eyecite-ts 0.2.0 → 0.4.0

package/dist/index.cjs.map

@@ -1 +1 @@
-{"version":3,"file":"index.cjs","names":[],"sources":["../src/types/guards.ts","../src/clean/cleaners.ts","../src/clean/cleanText.ts","../src/patterns/casePatterns.ts","../src/patterns/statutePatterns.ts","../src/patterns/journalPatterns.ts","../src/patterns/neutralPatterns.ts","../src/patterns/shortForm.ts","../src/tokenize/tokenizer.ts","../src/extract/extractCase.ts","../src/extract/extractStatute.ts","../src/extract/extractJournal.ts","../src/extract/extractNeutral.ts","../src/extract/extractPublicLaw.ts","../src/extract/extractFederalRegister.ts","../src/extract/extractShortForms.ts","../src/resolve/scopeBoundary.ts","../src/resolve/levenshtein.ts","../src/resolve/DocumentResolver.ts","../src/resolve/index.ts","../src/extract/extractCitations.ts"],"sourcesContent":["import type { Citation, CitationType, CitationOfType, FullCitation, ShortFormCitation, FullCaseCitation } from \"./citation\"\n\n/**\n * Type guard: narrows Citation to a full citation (case, statute, journal, neutral, publicLaw, federalRegister).\n */\nexport function isFullCitation(citation: Citation): citation is FullCitation {\n  return citation.type === 'case'\n    || citation.type === 'statute'\n    || citation.type === 'journal'\n    || citation.type === 'neutral'\n    || citation.type === 'publicLaw'\n    || citation.type === 'federalRegister'\n}\n\n/**\n * Type guard: narrows Citation to a short-form citation (id, supra, shortFormCase).\n */\nexport function isShortFormCitation(citation: Citation): citation is ShortFormCitation {\n  return citation.type === 'id'\n    || citation.type === 'supra'\n    || citation.type === 'shortFormCase'\n}\n\n/**\n * Type guard: narrows Citation to a full case citation.\n */\nexport function isCaseCitation(citation: Citation): citation is FullCaseCitation {\n  return citation.type === 'case'\n}\n\n/**\n * Generic type guard that narrows a Citation to a specific type.\n * Useful when the target type is dynamic or generic.\n */\nexport function isCitationType<T extends CitationType>(\n  citation: Citation,\n  type: T\n): citation is CitationOfType<T> {\n  return citation.type === type\n}\n\n/**\n * Exhaustiveness helper for switch statements on discriminated unions.\n *\n * Place in the `default` branch to get a compile-time error if a new\n * variant is added but not handled.\n *\n * @example\n * ```typescript\n * switch (citation.type) {\n *   case 'case': ...\n *   case 'statute': ...\n *   // If you forget a variant, TypeScript errors here:\n *   default: assertUnreachable(citation.type)\n * }\n * ```\n */\nexport function assertUnreachable(x: never): never {\n  throw new Error(`Unexpected value: ${x}`)\n}\n","/**\n * Built-in text cleaner functions for preprocessing legal documents.\n *\n * Each cleaner is a simple transformation: (text: string) => string\n * Cleaners can be composed via the cleanText() pipeline.\n */\n\n/**\n * Remove all HTML tags from text.\n *\n * @example\n * stripHtmlTags(\"Smith v. <b>Doe</b>, 500 F.2d 123\")\n * // => \"Smith v. Doe, 500 F.2d 123\"\n */\nexport function stripHtmlTags(text: string): string {\n\treturn text.replace(/<[^>]+>/g, \"\")\n}\n\n/**\n * Normalize whitespace: convert tabs/newlines to spaces, collapse multiple spaces.\n *\n * @example\n * normalizeWhitespace(\"Smith  v.  Doe,   500 F.2d 123\")\n * // => \"Smith v. Doe, 500 F.2d 123\"\n */\nexport function normalizeWhitespace(text: string): string {\n\treturn text.replace(/[\\t\\n\\r]+/g, \" \").replace(/ {2,}/g, \" \")\n}\n\n/**\n * Apply Unicode NFKC normalization (ligatures → separate chars).\n *\n * @example\n * normalizeUnicode(\"Smith v. Doe, 500 F.2d 123\") // with ligature \"fi\"\n * // => \"Smith v. Doe, 500 F.2d 123\" // normalized\n */\nexport function normalizeUnicode(text: string): string {\n\treturn text.normalize(\"NFKC\")\n}\n\n/**\n * Replace curly quotes and apostrophes with straight quotes.\n *\n * @example\n * fixSmartQuotes(\"\"Smith\" v. 'Doe', 500 F.2d 123\")\n * // => \"\\\"Smith\\\" v. 'Doe', 500 F.2d 123\"\n */\nexport function fixSmartQuotes(text: string): string {\n\treturn text\n\t\t.replace(/[\\u201C\\u201D]/g, '\"') // curly double quotes\n\t\t.replace(/[\\u2018\\u2019]/g, \"'\") // curly single quotes/apostrophes\n}\n\n/**\n * Remove underscore OCR artifacts (common in scanned documents).\n *\n * @example\n * removeOcrArtifacts(\"Smith v. Doe, 500 F._2d 123\")\n * // => \"Smith v. Doe, 500 F.2d 123\"\n */\nexport function removeOcrArtifacts(text: string): string {\n\treturn text.replace(/_/g, \"\")\n}\n","import type { TransformationMap } from \"../types/span\"\nimport type { Warning } from \"../types/citation\"\nimport {\n\tfixSmartQuotes,\n\tnormalizeUnicode,\n\tnormalizeWhitespace,\n\tstripHtmlTags,\n} from \"./cleaners\"\n\n/**\n * Result of text cleaning operation.\n */\nexport interface CleanTextResult {\n\t/** Cleaned text after all transformations */\n\tcleaned: string\n\n\t/** Position mappings between cleaned and original text */\n\ttransformationMap: TransformationMap\n\n\t/** Warnings generated during cleaning (currently unused) */\n\twarnings: Warning[]\n}\n\n/**\n * Clean text using a pipeline of transformation functions.\n *\n * Applies cleaners sequentially while maintaining accurate position mappings\n * between the original and cleaned text. This enables citation extraction from\n * cleaned text while reporting positions in the original text.\n *\n * @param original - Original input text\n * @param cleaners - Array of cleaner functions to apply (default: stripHtmlTags, normalizeWhitespace, normalizeUnicode, fixSmartQuotes)\n * @returns Cleaned text with position mappings and warnings\n *\n * @example\n * const result = cleanText(\"Smith v. <b>Doe</b>, 500 F.2d 123\")\n * // result.cleaned: \"Smith v. Doe, 500 F.2d 123\"\n * // result.transformationMap tracks position shifts from HTML removal\n */\nexport function cleanText(\n\toriginal: string,\n\tcleaners: Array<(text: string) => string> = [\n\t\tstripHtmlTags,\n\t\tnormalizeWhitespace,\n\t\tnormalizeUnicode,\n\t\tfixSmartQuotes,\n\t],\n): CleanTextResult {\n\t// Initialize 1:1 position mapping\n\tlet currentText = original\n\tlet cleanToOriginal = new Map<number, number>()\n\tlet originalToClean = new Map<number, number>()\n\n\t// Identity mapping: cleanToOriginal[i] = i, originalToClean[i] = i\n\tfor (let i = 0; i <= original.length; i++) {\n\t\tcleanToOriginal.set(i, i)\n\t\toriginalToClean.set(i, i)\n\t}\n\n\t// Apply each cleaner sequentially, rebuilding position maps\n\tfor (const cleaner of cleaners) {\n\t\tconst beforeText = currentText\n\t\tconst afterText = cleaner(currentText)\n\n\t\tif (beforeText !== afterText) {\n\t\t\t// Text changed - rebuild position maps\n\t\t\tconst { newCleanToOriginal, newOriginalToClean } = rebuildPositionMaps(\n\t\t\t\tbeforeText,\n\t\t\t\tafterText,\n\t\t\t\tcleanToOriginal,\n\t\t\t\toriginalToClean,\n\t\t\t)\n\n\t\t\tcleanToOriginal = newCleanToOriginal\n\t\t\toriginalToClean = newOriginalToClean\n\t\t\tcurrentText = afterText\n\t\t}\n\t}\n\n\tconst transformationMap: TransformationMap = {\n\t\tcleanToOriginal,\n\t\toriginalToClean,\n\t}\n\n\treturn {\n\t\tcleaned: currentText,\n\t\ttransformationMap,\n\t\twarnings: [],\n\t}\n}\n\n/**\n * Rebuild position maps after a text transformation.\n *\n * Uses a simplified algorithm that scans through both strings, matching\n * characters where possible and tracking the offset accumulation.