizeMathML(element.outerHTML);
+    }
+    const dataMath = getDataMathML(element);
+    if (dataMath) {
+        const canonical = canonicalizeMathML(dataMath);
+        if (canonical)
+            return canonical;
+    }
+    if (isMathJaxElement(element)) {
+        const assistive = findAssistiveMathML(element);
+        if (assistive) {
+            const canonical = canonicalizeMathML(assistive);
+            if (canonical)
+                return canonical;
+        }
+    }
+    return null;
+};
+const isMathContainer = (element) => Boolean(findCanonicalMathML(element));
+const TEXT_CHUNK_SOURCE_TAGS = new Set([
+    "P",
+    "H1",
+    "H2",
+    "H3",
+    "H4",
+    "H5",
+    "H6",
+    "LI",
+    "TD",
+    "TH",
+    "DD",
+    "DT",
+    "FIGCAPTION",
+    "LABEL",
+    "BUTTON",
+]);
+const TEXT_CHUNK_SOURCE_ROLES = new Set([
+    "heading",
+    "listitem",
+    "option",
+    "radio",
+    "row",
+    "cell",
+    "columnheader",
+    "rowheader",
+]);
+const resolveTextChunkSourceElement = (textNode, root) => {
+    let current = textNode.parentElement;
+    let best = null;
+    while (current && current !== root) {
+        const role = (current.getAttribute("role") || "").toLowerCase();
+        const tagName = current.tagName.toUpperCase();
+        if (TEXT_CHUNK_SOURCE_TAGS.has(tagName) ||
+            TEXT_CHUNK_SOURCE_ROLES.has(role)) {
+            best = current;
+            break;
+        }
+        current = current.parentElement;
+    }
+    return best || root;
+};
+const hasMathCandidate = (element) => Boolean(typeof element.querySelector === "function" &&
+    element.querySelector("math, [data-mathml], .MathJax, mjx-container, mjx-assistive-mml")) ||
+    element.tagName?.toLowerCase() === "math" ||
+    element.hasAttribute?.("data-mathml") ||
+    isMathJaxElement(element);
+const appendCharacter = (acc, character, mapping) => {
+    if (/[.,;:!?]/.test(character) &&
+        acc.chars.length > 0 &&
+        acc.chars[acc.chars.length - 1] === " ") {
+        acc.chars.pop();
+        acc.position--;
+        acc.map.delete(acc.position);
+    }
+    acc.chars.push(character);
+    if (mapping) {
+        acc.map.set(acc.position, mapping);
+        acc.lastMapped = mapping;
+    }
+    else if (acc.lastMapped) {
+        acc.map.set(acc.position, acc.lastMapped);
+    }
+    acc.position++;
+};
+const shouldInsertBoundarySpace = (acc, nextCharacter) => {
+    if (acc.inLeadingWhitespace ||
+        acc.lastCharWasWhitespace ||
+        acc.chars.length === 0) {
+        return false;
+    }
+    const previous = acc.chars[acc.chars.length - 1];
+    // Defer to the shared locale-aware decision (Intl.Segmenter, falling back to
+    // alnum) so adjacent text nodes split into words consistently with the main
+    // visible-text collector, instead of an ASCII-only heuristic.
+    return shouldInsertWordBoundarySpace(previous, nextCharacter, acc.options);
+};
+const appendTextNode = (acc, textNode) => {
+    const raw = textNode.textContent || "";
+    let appendedNonWhitespaceInNode = false;
+    for (let i = 0; i < raw.length; i++) {
+        const character = raw[i];
+        const isWhitespace = /\s/.test(character);
+        if (acc.inLeadingWhitespace) {
+            if (!isWhitespace) {
+                acc.inLeadingWhitespace = false;
+                appendCharacter(acc, character, { node: textNode, offset: i });
+                appendedNonWhitespaceInNode = true;
+                acc.lastCharWasWhitespace = false;
+            }
+            continue;
+        }
+        if (isWhitespace) {
+            if (!acc.lastCharWasWhitespace) {
+                appendCharacter(acc, " ", { node: textNode, offset: i });
+            }
+            acc.lastCharWasWhitespace = true;
+            continue;
+        }
+        if (!appendedNonWhitespaceInNode &&
+            shouldInsertBoundarySpace(acc, character)) {
+            appendCharacter(acc, " ", { node: textNode, offset: i });
+        }
+        appendCharacter(acc, character, { node: textNode, offset: i });
+        appendedNonWhitespaceInNode = true;
+        acc.lastCharWasWhitespace = false;
+    }
+};
+const appendCollectedText = (acc, collected) => {
+    for (let i = 0; i < collected.text.length; i++) {
+        const character = collected.text[i];
+        const isWhitespace = /\s/.test(character);
+        const mapping = collected.map.get(i) || null;
+        const syntheticNode = mapping?.node || acc.lastMapped?.node || null;
+        const syntheticOffset = mapping?.offset ?? acc.lastMapped?.offset ?? 0;
+        const normalizedMapping = mapping
+            ? { node: mapping.node, offset: mapping.offset }
+            : syntheticNode
+                ? { node: syntheticNode, offset: syntheticOffset }
+                : null;
+        if (acc.inLeadingWhitespace) {
+            if (isWhitespace)
+                continue;
+            acc.inLeadingWhitespace = false;
+        }
+        if (isWhitespace) {
+            if (!acc.lastCharWasWhitespace) {
+                appendCharacter(acc, " ", normalizedMapping);
+            }
+            acc.lastCharWasWhitespace = true;
+            continue;
+        }
+        if (i === 0 && shouldInsertBoundarySpace(acc, character)) {
+            appendCharacter(acc, " ", normalizedMapping);
+        }
+        appendCharacter(acc, character, normalizedMapping);
+        acc.lastCharWasWhitespace = false;
+        if (mapping) {
+            acc.map.set(acc.position - 1, {
+                node: mapping.node,
+                offset: mapping.offset,
+            });
+            acc.lastMapped = { node: mapping.node, offset: mapping.offset };
+        }
+        else if (syntheticNode) {
+            acc.map.set(acc.position - 1, {
+                node: syntheticNode,
+                offset: syntheticOffset,
+            });
+        }
+    }
+};
+const textFallbackFromMathML = (mathml) => {
+    if (typeof DOMParser === "undefined")
+        return "";
+    const parsed = new DOMParser().parseFromString(mathml, "application/xml");
+    if (parsed.getElementsByTagName("parsererror").length > 0)
+        return "";
+    const parts = [];
+    const visit = (node) => {
+        if (node.nodeType === 3) {
+            const text = node.textContent || "";
+            if (text.trim())
+                parts.push(text);
+            return;
+        }
+        if (node.nodeType !== 1)
+            return;
+        const element = node;
+        if (isAnnotationElement(element))
+            return;
+        for (const child of Array.from(element.childNodes)) {
+            visit(child);
+        }
+    };
+    visit(parsed.documentElement);
+    return normalizeTextForSpeech(parts.join(" "));
+};
+const collectVisibleMathFallback = (element, canonicalMathML, options) => {
+    const mathAcc = createAccumulator(options);
+    const visit = (node) => {
+        if (isNodeHiddenForTTS(node, element))
+            return;
+        if (node.nodeType === 3) {
+            appendTextNode(mathAcc, node);
+            return;
+        }
+        if (node.nodeType !== 1)
+            return;
+        const childElement = node;
+        if (isAssistiveMathElement(childElement) ||
+            isAnnotationElement(childElement)) {
+            return;
+        }
+        for (const child of Array.from(childElement.childNodes)) {
+            visit(child);
+        }
+    };
+    visit(element);
+    trimTrailingWhitespace(mathAcc);
+    const text = normalizeTextForSpeech(mathAcc.chars.join(""));
+    if (text) {
+        return { text, map: mathAcc.map };
+    }
+    const fallbackText = textFallbackFromMathML(canonicalMathML);
+    if (!fallbackText) {
+        return collectVisibleTextAndMap(element, options);
+    }
+    return {
+        text: fallbackText,
+        map: new Map(),
+    };
+};
+const trimTrailingWhitespace = (acc) => {
+    while (acc.chars.length > 0 && /\s/.test(acc.chars[acc.chars.length - 1])) {
+        acc.chars.pop();
+        acc.position--;
+        acc.map.delete(acc.position);
+    }
+};
+const collectMathAware = (root, options) => {
+    const acc = createAccumulator(options);
+    const chunks = [];
+    let textChunkStart = 0;
+    let textChunkSourceElement;
+    let containsMathMarkup = false;
+    const flushTextChunk = () => {
+        const text = normalizeTextForSpeech(acc.chars.slice(textChunkStart, acc.chars.length).join(""));
+        if (text) {
+            chunks.push({ type: "text", text, sourceElement: textChunkSourceElement });
+        }
+        textChunkStart = acc.chars.length;
+        textChunkSourceElement = undefined;
+    };
+    const processNode = (node) => {
+        if (isNodeHiddenForTTS(node, root))
+            return;
+        if (node.nodeType === Node.TEXT_NODE) {
+            const sourceElement = resolveTextChunkSourceElement(node, root);
+            if (textChunkSourceElement &&
+                sourceElement !== textChunkSourceElement) {
+                flushTextChunk();
+            }
+            textChunkSourceElement = sourceElement;
+            appendTextNode(acc, node);