\n *\n * @param beforeText - Text before transformation\n * @param afterText - Text after transformation\n * @param oldCleanToOriginal - Previous clean-to-original mapping\n * @param oldOriginalToClean - Previous original-to-clean mapping\n * @returns New position maps\n */\nfunction rebuildPositionMaps(\n\tbeforeText: string,\n\tafterText: string,\n\toldCleanToOriginal: Map<number, number>,\n\toldOriginalToClean: Map<number, number>,\n): {\n\tnewCleanToOriginal: Map<number, number>\n\tnewOriginalToClean: Map<number, number>\n} {\n\tconst newCleanToOriginal = new Map<number, number>()\n\tconst newOriginalToClean = new Map<number, number>()\n\n\tlet beforeIdx = 0\n\tlet afterIdx = 0\n\n\t// Scan through both strings, matching characters where possible\n\twhile (beforeIdx <= beforeText.length || afterIdx <= afterText.length) {\n\t\t// Both at end\n\t\tif (beforeIdx >= beforeText.length && afterIdx >= afterText.length) {\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewCleanToOriginal.set(afterIdx, originalPos)\n\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\tbreak\n\t\t}\n\n\t\t// Before text exhausted (expansion case)\n\t\tif (beforeIdx >= beforeText.length) {\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewCleanToOriginal.set(afterIdx, originalPos)\n\t\t\tafterIdx++\n\t\t\tcontinue\n\t\t}\n\n\t\t// After text exhausted (removal case)\n\t\tif (afterIdx >= afterText.length) {\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\tbeforeIdx++\n\t\t\tcontinue\n\t\t}\n\n\t\t// Characters match - carry forward the mapping\n\t\tif (beforeText[beforeIdx] === afterText[afterIdx]) {\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewCleanToOriginal.set(afterIdx, originalPos)\n\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\tbeforeIdx++\n\t\t\tafterIdx++\n\t\t} else {\n\t\t\t// Characters differ - need to determine if this is insertion/deletion/replacement\n\t\t\t// Look ahead to find next match\n\t\t\tlet foundMatch = false\n\t\t\tconst maxLookAhead = 20 // Limit lookahead to avoid performance issues\n\n\t\t\t// Check if something was deleted from before text\n\t\t\tfor (let lookAhead = 1; lookAhead <= maxLookAhead; lookAhead++) {\n\t\t\t\tif (beforeIdx + lookAhead >= beforeText.length) break\n\n\t\t\t\tif (beforeText[beforeIdx + lookAhead] === afterText[afterIdx]) {\n\t\t\t\t\t// Found a match - characters were deleted from before text\n\t\t\t\t\tfor (let i = 0; i < lookAhead; i++) {\n\t\t\t\t\t\tconst originalPos =\n\t\t\t\t\t\t\toldCleanToOriginal.get(beforeIdx + i) ?? beforeIdx + i\n\t\t\t\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\t\t\t}\n\t\t\t\t\tbeforeIdx += lookAhead\n\t\t\t\t\tfoundMatch = true\n\t\t\t\t\tbreak\n\t\t\t\t}\n\t\t\t}\n\n\t\t\tif (foundMatch) continue\n\n\t\t\t// Check if something was inserted into after text\n\t\t\tfor (let lookAhead = 1; lookAhead <= maxLookAhead; lookAhead++) {\n\t\t\t\tif (afterIdx + lookAhead >= afterText.length) break\n\n\t\t\t\tif (beforeText[beforeIdx] === afterText[afterIdx + lookAhead]) {\n\t\t\t\t\t// Found a match - characters were inserted into after text\n\t\t\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\t\t\tfor (let i = 0; i < lookAhead; i++) {\n\t\t\t\t\t\tnewCleanToOriginal.set(afterIdx + i, originalPos)\n\t\t\t\t\t}\n\t\t\t\t\tafterIdx += lookAhead\n\t\t\t\t\tfoundMatch = true\n\t\t\t\t\tbreak\n\t\t\t\t}\n\t\t\t}\n\n\t\t\tif (foundMatch) continue\n\n\t\t\t// No match found within lookahead - treat as replacement\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewCleanToOriginal.set(afterIdx, originalPos)\n\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\tbeforeIdx++\n\t\t\tafterIdx++\n\t\t}\n\t}\n\n\treturn { newCleanToOriginal, newOriginalToClean }\n}\n","/**\n * Case Citation Regex Patterns\n *\n * These patterns are designed for tokenization (broad matching) not extraction.\n * They identify potential case citations in text for the tokenizer (Plan 3).\n * Metadata parsing and validation against reporters-db happens in Phase 2 Plan 5 (extraction layer).\n *\n * Pattern Design Principles (from RESEARCH.md):\n * - Use \\b word boundaries to avoid matching \"F.\" in \"F.B.I.\"\n * - Avoid nested quantifiers: (a+)+ causes ReDoS\n * - Keep patterns simple: tokenization only needs to find candidates\n * - Use global flag /g for matchAll()\n */\n\nimport type { FullCitationType } from '@/types/citation'\n\nexport interface Pattern {\n  id: string\n  regex: RegExp\n  description: string\n  type: FullCitationType\n}\n\nexport const casePatterns: Pattern[] = [\n  {\n    id: 'federal-reporter',\n    regex: /\\b(\\d+)\\s+(F\\.|F\\.2d|F\\.3d|F\\.\\s?Supp\\.|F\\.\\s?Supp\\.\\s?2d|F\\.\\s?Supp\\.\\s?3d)\\s+(\\d+)\\b/g,\n    description: 'Federal Reporter (F., F.2d, F.3d, F.Supp., etc.)',\n    type: 'case',\n  },\n  {\n    id: 'supreme-court',\n    regex: /\\b(\\d+)\\s+(U\\.S\\.|S\\.\\s?Ct\\.|L\\.\\s?Ed\\.(?:\\s?2d)?)\\s+(\\d+)\\b/g,\n    description: 'U.S. Supreme Court reporters',\n    type: 'case',\n  },\n  {\n    id: 'state-reporter',\n    regex: /\\b(\\d+)\\s+([A-Z][A-Za-z\\.]+(?:\\s?2d|\\s?3d)?)\\s+(\\d+)\\b/g,\n    description: 'State reporters (broad pattern, validated against reporters-db in Phase 3)',\n    type: 'case',\n  },\n]\n","/**\n * Statute Citation Regex Patterns\n *\n * Patterns for U.S. Code and state code citations.\n * These are intentionally broad for tokenization - validation against\n * actual code databases happens in Phase 2 Plan 5 (extraction layer).\n *\n * Pattern Design:\n * - Simple structure to avoid ReDoS\n * - Matches both \"§\" and \"Section\" formats\n * - State codes use broad pattern (validated later)\n */\n\nimport type { Pattern } from './casePatterns'\n\nexport const statutePatterns: Pattern[] = [\n  {\n    id: 'usc',\n    regex: /\\b(\\d+)\\s+U\\.S\\.C\\.?\\s+§+\\s*(\\d+)\\b/g,\n    description: 'U.S. Code citations (e.g., \"42 U.S.C. § 1983\")',\n    type: 'statute',\n  },\n  {\n    id: 'state-code',\n    regex: /\\b([A-Z][a-z]+\\.?\\s+[A-Za-z\\.]+\\s+Code)\\s+§\\s*(\\d+)\\b/g,\n    description: 'State code citations (broad pattern, e.g., \"Cal. Penal Code § 187\")',\n    type: 'statute',\n  },\n]\n","/**\n * Journal Citation Regex Patterns\n *\n * Patterns for law review and journal citations.\n * These are intentionally broad for tokenization - validation against\n * journals-db happens in Phase 3 (extraction layer).\n *\n * Pattern Design:\n * - Matches volume-journal-page format\n * - Broad journal name matching (validated later)\n * - Simple structure to avoid ReDoS\n */\n\nimport type { Pattern } from './casePatterns'\n\nexport const journalPatterns: Pattern[] = [\n  {\n    id: 'law-review',\n    regex: /\\b(\\d+)\\s+([A-Z][A-Za-z\\.\\s]+)\\s+(\\d+)\\b/g,\n    description: 'Law review citations (e.g., \"120 Harv. L. Rev. 500\"), validated against journals-db in Phase 3',\n    type: 'journal',\n  },\n]\n","/**\n * Neutral and Online Citation Regex Patterns\n *\n * Patterns for WestLaw, LexisNexis, public laws, and Federal Register citations.\n * These have predictable formats and don't require external validation.\n *\n * Pattern Design:\n * - Matches year-database-number format for online citations\n * - Matches Pub. L. No. format for public laws\n * - Matches volume-Fed. Reg.-page for Federal Register\n * - Simple structure to avoid ReDoS\n */\n\nimport type { Pattern } from './casePatterns'\n\nexport const neutralPatterns: Pattern[] = [\n  {\n    id: 'westlaw',\n    regex: /\\b(\\d{4})\\s+WL\\s+(\\d+)\\b/g,\n    description: 'WestLaw citations (e.g., \"2021 WL 123456\")',\n    type: 'neutral',\n  },\n  {\n    id: 'lexis',\n    regex: /\\b(\\d{4})\\s+U\\.S\\.\\s+LEXIS\\s+(\\d+)\\b/g,\n    description: 'LexisNexis citations (e.g., \"2021 U.S. LEXIS 5000\")',\n    type: 'neutral',\n  },\n  {\n    id: 'public-law',\n    regex: /\\bPub\\.\\s?L\\.\\s?No\\.\\s?(\\d+-\\d+)\\b/g,\n    description: 'Public Law citations (e.g., \"Pub. L. No. 117-58\")',\n    type: 'publicLaw',\n  },\n  {\n    id: 'federal-register',\n    regex: /\\b(\\d+)\\s+Fed\\.\\s?Reg\\.\\s+(\\d+)\\b/g,\n    description: 'Federal Register citations (e.g., \"86 Fed. Reg. 12345\")',\n    type: 'federalRegister',\n  },\n]\n","/**\n * Short-form Citation Regex Patterns\n *\n * Patterns for Id., Ibid., supra, and short-form case citations.\n * These refer to earlier citations in the document.