+            return;
+        }
+        if (node.nodeType !== Node.ELEMENT_NODE)
+            return;
+        const element = node;
+        if (isAssistiveMathElement(element))
+            return;
+        const canonicalMathML = findCanonicalMathML(element);
+        if (canonicalMathML) {
+            flushTextChunk();
+            const collected = collectVisibleMathFallback(element, canonicalMathML, options);
+            appendCollectedText(acc, collected);
+            const fallbackText = normalizeTextForSpeech(collected.text);
+            chunks.push({
+                type: "math",
+                mathml: canonicalMathML,
+                fallbackText,
+                sourceElement: element,
+            });
+            textChunkStart = acc.chars.length;
+            containsMathMarkup = true;
+            return;
+        }
+        for (const child of Array.from(element.childNodes)) {
+            processNode(child);
+        }
+    };
+    for (const child of Array.from(root.childNodes)) {
+        processNode(child);
+    }
+    trimTrailingWhitespace(acc);
+    flushTextChunk();
+    const visibleText = acc.chars.join("");
+    while (acc.map.has(visibleText.length)) {
+        acc.map.delete(visibleText.length);
+    }
+    return {
+        visibleText,
+        map: acc.map,
+        chunks,
+        containsMathMarkup,
+    };
+};
+export const collectMathAwareTextAndMap = (element, options) => {
+    if (!hasMathCandidate(element)) {
+        const { text, map } = collectVisibleTextAndMap(element, options);
+        return {
+            visibleText: text,
+            map,
+            chunks: text ? [{ type: "text", text }] : [],
+            containsMathMarkup: false,
+        };
+    }
+    return collectMathAware(element, options);
+};
+//# sourceMappingURL=math-aware-text-processing.js.map

package/dist/services/tts/math-aware-text-processing.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"math-aware-text-processing.js","sourceRoot":"","sources":["../../../src/services/tts/math-aware-text-processing.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,MAAM,0BAA0B,CAAC;AAC9D,OAAO,EACN,wBAAwB,EAGxB,kBAAkB,EAClB,sBAAsB,EACtB,6BAA6B,GAC7B,MAAM,sBAAsB,CAAC;AAgC9B,MAAM,iBAAiB,GAAG,CACzB,OAA+B,EACb,EAAE,CAAC,CAAC;IACtB,KAAK,EAAE,EAAE;IACT,GAAG,EAAE,IAAI,GAAG,EAAE;IACd,QAAQ,EAAE,CAAC;IACX,mBAAmB,EAAE,IAAI;IACzB,qBAAqB,EAAE,KAAK;IAC5B,UAAU,EAAE,IAAI;IAChB,OAAO;CACP,CAAC,CAAC;AAEH,MAAM,gBAAgB,GAAG,CAAC,OAAgB,EAAW,EAAE,CACtD,OAAO,CAAC,SAAS,EAAE,QAAQ,CAAC,SAAS,CAAC;IACtC,OAAO,CAAC,OAAO,EAAE,WAAW,EAAE,KAAK,eAAe,CAAC;AAEpD,MAAM,sBAAsB,GAAG,CAAC,OAAgB,EAAW,EAAE,CAC5D,OAAO,CAAC,OAAO,EAAE,WAAW,EAAE,KAAK,mBAAmB,CAAC;AAExD,MAAM,mBAAmB,GAAG,CAAC,OAAgB,EAAW,EAAE;IACzD,MAAM,OAAO,GAAG,OAAO,CAAC,OAAO,EAAE,WAAW,EAAE,CAAC;IAC/C,OAAO,OAAO,KAAK,YAAY,IAAI,OAAO,KAAK,gBAAgB,CAAC;AACjE,CAAC,CAAC;AAEF,MAAM,aAAa,GAAG,CAAC,OAAgB,EAAiB,EAAE;IACzD,MAAM,KAAK,GAAG,OAAO,CAAC,YAAY,CAAC,aAAa,CAAC,CAAC;IAClD,OAAO,KAAK,IAAI,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC;AAC7C,CAAC,CAAC;AAEF,MAAM,mBAAmB,GAAG,CAAC,OAAgB,EAAiB,EAAE;IAC/D,MAAM,SAAS,GAAG,OAAO,CAAC,aAAa,CAAC,wBAAwB,CAAC,CAAC;IAClE,OAAO,SAAS,EAAE,SAAS,IAAI,IAAI,CAAC;AACrC,CAAC,CAAC;AAEF,MAAM,mBAAmB,GAAG,CAAC,OAAgB,EAAiB,EAAE;IAC/D,IAAI,OAAO,CAAC,OAAO,CAAC,WAAW,EAAE,KAAK,MAAM,EAAE,CAAC;QAC9C,OAAO,kBAAkB,CAAC,OAAO,CAAC,SAAS,CAAC,CAAC;IAC9C,CAAC;IACD,MAAM,QAAQ,GAAG,aAAa,CAAC,OAAO,CAAC,CAAC;IACxC,IAAI,QAAQ,EAAE,CAAC;QACd,MAAM,SAAS,GAAG,kBAAkB,CAAC,QAAQ,CAAC,CAAC;QAC/C,IAAI,SAAS;YAAE,OAAO,SAAS,CAAC;IACjC,CAAC;IACD,IAAI,gBAAgB,CAAC,OAAO,CAAC,EAAE,CAAC;QAC/B,MAAM,SAAS,GAAG,mBAAmB,CAAC,OAAO,CAAC,CAAC;QAC/C,IAAI,SAAS,EAAE,CAAC;YACf,MAAM,SAAS,GAAG,kBAAkB,CAAC,SAAS,CAAC,CAAC;YAChD,IAAI,SAAS;gBAAE,OAAO,SAAS,CAAC;QACjC,CAAC;IACF,CAAC;IACD,OAAO,IAAI,CAAC;AACb,CAAC,CAAC;AAEF,MAAM,eAAe,GAAG,CAAC,OAAgB,EAAW,EAAE,CACrD,OAAO,CAAC,mBAAmB,CAAC,OAAO,CAAC,CAAC,CAAC;AAEvC,MAAM,sBAAsB,GAAG,IAAI,GAAG,CAAC;IACtC,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,YAAY;IACZ,OAAO;IACP,QAAQ;CACR,CAAC,CAAC;AAEH,MAAM,uBAAuB,GAAG,IAAI,GAAG,CAAC;IACvC,SAAS;IACT,UAAU;IACV,QAAQ;IACR,OAAO;IACP,KAAK;IACL,MAAM;IACN,cAAc;IACd,WAAW;CACX,CAAC,CAAC;AAEH,MAAM,6BAA6B,GAAG,CACrC,QAAc,EACd,IAAa,EACH,EAAE;IACZ,IAAI,OAAO,GAAG,QAAQ,CAAC,aAAa,CAAC;IACrC,IAAI,IAAI,GAAmB,IAAI,CAAC;IAChC,OAAO,OAAO,IAAI,OAAO,KAAK,IAAI,EAAE,CAAC;QACpC,MAAM,IAAI,GAAG,CAAC,OAAO,CAAC,YAAY,CAAC,MAAM,CAAC,IAAI,EAAE,CAAC,CAAC,WAAW,EAAE,CAAC;QAChE,MAAM,OAAO,GAAG,OAAO,CAAC,OAAO,CAAC,WAAW,EAAE,CAAC;QAC9C,IACC,sBAAsB,CAAC,GAAG,CAAC,OAAO,CAAC;YACnC,uBAAuB,CAAC,GAAG,CAAC,IAAI,CAAC,EAChC,CAAC;YACF,IAAI,GAAG,OAAO,