\n *\n * Pattern Design:\n * - Simple structure to avoid ReDoS (no nested quantifiers)\n * - Broad matching for tokenization; validation happens in extraction layer\n * - Word boundaries to prevent false positives (e.g., \"Idaho\" vs \"Id.\")\n */\n\nimport type { Pattern } from './casePatterns'\n\n/** Id. with optional pincite: \"Id.\" or \"Id. at 253\" */\nexport const ID_PATTERN: RegExp = /\\b[Ii]d\\.(?:\\s+at\\s+(\\d+))?/g\n\n/** Ibid. with optional pincite (less common variant) */\nexport const IBID_PATTERN: RegExp = /\\b[Ii]bid\\.(?:\\s+at\\s+(\\d+))?/g\n\n/**\n * Supra with party name and optional pincite.\n * Pattern: word(s), supra [, at page]\n * Captures: (1) party name, (2) pincite\n * Note: Matches party names including \"v.\" (e.g., \"Smith v. Jones, supra\")\n */\nexport const SUPRA_PATTERN: RegExp = /\\b([A-Z][a-zA-Z]+(?:(?:\\s+v\\.?\\s+|\\s+)[A-Z][a-zA-Z]+)*),?\\s+supra(?:,?\\s+at\\s+(\\d+))?/g\n\n/**\n * Short-form case: volume reporter at page\n * Pattern: number space abbreviation space \"at\" space number\n * Simplified detection; full parsing in extraction layer\n */\nexport const SHORT_FORM_CASE_PATTERN: RegExp = /\\b(\\d+)\\s+([A-Z][A-Za-z.\\s]+?(?:\\d[a-z])?)\\s+at\\s+(\\d+)\\b/g\n\n/** All short-form patterns for tokenization */\nexport const SHORT_FORM_PATTERNS: readonly RegExp[] = [\n  ID_PATTERN,\n  IBID_PATTERN,\n  SUPRA_PATTERN,\n  SHORT_FORM_CASE_PATTERN,\n] as const\n\n/** Pattern objects for consistency with other pattern modules */\nexport const shortFormPatterns: Pattern[] = [\n  {\n    id: 'id',\n    regex: ID_PATTERN,\n    description: 'Id. citations (e.g., \"Id.\" or \"Id. at 253\")',\n    type: 'case', // Will be typed as 'id' in extraction layer\n  },\n  {\n    id: 'ibid',\n    regex: IBID_PATTERN,\n    description: 'Ibid. citations (e.g., \"Ibid.\" or \"Ibid. at 125\")',\n    type: 'case', // Will be typed as 'id' in extraction layer\n  },\n  {\n    id: 'supra',\n    regex: SUPRA_PATTERN,\n    description: 'Supra citations (e.g., \"Smith, supra\" or \"Smith, supra, at 460\")',\n    type: 'case', // Will be typed as 'supra' in extraction layer\n  },\n  {\n    id: 'shortFormCase',\n    regex: SHORT_FORM_CASE_PATTERN,\n    description: 'Short-form case citations (e.g., \"500 F.2d at 125\")',\n    type: 'case', // Will be typed as 'shortFormCase' in extraction layer\n  },\n]\n","/**\n * Tokenization Layer for Citation Extraction\n *\n * Applies regex patterns to cleaned text to produce citation candidate tokens.\n * This is the second stage of the parsing pipeline:\n *   1. Clean text (remove HTML, normalize Unicode)\n *   2. Tokenize (apply patterns to find candidates) ← THIS MODULE\n *   3. Extract (parse metadata, validate against reporters-db)\n *\n * Tokenization is intentionally broad - it finds potential citations without\n * validating them. The extraction layer (Plan 5) validates tokens against\n * reporters-db and parses metadata.\n *\n * @module tokenize\n */\n\nimport type { Span } from '@/types/span'\nimport type { Pattern } from '@/patterns'\nimport {\n  casePatterns,\n  statutePatterns,\n  journalPatterns,\n  neutralPatterns,\n} from '@/patterns'\nimport { shortFormPatterns } from '@/patterns/shortForm'\n\n/**\n * A token representing a potential citation found in cleaned text.\n *\n * Tokens are produced by applying regex patterns to cleaned text.\n * They include matched text, position in cleaned text, and pattern metadata\n * for use in the extraction layer.\n */\nexport interface Token {\n  /** Matched text from input */\n  text: string\n\n  /** Position in cleaned text (cleanStart/cleanEnd only, no original positions yet) */\n  span: Pick<Span, 'cleanStart' | 'cleanEnd'>\n\n  /** Pattern type that matched this token */\n  type: Pattern['type']\n\n  /** Pattern ID that matched this token */\n  patternId: string\n}\n\n/**\n * Tokenizes cleaned text by applying regex patterns to find citation candidates.\n *\n * For each pattern in the patterns array:\n *   1. Apply pattern.regex.matchAll(cleanedText)\n *   2. Create Token for each match with position, text, and pattern metadata\n *   3. Collect all tokens from all patterns\n *   4. Sort by cleanStart position (ascending)\n *\n * Timeout protection: If a pattern throws (e.g., ReDoS), skip it and continue\n * with remaining patterns. Logs warning to console.\n *\n * Note: This function is synchronous because regex matching is inherently\n * synchronous. This enables both sync (extractCitations) and async\n * (extractCitationsAsync) APIs in Plan 6.\n *\n * @param cleanedText - Text that has been cleaned by cleanText() from Plan 1\n * @param patterns - Regex patterns to apply (defaults to all patterns from Plan 2)\n * @returns Array of tokens sorted by position (cleanStart ascending)\n *\n * @example\n * ```typescript\n * import { tokenize } from '@/tokenize'\n * import { cleanText } from '@/clean'\n *\n * const original = \"See Smith v. Doe, 500 F.2d 123 (9th Cir. 2020)\"\n * const { cleanedText } = cleanText(original)\n * const tokens = tokenize(cleanedText)\n * // tokens[0] = {\n * //   text: \"500 F.2d 123\",\n * //   span: { cleanStart: 18, cleanEnd: 30 },\n * //   type: \"case\",\n * //   patternId: \"federal-reporter\"\n * // }\n * ```\n */\nexport function tokenize(\n  cleanedText: string,\n  patterns: Pattern[] = [\n    ...casePatterns,\n    ...statutePatterns,\n    ...journalPatterns,\n    ...neutralPatterns,\n    ...shortFormPatterns,\n  ]\n): Token[] {\n  const tokens: Token[] = []\n\n  for (const pattern of patterns) {\n    try {\n      // Apply pattern to cleaned text\n      const matches = cleanedText.matchAll(pattern.regex)\n\n      for (const match of matches) {\n        // Create token from match\n        tokens.push({\n          text: match[0],\n          span: {\n            cleanStart: match.index!,\n            cleanEnd: match.index! + match[0].length,\n          },\n          type: pattern.type,\n          patternId: pattern.id,\n        })\n      }\n    } catch (error) {\n      // Timeout protection: If pattern throws (ReDoS, etc.), skip it\n      console.warn(\n        `Pattern ${pattern.id} threw error, skipping:`,\n        error instanceof Error ? error.message : String(error)\n      )\n      continue\n    }\n  }\n\n  // Sort tokens by position (cleanStart ascending)\n  tokens.sort((a, b) => a.span.cleanStart - b.span.cleanStart)\n\n  return tokens\n}\n","/**\n * Case Citation Extraction\n *\n * Parses tokenized case citations to extract volume, reporter, page, and\n * optional metadata (pincite, court, year). This is the third stage of\n * the parsing pipeline:\n *   1. Clean text (remove HTML, normalize Unicode)\n *   2. Tokenize (apply patterns to find candidates)\n *   3. Extract (parse metadata, validate) ← THIS MODULE\n *\n * Extraction parses structured data from token text. Validation against\n * reporters-db happens in Phase 3 (resolution layer).\n *\n * @module extract/extractCase\n */\n\nimport type { Token } from '@/tokenize'\nimport type { FullCaseCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts case citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Volume: Leading digits (e.g., \"500\" from \"500 F.2d 123\")\n * - Reporter: Alphabetic abbreviation (e.g., \"F.2d\")\n * - Page: Trailing digits after reporter (e.g., \"123\")\n * - Pincite: Optional page reference after comma (e.g., \", 125\")\n * - Court: Optional court abbreviation in parentheses (e.g., \"(9th Cir.)\")\n * - Year: Optional year in parentheses (e.g., \"(2020)\")\n *\n * Confidence scoring:\n * - Base: 0.5\n * - Common reporter pattern (F., U.S., etc.): +0.3\n * - Valid year (not future): +0.2\n * - Capped at 1.0\n *\n * Position translation:\n * - Uses TransformationMap to convert clean positions → original positions\n * - cleanStart/cleanEnd from token span\n * - originalStart/originalEnd via transformationMap.cleanToOriginal\n *\n * Note: This function does NOT validate against reporters-db. That happens\n * in Phase 3 (resolution layer). Phase 2 extraction only parses structure.\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns FullCaseCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"500 F.2d 123, 125\",\n *   span: { cleanStart: 10, cleanEnd: 27 },\n *   type: \"case\",\n *   patternId: \"federal-reporter\"\n * }\n * const citation = extractCase(token, transformationMap)\n * // citation = {\n * //   type: \"case\",\n * //   text: \"500 F.2d 123, 125\",\n * //   volume: 500,\n * //   reporter: \"F.2d\",\n * //   page: 123,\n * //   pincite: 125,\n * //   span: { cleanStart: 10, cleanEnd: 27, originalStart: 10, originalEnd: 27 },\n * //   confidence: 0.8,\n * //   ...\n * // }\n * ```\n */\nexport function extractCase(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): FullCaseCitation {\n\tconst { text, span } = token\n\n\t// Parse volume-reporter-page using regex\n\t// Pattern: volume (digits) + reporter (letters/periods/spaces/numbers) + page (digits)\n\t// Use greedy matching for reporter to capture full abbreviation including spaces\n\tconst volumeReporterPageRegex = /^(\\d+)\\s+([A-Za-z0-9.\\s]+)\\s+(\\d+)/\n\tconst match = volumeReporterPageRegex.exec(text)\n\n\tif (!match) {\n\t\t// Fallback if pattern doesn't match (shouldn't happen if tokenizer is correct)\n\t\tthrow new Error(`Failed to parse case citation: ${text}`)\n\t}\n\n\tconst volume = Number.parseInt(match[1], 10)\n\tconst reporter = match[2].trim()\n\tconst page = Number.parseInt(match[3], 10)\n\n\t// Extract optional pincite (page reference after comma)\n\t// Pattern: \", digits\" (e.g., \", 125\")\n\tconst pinciteRegex = /,\\s*(\\d+)/\n\tconst pinciteMatch = pinciteRegex.exec(text)\n\tconst pincite = pinciteMatch ? Number.parseInt(pinciteMatch[1], 10) : undefined\n\n\t// Extract optional year in parentheses (extract first for better matching)\n\t// Pattern: 4-digit year anywhere in parentheses\n\tconst yearRegex = /\\((?:[^)]*\\s)?(\\d{4})\\)/\n\tconst yearMatch = yearRegex.exec(text)\n\tconst year = yearMatch ? Number.parseInt(yearMatch[1], 10) : undefined\n\n\t// Extract optional court abbreviation in parentheses\n\t// Pattern: \"(text)\" where text contains letters (captures full parenthetical)\n\tconst courtRegex = /\\(([^)]*[A-Za-z][^)]*)\\)/\n\tconst courtMatch = courtRegex.exec(text)\n\tconst court = courtMatch ? courtMatch[1].trim() : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Calculate confidence score\n\tlet confidence = 0.5 // Base confidence\n\n\t// Common reporter patterns (F., U.S., S. Ct., etc.)\n\tconst commonReporters = [\n\t\t'F.',\n\t\t'F.2d',\n\t\t'F.3d',\n\t\t'F.4th',\n\t\t'U.S.',\n\t\t'S. Ct.',\n\t\t'L. Ed.',\n\t\t'P.',\n\t\t'P.2d',\n\t\t'P.3d',\n\t\t'A.',\n\t\t'A.2d',\n\t\t'A.3d',\n\t\t'N.E.',\n\t\t'N.E.2d',\n\t\t'N.E.3d',\n\t\t'N.W.',\n\t\t'N.W.2d',\n\t\t'S.E.',\n\t\t'S.E.2d',\n\t\t'S.W.',\n\t\t'S.W.2d',\n\t\t'S.W.3d',\n\t\t'So.',\n\t\t'So. 2d',\n\t\t'So. 3d',\n\t]\n\n\tif (commonReporters.some((r) => reporter.includes(r))) {\n\t\tconfidence += 0.3\n\t}\n\n\t// Valid year check (not in future)\n\tif (year !== undefined) {\n\t\tconst currentYear = new Date().getFullYear()\n\t\tif (year <= currentYear) {\n\t\t\tconfidence += 0.2\n\t\t}\n\t}\n\n\t// Cap at 1.0\n\tconfidence = Math.min(confidence, 1.0)\n\n\treturn {\n\t\ttype: 'case',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0, // Placeholder - timing handled by orchestration layer\n\t\tpatternsChecked: 1, // Single token processed\n\t\tvolume,\n\t\treporter,\n\t\tpage,\n\t\tpincite,\n\t\tcourt,\n\t\tyear,\n\t}\n}\n","/**\n * Statute Citation Extraction\n *\n * Parses tokenized statute citations to extract title, code, section, and\n * optional subsections. Examples: \"42 U.S.C. § 1983\", \"Cal. Civ. Code § 1234(a)(1)\"\n *\n * @module extract/extractStatute\n */\n\nimport type { Token } from '@/tokenize'\nimport type { StatuteCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts statute citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Title: Optional leading digits (e.g., \"42\" from \"42 U.S.C. § 1983\")\n * - Code: Statutory code abbreviation (e.g., \"U.S.C.\", \"Cal. Civ. Code\")\n * - Section: Section number after § symbol (e.g., \"1983\")\n * - Subsections: Optional parenthetical subdivisions (e.g., \"(a)(1)\")\n *\n * Confidence scoring:\n * - Base: 0.5\n * - Known code pattern (U.S.C., C.F.R., state codes): +0.3\n * - Capped at 1.0\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns StatuteCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"42 U.S.C. § 1983\",\n *   span: { cleanStart: 10, cleanEnd: 26 },\n *   type: \"statute\",\n *   patternId: \"usc\"\n * }\n * const citation = extractStatute(token, transformationMap)\n * // citation = {\n * //   type: \"statute\",\n * //   title: 42,\n * //   code: \"U.S.C.\",\n * //   section: \"1983\",\n * //   ...\n * // }\n * ```\n */\nexport function extractStatute(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): StatuteCitation {\n\tconst { text, span } = token\n\n\t// Parse title-code-section using regex\n\t// Pattern: optional title (digits) + code (letters/periods/spaces) + § + section\n\tconst statuteRegex = /^(?:(\\d+)\\s+)?([A-Za-z.\\s]+?)\\s*§\\s*(\\d+[A-Za-z0-9\\-]*)/\n\tconst match = statuteRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse statute citation: ${text}`)\n\t}\n\n\tconst title = match[1] ? Number.parseInt(match[1], 10) : undefined\n\tconst code = match[2].trim()\n\tconst section = match[3]\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Calculate confidence score\n\tlet confidence = 0.5 // Base confidence\n\n\t// Known statutory code patterns\n\tconst knownCodes = [\n\t\t'U.S.C.',\n\t\t'C.F.R.',\n\t\t'Cal. Civ. Code',\n\t\t'Cal. Penal Code',\n\t\t'N.Y. Civ. Prac. L. & R.',\n\t\t'Tex. Civ. Prac. & Rem. Code',\n\t]\n\n\tif (knownCodes.some((c) => code.includes(c))) {\n\t\tconfidence += 0.3\n\t}\n\n\tconfidence = Math.min(confidence, 1.0)\n\n\treturn {\n\t\ttype: 'statute',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\ttitle,\n\t\tcode,\n\t\tsection,\n\t}\n}\n","/**\n * Journal Citation Extraction\n *\n * Parses tokenized journal citations to extract volume, journal name, page,\n * and optional metadata. Examples: \"123 Harv. L. Rev. 456\", \"75 Yale L.J. 789, 791\"\n *\n * @module extract/extractJournal\n */\n\nimport type { Token } from '@/tokenize'\nimport type { JournalCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts journal citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Volume: Leading digits (e.g., \"123\" from \"123 Harv. L. Rev. 456\")\n * - Journal: Journal abbreviation (e.g., \"Harv. L. Rev.\")\n * - Page: Starting page number (e.g., \"456\")\n * - Pincite: Optional specific page reference after comma (e.g., \", 458\")\n *\n * Confidence scoring:\n * - Base: 0.6 (journal validation happens in Phase 3)\n *\n * Note: Author and title extraction from preceding text is not implemented\n * in Phase 2. That requires context analysis in Phase 3.\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns JournalCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"123 Harv. L. Rev. 456\",\n *   span: { cleanStart: 10, cleanEnd: 31 },\n *   type: \"journal\",\n *   patternId: \"journal-standard\"\n * }\n * const citation = extractJournal(token, transformationMap)\n * // citation = {\n * //   type: \"journal\",\n * //   volume: 123,\n * //   journal: \"Harv. L. Rev.\",\n * //   abbreviation: \"Harv. L. Rev.\",\n * //   page: 456,\n * //   ...\n * // }\n * ```\n */\nexport function extractJournal(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): JournalCitation {\n\tconst { text, span } = token\n\n\t// Parse volume-journal-page using regex\n\t// Pattern: volume (digits) + journal (letters/periods/spaces) + page (digits)\n\tconst journalRegex = /^(\\d+)\\s+([A-Za-z.\\s]+?)