CAAC;YACf,MAAM;QACP,CAAC;QACD,OAAO,GAAG,OAAO,CAAC,aAAa,CAAC;IACjC,CAAC;IACD,OAAO,IAAI,IAAI,IAAI,CAAC;AACrB,CAAC,CAAC;AAEF,MAAM,gBAAgB,GAAG,CAAC,OAAgB,EAAW,EAAE,CACtD,OAAO,CACN,OAAO,OAAO,CAAC,aAAa,KAAK,UAAU;IAC1C,OAAO,CAAC,aAAa,CACpB,iEAAiE,CACjE,CACF;IACD,OAAO,CAAC,OAAO,EAAE,WAAW,EAAE,KAAK,MAAM;IACzC,OAAO,CAAC,YAAY,EAAE,CAAC,aAAa,CAAC;IACrC,gBAAgB,CAAC,OAAO,CAAC,CAAC;AAE3B,MAAM,eAAe,GAAG,CACvB,GAAoB,EACpB,SAAiB,EACjB,OAA8C,EACvC,EAAE;IACT,IACC,UAAU,CAAC,IAAI,CAAC,SAAS,CAAC;QAC1B,GAAG,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC;QACpB,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,KAAK,GAAG,EACtC,CAAC;QACF,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC;QAChB,GAAG,CAAC,QAAQ,EAAE,CAAC;QACf,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;IAC9B,CAAC;IACD,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;IAC1B,IAAI,OAAO,EAAE,CAAC;QACb,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAO,CAAC,CAAC;QACnC,GAAG,CAAC,UAAU,GAAG,OAAO,CAAC;IAC1B,CAAC;SAAM,IAAI,GAAG,CAAC,UAAU,EAAE,CAAC;QAC3B,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,QAAQ,EAAE,GAAG,CAAC,UAAU,CAAC,CAAC;IAC3C,CAAC;IACD,GAAG,CAAC,QAAQ,EAAE,CAAC;AAChB,CAAC,CAAC;AAEF,MAAM,yBAAyB,GAAG,CACjC,GAAoB,EACpB,aAAqB,EACX,EAAE;IACZ,IACC,GAAG,CAAC,mBAAmB;QACvB,GAAG,CAAC,qBAAqB;QACzB,GAAG,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EACrB,CAAC;QACF,OAAO,KAAK,CAAC;IACd,CAAC;IACD,MAAM,QAAQ,GAAG,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IACjD,6EAA6E;IAC7E,4EAA4E;IAC5E,8DAA8D;IAC9D,OAAO,6BAA6B,CAAC,QAAQ,EAAE,aAAa,EAAE,GAAG,CAAC,OAAO,CAAC,CAAC;AAC5E,CAAC,CAAC;AAEF,MAAM,cAAc,GAAG,CAAC,GAAoB,EAAE,QAAc,EAAQ,EAAE;IACrE,MAAM,GAAG,GAAG,QAAQ,CAAC,WAAW,IAAI,EAAE,CAAC;IACvC,IAAI,2BAA2B,GAAG,KAAK,CAAC;IACxC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,GAAG,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QACrC,MAAM,SAAS,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC;QACzB,MAAM,YAAY,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;QAC1C,IAAI,GAAG,CAAC,mBAAmB,EAAE,CAAC;YAC7B,IAAI,CAAC,YAAY,EAAE,CAAC;gBACnB,GAAG,CAAC,mBAAmB,GAAG,KAAK,CAAC;gBAChC,eAAe,CAAC,GAAG,EAAE,SAAS,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,CAAC,EAAE,CAAC,CAAC;gBAC/D,2BAA2B,GAAG,IAAI,CAAC;gBACnC,GAAG,CAAC,qBAAqB,GAAG,KAAK,CAAC;YACnC,CAAC;YACD,SAAS;QACV,CAAC;QACD,IAAI,YAAY,EAAE,CAAC;YAClB,IAAI,CAAC,GAAG,CAAC,qBAAqB,EAAE,CAAC;gBAChC,eAAe,CAAC,GAAG,EAAE,GAAG,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,CAAC,EAAE,CAAC,CAAC;YAC1D,CAAC;YACD,GAAG,CAAC,qBAAqB,GAAG,IAAI,CAAC;YACjC,SAAS;QACV,CAAC;QACD,IACC,CAAC,2BAA2B;YAC5B,yBAAyB,CAAC,GAAG,EAAE,SAAS,CAAC,EACxC,CAAC;YACF,eAAe,CAAC,GAAG,EAAE,GAAG,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,CAAC,EAAE,CAAC,CAAC;QAC1D,CAAC;QACD,eAAe,CAAC,GAAG,EAAE,SAAS,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,CAAC,EAAE,CAAC,CAAC;QAC/D,2BAA2B,GAAG,IAAI,CAAC;QACnC,GAAG,CAAC,qBAAqB,GAAG,KAAK,CAAC;IACnC,CAAC;AACF,CAAC,CAAC;AAEF,MAAM,mBAAmB,GAAG,CAC3B,GAAoB,EACpB,SAAmD,EAC5C,EAAE;IACT,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,SAAS,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QAChD,MAAM,SAAS,GAAG,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;QACpC,MAAM,YAAY,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;QAC1C,MAAM,OAAO,GAAG,SAAS,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,IAAI,CAAC;QAC7C,MAAM,aAAa,GAAG,OAAO,EAAE,IAAI,IAAI,GAAG,CAAC,UAAU,EAAE,IAAI,IAAI,IAAI,CAAC;QACpE,MAAM,eAAe,GAAG,OAAO,EAAE,MAAM,IAAI,GAAG,CAAC,UAAU,EAAE,MAAM,IAAI,CAAC,CAAC;QACvE,MAAM,iBAAiB,GAAG,OAAO;YAChC,CAAC,CAAC,EAAE,IAAI,EAAE,OAAO,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE;YAChD,CAAC,CAAC,aAAa;gBACd,CAAC,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,MAAM,EAAE,eAAe,EAAE;gBAClD,CAAC,CAAC,IAAI,CAAC;QACT,IAAI,GAAG,CAAC,mBAAmB,EAAE,CAAC;YAC7B,IAAI,YAAY;gBAAE,SAAS;YAC3B,GAAG,CAAC,mBAAmB,GAAG,KAAK,CAAC;QACjC,CAAC;QACD,IAAI,YAAY,EAAE,CAAC;YAClB,IAAI,CAAC,GAAG,CAAC,qBAAqB,EAAE,CAAC;gBAChC,eAAe,CAAC,GAAG,EAAE,GAAG,EAAE,iBAAiB,CAAC,CAAC;YAC9C,CAAC;YACD,GAAG,CAAC,qBAAqB,GAAG,IAAI,CAAC;YACjC,SAAS;QACV,CAAC;QACD,IAAI,CAAC,KAAK,CAAC,IAAI,yBAAyB,CAAC,GAAG,EAAE,SAAS,CAAC,EAAE,CAAC;YAC1D,eAAe,CAAC,GAAG,EAAE,GAAG,EAAE,iBAAiB,CAAC,CAAC;QAC9C,CAAC;QACD,eAAe,CAAC,GAAG,EAAE,SAAS,EAAE,iBAAiB,CAAC,CAAC;QACnD,GAAG,CAAC,qBAAqB,GAAG,KAAK,CAAC;QAClC,IAAI,OAAO,EAAE,CAAC;YACb,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,QAAQ,GAAG,CAAC,EAAE;gBAC7B,IAAI,EAAE,OAAO,CAAC,IAAI;gBAClB,MAAM,EAAE,OAAO,CAAC,MAAM;aACtB,CAAC,CAAC;YACH,GAAG,CAAC,UAAU,GAAG,EAAE,IAAI,EAAE,OAAO,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAC;QACjE,CAAC;aAAM,IAAI,aAAa,EAAE,CAAC;YAC1B,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,QAAQ,GAAG,CAAC,EAAE;gBAC7B,IAAI,EAAE,aAAa;gBACnB,MAAM,EAAE,eAAe;aACvB,CAAC,CAAC;QACJ,CAAC;IACF,CAAC;AACF,CAAC,CAAC;AAEF,MAAM,sBAAsB,GAAG,CAAC,MAAc,EAAU,EAAE;IACzD,IAAI,OAAO,SAAS,KAAK,WAAW;QAAE,OAAO,EAAE,CAAC;IAChD,MAAM,MAAM,GAAG,IAAI,SAAS,EAAE,CAAC,eAAe,CAAC,MAAM,EAAE,iBAAiB,CAAC,CAAC;IAC1E,IAAI,MAAM,CAAC,oBAAoB,CAAC,aAAa,CAAC,CAAC,MAAM,GAAG,CAAC;QAAE,OAAO,EAAE,CAAC;IACrE,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,MAAM,KAAK,GAAG,CAAC,IAAU,EAAQ,EAAE;QAClC,IAAI,IAAI,CAAC,QAAQ,KAAK,CAAC,EAAE,CAAC;YACzB,MAAM,IAAI,GAAG,IAAI,CAAC,WAAW,IAAI,EAAE,CAAC;YACpC,IAAI,IAAI,CAAC,IAAI,EAAE;gBAAE,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YAClC,OAAO;QACR,CAAC;QACD,IAAI,IAAI,CAAC,QAAQ,KAAK,CAAC;YAAE,OAAO;QAChC,MAAM,OAAO,GAAG,IAAe,CAAC;QAChC,IAAI,mBAAmB,CAAC,OAAO,CAAC;YAAE,OAAO;QACzC,KAAK,MAAM,KAAK,IAAI,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,UAAU,CAAC,EAAE,CAAC;YACpD,KAAK,CAAC,KAAK,CAAC,CAAC;QACd,CAAC;IACF,CAAC,CAAC;IACF,KAAK,CAAC,MAAM,CAAC,eAAe,CAAC,CAAC;IAC9B,OAAO,sBAAsB,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC;AAChD,CAAC,CAAC;AAEF,MAAM,0BAA0B,GAAG,CAClC,OAAgB,EAChB,eAAuB,EACvB,OAA+B,EACY,EAAE;IAC7C,MAAM,OAAO,GAAG,iBAAiB,CAAC,OAAO,CAAC,CAAC;IAC3C,MAAM,KAAK,GAAG,CAAC,IAAU,EAAQ,EAAE;QAClC,IAAI,kBAAkB,CAAC,IAAI,EAAE,OAAO,CAAC;YAAE,OAAO;QAC9C,IAAI,IAAI,CAAC,QAAQ,KAAK,CAAC,EAAE,CAAC;YACzB,cAAc,CAAC,OAAO,EAAE,IAAY,CAAC,CAAC;YACtC,OAAO;QACR,CAAC;QACD,IAAI,IAAI,CAAC,QAAQ,KAAK,CAAC;YAAE,OAAO;QAChC,MAAM,YAAY,GAAG,IAAe,CAAC;QACrC,IACC,sBAAsB,CAAC,YAAY,CAAC;YACpC,mBAAmB,CAAC,YAAY,CAAC,EAChC,CAAC;YACF,OAAO;QACR,CAAC;QACD,KAAK,MAAM,KAAK,IAAI,KAAK,CAAC,IAAI,CAAC,YAAY,CAAC,UAAU,CAAC,EAAE,CAAC;YACzD,KAAK,CAAC,KAAK,CAAC,CAAC;QACd,CAAC;IACF,CAAC,CAAC;IACF,KAAK,CAAC,OAAO,CAAC,CAAC;IACf,sBAAsB,CAAC,OAAO,CAAC,CAAC;IAChC,MAAM,IAAI,GAAG,sBAAsB,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC;IAC5D,IAAI,IAAI,EAAE,CAAC;QACV,OAAO,EAAE,IAAI,EAAE,GAAG,EAAE,OAAO,CAAC,GAAG,EAAE,CAAC;IACnC,CAAC;IACD,MAAM,YAAY,GAAG,sBAAsB,CAAC,eAAe,CAAC,CAAC;IAC7D,IAAI,CAAC,YAAY,EAAE,CAAC;QACnB,OAAO,wBAAwB,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;IACnD,CAAC;IACD,OAAO;QACN,IAAI,EAAE,YAAY;QAClB,GAAG,EAAE,IAAI,GAAG,EAAE;KACd,CAAC;AACH,CAAC,CAAC;AAEF,MAAM,sBAAsB,GAAG,CAAC,GAAoB,EAAQ,EAAE;IAC7D,OAAO,GAAG,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,IAAI,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC;QAC3E,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC;QAChB,GAAG,CAAC,QAAQ,EAAE,CAAC;QACf,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;IAC9B,CAAC;AACF,CAAC,CAAC;AAEF,MAAM,gBAAgB,GAAG,CACxB,IAAa,EACb,OAA+B,EACT,EAAE;IACxB,MAAM,GAAG,GAAG,iBAAiB,CAAC,OAAO,CAAC,CAAC;IACvC,MAAM,MAAM,GAA2B,EAAE,CAAC;IAC1C,IAAI,cAAc,GAAG,CAAC,CAAC;IACvB,IAAI,sBAA2C,CAAC;IAChD,IAAI,kBAAkB,GAAG,KAAK,CAAC;IAE/B,MAAM,cAAc,GAAG,GAAG,EAAE;QAC3B,MAAM,IAAI,GAAG,sBAAsB,CAClC,GAAG,CAAC,KAAK,CAAC,KAAK,CAAC,cAAc,EAAE,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAC1D,CAAC;QACF,IAAI,IAAI,EAAE,CAAC;YACV,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,aAAa,EAAE,sBAAsB,EAAE,CAAC,CAAC;QAC5E,CAAC;QACD,cAAc,GAAG,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC;QAClC,sBAAsB,GAAG,SAAS,CAAC;IACpC,CAAC,CAAC;IAEF,MAAM,WAAW,GAAG,CAAC,IAAU,EAAQ,EAAE;QACxC,IAAI,kBAAkB,CAAC,IAAI,EAAE,IAAI,CAAC;YAAE,OAAO;QAC3C,IAAI,IAAI,CAAC,QAAQ,KAAK,IAAI,CAAC,SAAS,EAAE,CAAC;YACtC,MAAM,aAAa,GAAG,6BAA6B,CAAC,IAAY,EAAE,IAAI,CAAC,CAAC;YACxE,IACC,sBAAsB;gBACtB,aAAa,KAAK,sBAAsB,EACvC,CAAC;gBACF,cAAc,EAAE,CAAC;YAClB,CAAC;YACD,sBAAsB,GAAG,aAAa,CAAC;YACvC,cAAc,CAAC,GAAG,EAAE,IAAY,CAAC,CAAC;YAClC,OAAO;QACR,CAAC;QACD,IAAI,IAAI,CAAC,QAAQ,KAAK,IAAI,CAAC,YAAY;YAAE,OAAO;QAChD,MAAM,OAAO,GAAG,IAAe,CAAC;QAChC,IAAI,sBAAsB,CAAC,OAAO,CAAC;YAAE,OAAO;QAC5C,MAAM,eAAe,GAAG,mBAAmB,CAAC,OAAO,CAAC,CAAC;QACrD,IAAI,eAAe,EAAE,CAAC;YACrB,cAAc,EAAE,CAAC;YACjB,MAAM,SAAS,GAAG,0BAA0B,CAC3C,OAAO,EACP,eAAe,EACf,OAAO,CACP,CAAC;YACF,mBAAmB,CAAC,GAAG,EAAE,SAAS,CAAC,CAAC;YACpC,MAAM,YAAY,GAAG,sBAAsB,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC;YAC5D,MAAM,CAAC,IAAI,CAAC;gBACX,IAAI,EAAE,MAAM;gBACZ,MAAM,EAAE,eAAe;gBACvB,YAAY;gBACZ,aAAa,EAAE,OAAO;aACtB,CAAC,CAAC;YACH,cAAc,GAAG,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC;YAClC,kBAAkB,GAAG,IAAI,CAAC;YAC1B,OAAO;QACR,CAAC;QACD,KAAK,MAAM,KAAK,IAAI,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,UAAU,CAAC,EAAE,CAAC;YACpD,WAAW,CAAC,KAAK,CAAC,CAAC;QACpB,CAAC;IACF,CAAC,CAAC;IAEF,KAAK,MAAM,KAAK,IAAI,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,EAAE,CAAC;QACjD,WAAW,CAAC,KAAK,CAAC,CAAC;IACpB,CAAC;IAED,sBAAsB,CAAC,GAAG,CAAC,CAAC;IAC5B,cAAc,EAAE,CAAC;IACjB,MAAM,WAAW,GAAG,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;IACvC,OAAO,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,WAAW,CAAC,MAAM,CAAC,EAAE,CAAC;QACxC,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC;IACpC,CAAC;IACD,OAAO;QACN,WAAW;QACX,GAAG,EAAE,GAAG,CAAC,GAAG;QACZ,MAAM;QACN,kBAAkB;KAClB,CAAC;AACH,CAAC,CAAC;AAEF,MAAM,CAAC,MAAM,0BAA0B,GAAG,CACzC,OAAgB,EAChB,OAA+B,EACT,EAAE;IACxB,IAAI,CAAC,gBAAgB,CAAC,OAAO,CAAC,EAAE,CAAC;QAChC,MAAM,EAAE,IAAI,EAAE,GAAG,EAAE,GAAG,wBAAwB,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;QACjE,OAAO;YACN,WAAW,EAAE,IAAI;YACjB,GAAG;YACH,MAAM,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE;YAC5C,kBAAkB,EAAE,KAAK;SACzB,CAAC;IACH,CAAC;IACD,OAAO,gBAAgB,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;AAC3C,CAAC,CAAC","sourcesContent":["import { canonicalizeMathML } from \"./mathml-sanitization.js\";\nimport {\n\tcollectVisibleTextAndMap,\n\ttype NormalizedTextMap,\n\ttype TextProcessingOptions,\n\tisNodeHiddenForTTS,\n\tnormalizeTextForSpeech,\n\tshouldInsertWordBoundarySpace,\n} from \"./text-processing.js\";\n\nexport type MathAwareSpeechChunk =\n\t| {\n\t\t\ttype: \"text\";\n\t\t\ttext: string;\n\t\t\tsourceElement?: Element;\n\t  }\n\t| {\n\t\t\ttype: \"math\";\n\t\t\tmathml: string;\n\t\t\tfallbackText: string;\n\t\t\tsourceElement?: Element;\n\t  };\n\nexport interface MathAwareTextResult {\n\tvisibleText: string;\n\tmap: NormalizedTextMap;\n\tchunks: MathAwareSpeechChunk[];\n\tcontainsMathMarkup: boolean;\n}\n\ninterface TextAccumulator {\n\tchars: string[];\n\tmap: NormalizedTextMap;\n\tposition: number;\n\tinLeadingWhitespace: boolean;\n\tlastCharWasWhitespace: boolean;\n\tlastMapped: { node: Text; offset: number } | null;\n\toptions?: TextProcessingOptions;\n}\n\nconst createAccumulator = (\n\toptions?: TextProcessingOptions,\n): TextAccumulator => ({\n\tchars: [],\n\tmap: new Map(),\n\tposition: 0,\n\tinLeadingWhitespace: true,\n\tlastCharWasWhitespace: false,\n\tlastMapped: null,\n\toptions,\n});\n\nconst isMathJaxElement = (element: Element): boolean =>\n\telement.classList?.contains(\"MathJax\") ||\n\telement.tagName?.toLowerCase() === \"mjx-container\";\n\nconst isAssistiveMathElement = (element: Element): boolean =>\n\telement.tagName?.toLowerCase() === \"mjx-assistive-mml\";\n\nconst isAnnotationElement = (element: Element): boolean => {\n\tconst tagName = element.tagName?.toLowerCase();\n\treturn tagName === \"annotation\" || tagName === \"annotation-xml\";\n};\n\nconst getDataMathML = (element: Element): string | null => {\n\tconst value = element.getAttribute(\"data-mathml\");\n\treturn value && value.trim() ? value : null;\n};\n\nconst findAssistiveMathML = (element: Element): string | null => {\n\tconst assistive = element.querySelector(\"mjx-assistive-mml math\");\n\treturn assistive?.outerHTML || null;\n};\n\nconst findCanonicalMathML = (element: Element): string | null => {\n\tif (element.tagName.toLowerCase() === \"math\") {\n\t\treturn canonicalizeMathML(element.outerHTML);\n\t}\n\tconst dataMath = getDataMathML(element);\n\tif (dataMath) {\n\t\tconst canonical = canonicalizeMathML(dataMath);\n\t\tif (canonical) return canonical;\n\t}\n\tif (isMathJaxElement(element)) {\n\t\tconst assistive = findAssistiveMathML(element);\n\t\tif (assistive) {\n\t\t\tconst canonical = canonicalizeMathML(assistive);\n\t\t\tif (canonical) return canonical;\n\t\t}\n\t}\n\treturn null;\n};\n\nconst isMathContainer = (element: Element): boolean =>\n\tBoolean(findCanonicalMathML(element));\n\nconst TEXT_CHUNK_SOURCE_TAGS = new Set([\n\t\"P\",\n\t\"H1\",\n\t\"H2\",\n\t\"H3\",\n\t\"H4\",\n\t\"H5\",\n\t\"H6\",\n\t\"LI\",\n\t\"TD\",\n\t\"TH\",\n\t\"DD\",\n\t\"DT\",\n\t\"FIGCAPTION\",\n\t\"LABEL\",\n\t\"BUTTON\",\n]);\n\nconst TEXT_CHUNK_SOURCE_ROLES = new Set([\n\t\"heading\",\n\t\"listitem\",\n\t\"option\",\n\t\"radio\",\n\t\"row\",\n\t\"cell\",\n\t\"columnheader\",\n\t\"rowheader\",\n]);\n\nconst resolveTextChunkSourceElement = (\n\ttextNode: Text,\n\troot: Element,\n): Element => {\n\tlet current = textNode.parentElement;\n\tlet best: Element | null = null;\n\twhile (current && current !== root) {\n\t\tconst role = (current.getAttribute(\"role\") || \"\").toLowerCase();\n\t\tconst tagName = current.tagName.toUpperCase();\n\t\tif (\n\t\t\tTEXT_CHUNK_SOURCE_TAGS.has(tagName) ||\n\t\t\tTEXT_CHUNK_SOURCE_ROLES.has(role)\n\t\t) {\n\t\t\tbest = current;\n\t\t\tbreak;\n\t\t}\n\t\tcurrent = current.parentElement;\n\t}\n\treturn best || root;\n};\n\nconst hasMathCandidate = (element: Element): boolean =>\n\tBoolean(\n\t\ttypeof element.querySelector === \"function\" &&\n\t\t\telement.querySelector(\n\t\t\t\t\"math, [data-mathml], .MathJax, mjx-container, mjx-assistive-mml\",\n\t\t\t),\n\t) ||\n\telement.tagName?.toLowerCase() === \"math\" ||\n\telement.hasAttribute?.