\\s+(\\d+)/\n\tconst match = journalRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse journal citation: ${text}`)\n\t}\n\n\tconst volume = Number.parseInt(match[1], 10)\n\tconst journal = match[2].trim()\n\tconst page = Number.parseInt(match[3], 10)\n\n\t// Extract optional pincite (page reference after comma)\n\tconst pinciteRegex = /,\\s*(\\d+)/\n\tconst pinciteMatch = pinciteRegex.exec(text)\n\tconst pincite = pinciteMatch ? Number.parseInt(pinciteMatch[1], 10) : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.6 base (journal validation against database happens in Phase 3)\n\tconst confidence = 0.6\n\n\treturn {\n\t\ttype: 'journal',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tvolume,\n\t\tjournal,\n\t\tabbreviation: journal, // For Phase 2, abbreviation = journal name\n\t\tpage,\n\t\tpincite,\n\t}\n}\n","/**\n * Neutral Citation Extraction\n *\n * Parses tokenized neutral (vendor-neutral) citations to extract year, court,\n * and document number. Examples: \"2020 WL 123456\", \"2020 U.S. LEXIS 456\"\n *\n * @module extract/extractNeutral\n */\n\nimport type { Token } from '@/tokenize'\nimport type { NeutralCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts neutral citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Year: 4-digit year (e.g., \"2020\")\n * - Court: Vendor identifier (e.g., \"WL\", \"U.S. LEXIS\")\n * - Document number: Unique document identifier (e.g., \"123456\")\n *\n * Confidence scoring:\n * - 1.0 (neutral format is unambiguous and standardized)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns NeutralCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"2020 WL 123456\",\n *   span: { cleanStart: 10, cleanEnd: 24 },\n *   type: \"neutral\",\n *   patternId: \"westlaw-neutral\"\n * }\n * const citation = extractNeutral(token, transformationMap)\n * // citation = {\n * //   type: \"neutral\",\n * //   year: 2020,\n * //   court: \"WL\",\n * //   documentNumber: \"123456\",\n * //   confidence: 1.0,\n * //   ...\n * // }\n * ```\n */\nexport function extractNeutral(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): NeutralCitation {\n\tconst { text, span } = token\n\n\t// Parse year-court-documentNumber using regex\n\t// Pattern: 4-digit year + court identifier (WL, LEXIS, etc.) + document number\n\tconst neutralRegex = /^(\\d{4})\\s+(WL|LEXIS|U\\.S\\.\\s+LEXIS)\\s+(\\d+)/\n\tconst match = neutralRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse neutral citation: ${text}`)\n\t}\n\n\tconst year = Number.parseInt(match[1], 10)\n\tconst court = match[2]\n\tconst documentNumber = match[3]\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 1.0 (neutral format is unambiguous)\n\tconst confidence = 1.0\n\n\treturn {\n\t\ttype: 'neutral',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tyear,\n\t\tcourt,\n\t\tdocumentNumber,\n\t}\n}\n","/**\n * Public Law Citation Extraction\n *\n * Parses tokenized public law citations to extract congress number and law number.\n * Examples: \"Pub. L. No. 116-283\", \"Pub. L. 117-58\"\n *\n * @module extract/extractPublicLaw\n */\n\nimport type { Token } from '@/tokenize'\nimport type { PublicLawCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts public law citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Congress: Congress number (e.g., \"116\" from \"Pub. L. No. 116-283\")\n * - Law number: Law number within that Congress (e.g., \"283\")\n *\n * Confidence scoring:\n * - 0.9 (public law format is fairly standard)\n *\n * Note: Bill title extraction from nearby text is not implemented in Phase 2.\n * That requires context analysis in Phase 3.\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns PublicLawCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"Pub. L. No. 116-283\",\n *   span: { cleanStart: 10, cleanEnd: 29 },\n *   type: \"publicLaw\",\n *   patternId: \"public-law\"\n * }\n * const citation = extractPublicLaw(token, transformationMap)\n * // citation = {\n * //   type: \"publicLaw\",\n * //   congress: 116,\n * //   lawNumber: 283,\n * //   confidence: 0.9,\n * //   ...\n * // }\n * ```\n */\nexport function extractPublicLaw(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): PublicLawCitation {\n\tconst { text, span } = token\n\n\t// Parse congress-lawNumber using regex\n\t// Pattern: \"Pub. L.\" (with optional \"No.\") + congress number + \"-\" + law number\n\tconst publicLawRegex = /Pub\\.\\s?L\\.(?:\\s?No\\.)?\\s?(\\d+)-(\\d+)/\n\tconst match = publicLawRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse public law citation: ${text}`)\n\t}\n\n\tconst congress = Number.parseInt(match[1], 10)\n\tconst lawNumber = Number.parseInt(match[2], 10)\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.9 (public law format is fairly standard)\n\tconst confidence = 0.9\n\n\treturn {\n\t\ttype: 'publicLaw',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tcongress,\n\t\tlawNumber,\n\t}\n}\n","/**\n * Federal Register Citation Extraction\n *\n * Parses tokenized Federal Register citations to extract volume, page, and\n * optional year. Examples: \"85 Fed. Reg. 12345\", \"86 Fed. Reg. 56789 (Jan. 15, 2021)\"\n *\n * @module extract/extractFederalRegister\n */\n\nimport type { Token } from '@/tokenize'\nimport type { FederalRegisterCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts Federal Register citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Volume: Federal Register volume number (e.g., \"85\")\n * - Page: Page number (e.g., \"12345\")\n * - Year: Optional publication year in parentheses (e.g., \"(2021)\")\n *\n * Confidence scoring:\n * - 0.9 (Federal Register format is standardized)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns FederalRegisterCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"85 Fed. Reg. 12345\",\n *   span: { cleanStart: 10, cleanEnd: 28 },\n *   type: \"federalRegister\",\n *   patternId: \"federal-register\"\n * }\n * const citation = extractFederalRegister(token, transformationMap)\n * // citation = {\n * //   type: \"federalRegister\",\n * //   volume: 85,\n * //   page: 12345,\n * //   confidence: 0.9,\n * //   ...\n * // }\n * ```\n */\nexport function extractFederalRegister(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): FederalRegisterCitation {\n\tconst { text, span } = token\n\n\t// Parse volume-page using regex\n\t// Pattern: volume (digits) + \"Fed. Reg.\" + page (digits)\n\tconst federalRegisterRegex = /^(\\d+)\\s+Fed\\.\\s?Reg\\.\\s+(\\d+)/\n\tconst match = federalRegisterRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse Federal Register citation: ${text}`)\n\t}\n\n\tconst volume = Number.parseInt(match[1], 10)\n\tconst page = Number.parseInt(match[2], 10)\n\n\t// Extract optional year in parentheses\n\t// Pattern: \"(year)\" or \"(month day, year)\"\n\tconst yearRegex = /\\((?:.*?\\s)?(\\d{4})\\)/\n\tconst yearMatch = yearRegex.exec(text)\n\tconst year = yearMatch ? Number.parseInt(yearMatch[1], 10) : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.9 (Federal Register format is standardized)\n\tconst confidence = 0.9\n\n\treturn {\n\t\ttype: 'federalRegister',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tvolume,\n\t\tpage,\n\t\tyear,\n\t}\n}\n","/**\n * Short-form Citation Extraction\n *\n * Parses tokenized short-form citations (Id., supra, short-form case) to extract\n * metadata. Short-form citations refer to earlier citations in the document.\n *\n * @module extract/extractShortForms\n */\n\nimport type { Token } from '@/tokenize'\nimport type { IdCitation, SupraCitation, ShortFormCaseCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts Id. citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Pincite: Optional page reference (e.g., \"253\" from \"Id. at 253\")\n *\n * Confidence scoring:\n * - 1.0 (Id. format is unambiguous and standardized)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns IdCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"Id. at 253\",\n *   span: { cleanStart: 10, cleanEnd: 20 },\n *   type: \"case\",\n *   patternId: \"id\"\n * }\n * const citation = extractId(token, transformationMap)\n * // citation = {\n * //   type: \"id\",\n * //   pincite: 253,\n * //   confidence: 1.0,\n * //   ...\n * // }\n * ```\n */\nexport function extractId(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): IdCitation {\n\tconst { text, span } = token\n\n\t// Parse Id. with optional pincite\n\t// Pattern: Id. or Ibid. with optional \"at [page]\"\n\tconst idRegex = /[Ii](?:d|bid)\\.(?:\\s+at\\s+(\\d+))?/\n\tconst match = idRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse Id. citation: ${text}`)\n\t}\n\n\t// Extract pincite if present\n\tconst pincite = match[1] ? Number.parseInt(match[1], 10) : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 1.0 (Id. format is unambiguous)\n\tconst confidence = 1.0\n\n\treturn {\n\t\ttype: 'id',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tpincite,\n\t}\n}\n\n/**\n * Extracts supra citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Party name: Name preceding \"supra\" (e.g., \"Smith\" from \"Smith, supra\")\n * - Pincite: Optional page reference (e.g., \"460\" from \"Smith, supra, at 460\")\n *\n * Confidence scoring:\n * - 0.9 (supra format is fairly standard but party name extraction can vary)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns SupraCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"Smith, supra, at 460\",\n *   span: { cleanStart: 10, cleanEnd: 30 },\n *   type: \"case\",\n *   patternId: \"supra\"\n * }\n * const citation = extractSupra(token, transformationMap)\n * // citation = {\n * //   type: \"supra\",\n * //   partyName: \"Smith\",\n * //   pincite: 460,\n * //   confidence: 0.9,\n * //   ...\n * // }\n * ```\n */\nexport function extractSupra(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): SupraCitation {\n\tconst { text, span } = token\n\n\t// Parse party name and optional pincite\n\t// Pattern: word(s), supra [, at page]\n\t// Note: Matches party names including \"v.\" (e.g., \"Smith v. Jones\")\n\tconst supraRegex = /\\b([A-Z][a-zA-Z]+(?:(?:\\s+v\\.?\\s+|\\s+)[A-Z][a-zA-Z]+)*),?\\s+supra(?:,?\\s+at\\s+(\\d+))?/\n\tconst match = supraRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse supra citation: ${text}`)\n\t}\n\n\tconst partyName = match[1]\n\tconst pincite = match[2] ? Number.parseInt(match[2], 10) : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.9 (supra format is fairly standard)\n\tconst confidence = 0.9\n\n\treturn {\n\t\ttype: 'supra',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tpartyName,\n\t\tpincite,\n\t}\n}\n\n/**\n * Extracts short-form case citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Volume: Volume number\n * - Reporter: Reporter abbreviation\n * - Pincite: Page reference (from \"at [page]\" pattern)\n *\n * Confidence scoring:\n * - 0.7 (short-form case citations are more ambiguous than full citations)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns ShortFormCaseCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"500 F.2d at 125\",\n *   span: { cleanStart: 10, cleanEnd: 25 },\n *   type: \"case\",\n *   patternId: \"short-form-case\"\n * }\n * const citation = extractShortFormCase(token, transformationMap)\n * // citation = {\n * //   type: \"shortFormCase\",\n * //   volume: 500,\n * //   reporter: \"F.2d\",\n * //   pincite: 125,\n * //   confidence: 0.7,\n * //   ...\n * // }\n * ```\n */\nexport function extractShortFormCase(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): ShortFormCaseCitation {\n\tconst { text, span } = token\n\n\t// Parse volume-reporter-at-page\n\t// Pattern: number space abbreviation space \"at\" space number\n\tconst shortFormRegex = /(\\d+)\\s+([A-Z][A-Za-z.\\s]+?(?:\\d[a-z])?)\\s+at\\s+(\\d+)/\n\tconst match = shortFormRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse short-form case citation: ${text}`)\n\t}\n\n\tconst volume = Number.parseInt(match[1], 10)\n\tconst reporter = match[2].trim() // Remove trailing spaces\n\tconst pincite = Number.parseInt(match[3], 10)\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.7 (short-form citations are more ambiguous)\n\tconst confidence = 0.7\n\n\treturn {\n\t\ttype: 'shortFormCase',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tvolume,\n\t\treporter,\n\t\tpincite,\n\t}\n}\n","/**\n * Scope Boundary Detection\n *\n * Detects paragraph/section boundaries in text and validates whether\n * an antecedent citation is within the resolution scope.\n */\n\nimport type { Citation } from '../types/citation'\nimport type { ScopeStrategy } from './types'\n\n/**\n * Detects paragraph boundaries from text and assigns each citation to a paragraph.\n *\n * @param text - Original document text\n * @param citations - Extracted citations with position spans\n * @param boundaryPattern - Regex pattern to detect boundaries (default: /\\n\\n+/)\n * @returns Map of citation index to paragraph number (0-based)\n */\nexport function detectParagraphBoundaries(\n  text: string,\n  citations: Citation[],\n  boundaryPattern: RegExp = /\\n\\n+/g\n): Map<number, number> {\n  const paragraphMap = new Map<number, number>()\n\n  // Find all paragraph boundaries (positions in text)\n  const boundaries: number[] = [0] // Start of document is first boundary\n  let match: RegExpExecArray | null\n\n  while ((match = boundaryPattern.exec(text)) !== null) {\n    // Boundary is at end of match (start of next paragraph)\n    boundaries.push(match.index + match[0].length)\n  }\n\n  boundaries.push(text.length) // End of document\n\n  // Assign each citation to a paragraph\n  for (let i = 0; i < citations.length; i++) {\n    const citation = citations[i]\n    const citationStart = citation.span.originalStart\n\n    // Find which paragraph this citation belongs to\n    let paragraphNum = 0\n    for (let j = 0; j < boundaries.length - 1; j++) {\n      if (citationStart >= boundaries[j] && citationStart < boundaries[j + 1]) {\n        paragraphNum = j\n        break\n      }\n    }\n\n    paragraphMap.set(i, paragraphNum)\n  }\n\n  return paragraphMap\n}\n\n/**\n * Checks if an antecedent citation is within resolution scope.\n *\n * @param antecedentIndex - Index of the antecedent citation\n * @param currentIndex - Index of current citation being resolved\n * @param paragraphMap - Map of citation index to paragraph number\n * @param strategy - Scope boundary strategy\n * @returns true if antecedent is within scope, false otherwise\n */\nexport function isWithinBoundary(\n  antecedentIndex: number,\n  currentIndex: number,\n  paragraphMap: Map<number, number>,\n  strategy: ScopeStrategy\n): boolean {\n  if (strategy === 'none') {\n    // No boundary restriction - can resolve across entire document\n    return true\n  }\n\n  // Get paragraph numbers for both citations\n  const antecedentParagraph = paragraphMap.get(antecedentIndex)\n  const currentParagraph = paragraphMap.get(currentIndex)\n\n  // If either is undefined, default to allowing resolution\n  if (antecedentParagraph === undefined || currentParagraph === undefined) {\n    return true\n  }\n\n  // For paragraph/section/footnote strategies, citations must be in same boundary\n  // (In this MVP, section and footnote behave same as paragraph - future enhancement)\n  return antecedentParagraph === currentParagraph\n}\n","/**\n * Levenshtein Distance\n *\n * Calculates edit distance between strings for fuzzy party name matching\n * in supra citation resolution.\n *\n * Uses dynamic programming for O(m*n) time complexity.\n */\n\n/**\n * Calculates Levenshtein distance (edit distance) between two strings.\n *\n * The edit distance is the minimum number of single-character edits\n * (insertions, deletions, substitutions) needed to change one string into the other.