(\"data-mathml\") ||\n\tisMathJaxElement(element);\n\nconst appendCharacter = (\n\tacc: TextAccumulator,\n\tcharacter: string,\n\tmapping: { node: Text; offset: number } | null,\n): void => {\n\tif (\n\t\t/[.,;:!?]/.test(character) &&\n\t\tacc.chars.length > 0 &&\n\t\tacc.chars[acc.chars.length - 1] === \" \"\n\t) {\n\t\tacc.chars.pop();\n\t\tacc.position--;\n\t\tacc.map.delete(acc.position);\n\t}\n\tacc.chars.push(character);\n\tif (mapping) {\n\t\tacc.map.set(acc.position, mapping);\n\t\tacc.lastMapped = mapping;\n\t} else if (acc.lastMapped) {\n\t\tacc.map.set(acc.position, acc.lastMapped);\n\t}\n\tacc.position++;\n};\n\nconst shouldInsertBoundarySpace = (\n\tacc: TextAccumulator,\n\tnextCharacter: string,\n): boolean => {\n\tif (\n\t\tacc.inLeadingWhitespace ||\n\t\tacc.lastCharWasWhitespace ||\n\t\tacc.chars.length === 0\n\t) {\n\t\treturn false;\n\t}\n\tconst previous = acc.chars[acc.chars.length - 1];\n\t// Defer to the shared locale-aware decision (Intl.Segmenter, falling back to\n\t// alnum) so adjacent text nodes split into words consistently with the main\n\t// visible-text collector, instead of an ASCII-only heuristic.\n\treturn shouldInsertWordBoundarySpace(previous, nextCharacter, acc.options);\n};\n\nconst appendTextNode = (acc: TextAccumulator, textNode: Text): void => {\n\tconst raw = textNode.textContent || \"\";\n\tlet appendedNonWhitespaceInNode = false;\n\tfor (let i = 0; i < raw.length; i++) {\n\t\tconst character = raw[i];\n\t\tconst isWhitespace = /\\s/.test(character);\n\t\tif (acc.inLeadingWhitespace) {\n\t\t\tif (!isWhitespace) {\n\t\t\t\tacc.inLeadingWhitespace = false;\n\t\t\t\tappendCharacter(acc, character, { node: textNode, offset: i });\n\t\t\t\tappendedNonWhitespaceInNode = true;\n\t\t\t\tacc.lastCharWasWhitespace = false;\n\t\t\t}\n\t\t\tcontinue;\n\t\t}\n\t\tif (isWhitespace) {\n\t\t\tif (!acc.lastCharWasWhitespace) {\n\t\t\t\tappendCharacter(acc, \" \", { node: textNode, offset: i });\n\t\t\t}\n\t\t\tacc.lastCharWasWhitespace = true;\n\t\t\tcontinue;\n\t\t}\n\t\tif (\n\t\t\t!appendedNonWhitespaceInNode &&\n\t\t\tshouldInsertBoundarySpace(acc, character)\n\t\t) {\n\t\t\tappendCharacter(acc, \" \", { node: textNode, offset: i });\n\t\t}\n\t\tappendCharacter(acc, character, { node: textNode, offset: i });\n\t\tappendedNonWhitespaceInNode = true;\n\t\tacc.lastCharWasWhitespace = false;\n\t}\n};\n\nconst appendCollectedText = (\n\tacc: TextAccumulator,\n\tcollected: { text: string; map: NormalizedTextMap },\n): void => {\n\tfor (let i = 0; i < collected.text.length; i++) {\n\t\tconst character = collected.text[i];\n\t\tconst isWhitespace = /\\s/.test(character);\n\t\tconst mapping = collected.map.get(i) || null;\n\t\tconst syntheticNode = mapping?.node || acc.lastMapped?.node || null;\n\t\tconst syntheticOffset = mapping?.offset ?? acc.lastMapped?.offset ?? 0;\n\t\tconst normalizedMapping = mapping\n\t\t\t? { node: mapping.node, offset: mapping.offset }\n\t\t\t: syntheticNode\n\t\t\t\t? { node: syntheticNode, offset: syntheticOffset }\n\t\t\t\t: null;\n\t\tif (acc.inLeadingWhitespace) {\n\t\t\tif (isWhitespace) continue;\n\t\t\tacc.inLeadingWhitespace = false;\n\t\t}\n\t\tif (isWhitespace) {\n\t\t\tif (!acc.lastCharWasWhitespace) {\n\t\t\t\tappendCharacter(acc, \" \", normalizedMapping);\n\t\t\t}\n\t\t\tacc.lastCharWasWhitespace = true;\n\t\t\tcontinue;\n\t\t}\n\t\tif (i === 0 && shouldInsertBoundarySpace(acc, character)) {\n\t\t\tappendCharacter(acc, \" \", normalizedMapping);\n\t\t}\n\t\tappendCharacter(acc, character, normalizedMapping);\n\t\tacc.lastCharWasWhitespace = false;\n\t\tif (mapping) {\n\t\t\tacc.map.set(acc.position - 1, {\n\t\t\t\tnode: mapping.node,\n\t\t\t\toffset: mapping.offset,\n\t\t\t});\n\t\t\tacc.lastMapped = { node: mapping.node, offset: mapping.offset };\n\t\t} else if (syntheticNode) {\n\t\t\tacc.map.set(acc.position - 1, {\n\t\t\t\tnode: syntheticNode,\n\t\t\t\toffset: syntheticOffset,\n\t\t\t});\n\t\t}\n\t}\n};\n\nconst textFallbackFromMathML = (mathml: string): string => {\n\tif (typeof DOMParser === \"undefined\") return \"\";\n\tconst parsed = new DOMParser().parseFromString(mathml, \"application/xml\");\n\tif (parsed.getElementsByTagName(\"parsererror\").length > 0) return \"\";\n\tconst parts: string[] = [];\n\tconst visit = (node: Node): void => {\n\t\tif (node.nodeType === 3) {\n\t\t\tconst text = node.textContent || \"\";\n\t\t\tif (text.trim()) parts.push(text);\n\t\t\treturn;\n\t\t}\n\t\tif (node.nodeType !