\n *\n * @param a - First string\n * @param b - Second string\n * @returns Number of edits required (0 = identical)\n */\nexport function levenshteinDistance(a: string, b: string): number {\n  // Handle empty strings\n  if (a.length === 0) return b.length\n  if (b.length === 0) return a.length\n\n  // Create 2D array for dynamic programming\n  // dp[i][j] = edit distance between a[0...i-1] and b[0...j-1]\n  const dp: number[][] = Array.from({ length: a.length + 1 }, () =>\n    Array(b.length + 1).fill(0)\n  )\n\n  // Initialize base cases\n  for (let i = 0; i <= a.length; i++) {\n    dp[i][0] = i // Distance from a[0...i-1] to empty string\n  }\n  for (let j = 0; j <= b.length; j++) {\n    dp[0][j] = j // Distance from empty string to b[0...j-1]\n  }\n\n  // Fill the DP table\n  for (let i = 1; i <= a.length; i++) {\n    for (let j = 1; j <= b.length; j++) {\n      if (a[i - 1] === b[j - 1]) {\n        // Characters match - no edit needed\n        dp[i][j] = dp[i - 1][j - 1]\n      } else {\n        // Characters differ - take minimum of:\n        // 1. Insert: dp[i][j-1] + 1\n        // 2. Delete: dp[i-1][j] + 1\n        // 3. Substitute: dp[i-1][j-1] + 1\n        dp[i][j] = 1 + Math.min(\n          dp[i - 1][j],     // Delete from a\n          dp[i][j - 1],     // Insert into a\n          dp[i - 1][j - 1]  // Substitute\n        )\n      }\n    }\n  }\n\n  return dp[a.length][b.length]\n}\n\n/**\n * Calculates normalized Levenshtein similarity (0-1 scale).\n *\n * Returns similarity score where:\n * - 1.0 = identical strings\n * - 0.0 = completely different\n *\n * Comparison is case-insensitive.\n *\n * @param a - First string\n * @param b - Second string\n * @returns Similarity score from 0 to 1\n */\nexport function normalizedLevenshteinDistance(a: string, b: string): number {\n  // Normalize to lowercase for case-insensitive comparison\n  const lowerA = a.toLowerCase()\n  const lowerB = b.toLowerCase()\n\n  // Calculate raw edit distance\n  const distance = levenshteinDistance(lowerA, lowerB)\n\n  // Normalize by max length\n  const maxLength = Math.max(lowerA.length, lowerB.length)\n  if (maxLength === 0) return 1.0 // Both empty strings\n\n  // Convert distance to similarity: 1 - (distance / maxLength)\n  return 1 - distance / maxLength\n}\n","/**\n * Document-Scoped Citation Resolver\n *\n * Resolves short-form citations (Id./supra/short-form case) to their full antecedent citations\n * by maintaining resolution context and enforcing scope boundaries.\n *\n * Resolution rules:\n * - Id. resolves to immediately preceding full citation (within scope)\n * - Supra resolves to full citation with matching party name (within scope)\n * - Short-form case resolves to full case with matching volume/reporter (within scope)\n */\n\nimport type {\n  Citation,\n  FullCaseCitation,\n  IdCitation,\n  SupraCitation,\n  ShortFormCaseCitation,\n} from '../types/citation'\nimport { isFullCitation } from '../types/guards'\nimport type {\n  ResolutionOptions,\n  ResolutionResult,\n  ResolvedCitation,\n  ResolutionContext,\n} from './types'\nimport { detectParagraphBoundaries, isWithinBoundary } from './scopeBoundary'\nimport { normalizedLevenshteinDistance } from './levenshtein'\n\n/**\n * Document-scoped resolver that processes citations sequentially\n * and resolves short-form citations to their antecedents.\n */\nexport class DocumentResolver {\n  private readonly citations: Citation[]\n  private readonly text: string\n  private readonly options: Required<ResolutionOptions>\n  private readonly context: ResolutionContext\n\n  /**\n   * Creates a new DocumentResolver.\n   *\n   * @param citations - All citations in document (in order of appearance)\n   * @param text - Original document text\n   * @param options - Resolution options\n   */\n  constructor(\n    citations: Citation[],\n    text: string,\n    options: ResolutionOptions = {}\n  ) {\n    this.citations = citations\n    this.text = text\n\n    // Apply defaults to options\n    this.options = {\n      scopeStrategy: options.scopeStrategy ?? 'paragraph',\n      autoDetectParagraphs: options.autoDetectParagraphs ?? true,\n      paragraphBoundaryPattern: options.paragraphBoundaryPattern ?? /\\n\\n+/g,\n      fuzzyPartyMatching: options.fuzzyPartyMatching ?? true,\n      partyMatchThreshold: options.partyMatchThreshold ?? 0.8,\n      allowNestedResolution: options.allowNestedResolution ?? false,\n      reportUnresolved: options.reportUnresolved ?? true,\n    }\n\n    // Initialize resolution context\n    this.context = {\n      citationIndex: 0,\n      allCitations: citations,\n      lastFullCitation: undefined,\n      fullCitationHistory: new Map(),\n      paragraphMap: new Map(),\n    }\n\n    // Detect paragraph boundaries if enabled\n    if (this.options.autoDetectParagraphs) {\n      this.context.paragraphMap = detectParagraphBoundaries(\n        text,\n        citations,\n        this.options.paragraphBoundaryPattern\n      )\n    }\n  }\n\n  /**\n   * Resolves all citations in the document.\n   *\n   * @returns Array of citations with resolution metadata\n   */\n  resolve(): ResolvedCitation[] {\n    const resolved: ResolvedCitation[] = []\n\n    for (let i = 0; i < this.citations.length; i++) {\n      this.context.citationIndex = i\n      const citation = this.citations[i]\n\n      // Resolve based on citation type\n      let resolution: ResolutionResult | undefined\n\n      switch (citation.type) {\n        case 'id':\n          resolution = this.resolveId(citation)\n          break\n        case 'supra':\n          resolution = this.resolveSupra(citation)\n          break\n        case 'shortFormCase':\n          resolution = this.resolveShortFormCase(citation)\n          break\n        default:\n          // Full citation - update context for future resolutions\n          if (isFullCitation(citation)) {\n            this.context.lastFullCitation = i\n            this.trackFullCitation(citation, i)\n          }\n          break\n      }\n\n      // Add citation with resolution metadata\n      // Type assertion is safe: runtime logic only sets resolution on short-form citations\n      resolved.push({\n        ...citation,\n        resolution,\n      } as ResolvedCitation)\n    }\n\n    return resolved\n  }\n\n  /**\n   * Resolves Id. citation to immediately preceding full case citation.\n   */\n  private resolveId(citation: IdCitation): ResolutionResult | undefined {\n    const currentIndex = this.context.citationIndex\n\n    // Find most recent full case citation (Id. only resolves to case citations, not statutes/journals)\n    let antecedentIndex: number | undefined\n    for (let i = currentIndex - 1; i >= 0; i--) {\n      const candidate = this.citations[i]\n      if (candidate.type === 'case') {\n        antecedentIndex = i\n        break\n      }\n    }\n\n    // Check if we have a previous case citation\n    if (antecedentIndex === undefined) {\n      return this.createFailureResult('No preceding full case citation found')\n    }\n\n    // Check scope boundary\n    if (!this.isWithinScope(antecedentIndex, currentIndex)) {\n      return this.createFailureResult('Antecedent citation outside scope boundary')\n    }\n\n    return {\n      resolvedTo: antecedentIndex,\n      confidence: 1.0, // Id. resolution is unambiguous when successful\n    }\n  }\n\n  /**\n   * Resolves supra citation by matching party name.