== 1) return;\n\t\tconst element = node as Element;\n\t\tif (isAnnotationElement(element)) return;\n\t\tfor (const child of Array.from(element.childNodes)) {\n\t\t\tvisit(child);\n\t\t}\n\t};\n\tvisit(parsed.documentElement);\n\treturn normalizeTextForSpeech(parts.join(\" \"));\n};\n\nconst collectVisibleMathFallback = (\n\telement: Element,\n\tcanonicalMathML: string,\n\toptions?: TextProcessingOptions,\n): { text: string; map: NormalizedTextMap } => {\n\tconst mathAcc = createAccumulator(options);\n\tconst visit = (node: Node): void => {\n\t\tif (isNodeHiddenForTTS(node, element)) return;\n\t\tif (node.nodeType === 3) {\n\t\t\tappendTextNode(mathAcc, node as Text);\n\t\t\treturn;\n\t\t}\n\t\tif (node.nodeType !== 1) return;\n\t\tconst childElement = node as Element;\n\t\tif (\n\t\t\tisAssistiveMathElement(childElement) ||\n\t\t\tisAnnotationElement(childElement)\n\t\t) {\n\t\t\treturn;\n\t\t}\n\t\tfor (const child of Array.from(childElement.childNodes)) {\n\t\t\tvisit(child);\n\t\t}\n\t};\n\tvisit(element);\n\ttrimTrailingWhitespace(mathAcc);\n\tconst text = normalizeTextForSpeech(mathAcc.chars.join(\"\"));\n\tif (text) {\n\t\treturn { text, map: mathAcc.map };\n\t}\n\tconst fallbackText = textFallbackFromMathML(canonicalMathML);\n\tif (!fallbackText) {\n\t\treturn collectVisibleTextAndMap(element, options);\n\t}\n\treturn {\n\t\ttext: fallbackText,\n\t\tmap: new Map(),\n\t};\n};\n\nconst trimTrailingWhitespace = (acc: TextAccumulator): void => {\n\twhile (acc.chars.length > 0 && /\\s/.test(acc.chars[acc.chars.length - 1])) {\n\t\tacc.chars.pop();\n\t\tacc.position--;\n\t\tacc.map.delete(acc.position);\n\t}\n};\n\nconst collectMathAware = (\n\troot: Element,\n\toptions?: TextProcessingOptions,\n): MathAwareTextResult => {\n\tconst acc = createAccumulator(options);\n\tconst chunks: MathAwareSpeechChunk[] = [];\n\tlet textChunkStart = 0;\n\tlet textChunkSourceElement: Element | undefined;\n\tlet containsMathMarkup = false;\n\n\tconst flushTextChunk = () => {\n\t\tconst text = normalizeTextForSpeech(\n\t\t\tacc.chars.slice(textChunkStart, acc.chars.length).join(\"\"),\n\t\t);\n\t\tif (text) {\n\t\t\tchunks.push({ type: \"text\", text, sourceElement: textChunkSourceElement });\n\t\t}\n\t\ttextChunkStart = acc.chars.length;\n\t\ttextChunkSourceElement = undefined;\n\t};\n\n\tconst processNode = (node: Node): void => {\n\t\tif (isNodeHiddenForTTS(node, root)) return;\n\t\tif (node.nodeType === Node.TEXT_NODE) {\n\t\t\tconst sourceElement = resolveTextChunkSourceElement(node as Text, root);\n\t\t\tif (\n\t\t\t\ttextChunkSourceElement &&\n\t\t\t\tsourceElement !== textChunkSourceElement\n\t\t\t) {\n\t\t\t\tflushTextChunk();\n\t\t\t}\n\t\t\ttextChunkSourceElement = sourceElement;\n\t\t\tappendTextNode(acc, node as Text);\n\t\t\treturn;\n\t\t}\n\t\tif (node.nodeType !== Node.ELEMENT_NODE) return;\n\t\tconst element = node as Element;\n\t\tif (isAssistiveMathElement(element)) return;\n\t\tconst canonicalMathML = findCanonicalMathML(element);\n\t\tif (canonicalMathML) {\n\t\t\tflushTextChunk();\n\t\t\tconst collected = collectVisibleMathFallback(\n\t\t\t\telement,\n\t\t\t\tcanonicalMathML,\n\t\t\t\toptions,\n\t\t\t);\n\t\t\tappendCollectedText(acc, collected);\n\t\t\tconst fallbackText = normalizeTextForSpeech(collected.text);\n\t\t\tchunks.push({\n\t\t\t\ttype: \"math\",\n\t\t\t\tmathml: canonicalMathML,\n\t\t\t\tfallbackText,\n\t\t\t\tsourceElement: element,\n\t\t\t});\n\t\t\ttextChunkStart = acc.chars.length;\n\t\t\tcontainsMathMarkup = true;\n\t\t\treturn;\n\t\t}\n\t\tfor (const child of Array.from(element.childNodes)) {\n\t\t\tprocessNode(child);\n\t\t}\n\t};\n\n\tfor (const child of Array.from(root.childNodes)) {\n\t\tprocessNode(child);\n\t}\n\n\ttrimTrailingWhitespace(acc);\n\tflushTextChunk();\n\tconst visibleText = acc.chars.join(\"\");\n\twhile (acc.map.has(visibleText.length)) {\n\t\tacc.map.delete(visibleText.length);\n\t}\n\treturn {\n\t\tvisibleText,\n\t\tmap: acc.map,\n\t\tchunks,\n\t\tcontainsMathMarkup,\n\t};\n};\n\nexport const collectMathAwareTextAndMap = (\n\telement: Element,\n\toptions?: TextProcessingOptions,\n): MathAwareTextResult => {\n\tif (!hasMathCandidate(element)) {\n\t\tconst { text, map } = collectVisibleTextAndMap(element, options);\n\t\treturn {\n\t\t\tvisibleText: text,\n\t\t\tmap,\n\t\t\tchunks: text ? [{ type: \"text\", text }] : [],\n\t\t\tcontainsMathMarkup: false,\n\t\t};\n\t}\n\treturn collectMathAware(element, options);\n};\n"]}