\n   */\n  private resolveSupra(citation: SupraCitation): ResolutionResult | undefined {\n    const currentIndex = this.context.citationIndex\n    const targetPartyName = this.normalizePartyName(citation.partyName)\n\n    // Search full citation history for matching party name\n    let bestMatch: { index: number; similarity: number } | undefined\n\n    for (const [partyName, citationIndex] of this.context.fullCitationHistory) {\n      // Check scope boundary\n      if (!this.isWithinScope(citationIndex, currentIndex)) {\n        continue\n      }\n\n      // Calculate similarity\n      const similarity = normalizedLevenshteinDistance(targetPartyName, partyName)\n\n      // Update best match if this is better\n      if (!bestMatch || similarity > bestMatch.similarity) {\n        bestMatch = { index: citationIndex, similarity }\n      }\n    }\n\n    // Check if we found a match above threshold\n    if (!bestMatch) {\n      return this.createFailureResult('No full citation found in scope')\n    }\n\n    if (bestMatch.similarity < this.options.partyMatchThreshold) {\n      return this.createFailureResult(\n        `Party name similarity ${bestMatch.similarity.toFixed(2)} below threshold ${this.options.partyMatchThreshold}`\n      )\n    }\n\n    // Return successful resolution with confidence based on similarity\n    const warnings: string[] = []\n    if (bestMatch.similarity < 1.0) {\n      warnings.push(`Fuzzy match: similarity ${bestMatch.similarity.toFixed(2)}`)\n    }\n\n    return {\n      resolvedTo: bestMatch.index,\n      confidence: bestMatch.similarity,\n      warnings: warnings.length > 0 ? warnings : undefined,\n    }\n  }\n\n  /**\n   * Resolves short-form case citation by matching volume/reporter.\n   */\n  private resolveShortFormCase(citation: ShortFormCaseCitation): ResolutionResult | undefined {\n    const currentIndex = this.context.citationIndex\n\n    // Search backwards for matching full case citation\n    for (let i = currentIndex - 1; i >= 0; i--) {\n      const candidate = this.citations[i]\n\n      // Only match against full case citations\n      if (candidate.type !== 'case') {\n        continue\n      }\n\n      // Check if volume and reporter match\n      if (\n        candidate.volume === citation.volume &&\n        this.normalizeReporter(candidate.reporter) === this.normalizeReporter(citation.reporter)\n      ) {\n        // Check scope boundary\n        if (!this.isWithinScope(i, currentIndex)) {\n          return this.createFailureResult('Matching citation outside scope boundary')\n        }\n\n        // Found a match\n        return {\n          resolvedTo: i,\n          confidence: 0.95, // High confidence but not perfect (multiple cases could have same volume/reporter)\n        }\n      }\n    }\n\n    return this.createFailureResult('No matching full case citation found')\n  }\n\n  /**\n   * Tracks a full citation in the resolution history.\n   * Extracts party name for supra resolution.\n   */\n  private trackFullCitation(citation: Citation, index: number): void {\n    // Only case citations have party names for supra resolution\n    if (citation.type === 'case') {\n      const partyName = this.extractPartyName(citation)\n      if (partyName) {\n        const normalized = this.normalizePartyName(partyName)\n        this.context.fullCitationHistory.set(normalized, index)\n      }\n    }\n  }\n\n  /**\n   * Extracts party name from full case citation text.\n   * Handles \"Party v. Party\" format by looking at text before citation span.\n   */\n  private extractPartyName(citation: FullCaseCitation): string | undefined {\n    // Look at text before citation span to find party names\n    // Case citations typically appear as: \"Smith v. Jones, 100 F.2d 10\"\n    // But tokenizer only captures \"100 F.2d 10\" - we need to look backwards in text\n\n    const citationStart = citation.span.originalStart\n    // Look backwards up to 100 characters for party name\n    const lookbackStart = Math.max(0, citationStart - 100)\n    const beforeText = this.text.substring(lookbackStart, citationStart)\n\n    // Match pattern: \"FirstParty v. SecondParty, \" before the citation\n    // Capture the first party name (handles single-letter party names like \"A\" or \"B\")\n    const vMatch = beforeText.match(/([A-Z][a-zA-Z]*(?:\\s+[A-Z][a-zA-Z]*)*)\\s+v\\.?\\s+[A-Z][a-zA-Z]*(?:\\s+[A-Z][a-zA-Z]*)*,\\s*$/)\n    if (vMatch) {\n      return vMatch[1].trim()\n    }\n\n    // Fallback: try to find any capitalized word(s) before comma\n    const beforeComma = beforeText.match(/([A-Z][a-zA-Z]*(?:\\s+[A-Z][a-zA-Z]*)*),\\s*$/)\n    return beforeComma?.[1].trim()\n  }\n\n  /**\n   * Normalizes party name for matching.\n   */\n  private normalizePartyName(name: string): string {\n    return name\n      .toLowerCase()\n      .replace(/\\s+/g, ' ') // Normalize whitespace\n      .trim()\n  }\n\n  /**\n   * Normalizes reporter abbreviation for matching.\n   */\n  private normalizeReporter(reporter: string): string {\n    return reporter\n      .toLowerCase()\n      .replace(/\\s+/g, '') // Remove spaces (F.2d vs F. 2d)\n      .replace(/\\./g, '')   // Remove periods\n  }\n\n  /**\n   * Checks if antecedent citation is within scope boundary.\n   */\n  private isWithinScope(antecedentIndex: number, currentIndex: number): boolean {\n    return isWithinBoundary(\n      antecedentIndex,\n      currentIndex,\n      this.context.paragraphMap,\n      this.options.scopeStrategy\n    )\n  }\n\n  /**\n   * Creates a failure result for unresolved citations.\n   */\n  private createFailureResult(reason: string): ResolutionResult | undefined {\n    if (this.options.reportUnresolved) {\n      return {\n        resolvedTo: undefined,\n        failureReason: reason,\n        confidence: 0.0,\n      }\n    }\n    return undefined\n  }\n}\n","/**\n * Citation Resolution\n *\n * Resolves short-form citations (Id./supra/short-form case) to their full antecedents.\n *\n * @example\n * ```ts\n * import { resolveCitations } from 'eyecite-ts/resolve'\n * import { extractCitations } from 'eyecite-ts'\n *\n * const text = 'See Smith v. Jones, 500 F.2d 100 (1974). Id. at 105.'\n * const citations = extractCitations(text)\n * const resolved = resolveCitations(citations, text)\n *\n * // resolved[1] is Id. citation with resolution.resolvedTo = 0\n * console.log(resolved[1].resolution?.resolvedTo) // 0 (points to Smith v. Jones)\n * ```\n */\n\nimport type { Citation } from '../types/citation'\nimport type { ResolutionOptions, ResolvedCitation } from './types'\nimport { DocumentResolver } from './DocumentResolver'\n\n/**\n * Resolves short-form citations to their full antecedents.\n *\n * Convenience wrapper around DocumentResolver that handles common use cases.\n *\n * @param citations - Extracted citations in order of appearance\n * @param text - Original document text\n * @param options - Resolution options\n * @returns Citations with resolution metadata\n */\nexport function resolveCitations(\n  citations: Citation[],\n  text: string,\n  options?: ResolutionOptions\n): ResolvedCitation[] {\n  const resolver = new DocumentResolver(citations, text, options)\n  return resolver.resolve()\n}\n\n// Re-export core types and classes\nexport { DocumentResolver } from './DocumentResolver'\nexport type {\n  ResolutionOptions,\n  ResolutionResult,\n  ResolvedCitation,\n  ScopeStrategy,\n} from './types'\n","/**\n * Main Citation Extraction Pipeline\n *\n * Orchestrates the complete citation extraction flow:\n *   1. Clean text (remove HTML, normalize Unicode)\n *   2. Tokenize (apply patterns to find candidates)\n *   3. Extract (parse metadata from tokens)\n *\n * This is the primary public API for citation extraction.\n *\n * @module extract/extractCitations\n */\n\nimport { cleanText } from '@/clean'\nimport { tokenize } from '@/tokenize'\nimport {\n\textractCase,\n\textractStatute,\n\textractJournal,\n\textractNeutral,\n\textractPublicLaw,\n\textractFederalRegister,\n} from '@/extract'\nimport { extractId, extractSupra, extractShortFormCase } from './extractShortForms'\nimport {\n\tcasePatterns,\n\tstatutePatterns,\n\tjournalPatterns,\n\tneutralPatterns,\n\tshortFormPatterns,\n} from '@/patterns'\nimport { resolveCitations } from '../resolve'\nimport type { Citation } from '@/types/citation'\nimport type { Pattern } from '@/patterns'\nimport type { ResolutionOptions, ResolvedCitation } from '../resolve/types'\n\n/**\n * Options for customizing citation extraction behavior.\n */\nexport interface ExtractOptions {\n\t/**\n\t * Custom text cleaners (overrides defaults).\n\t *\n\t * If provided, these cleaners replace the default pipeline:\n\t * [stripHtmlTags, normalizeWhitespace, normalizeUnicode, fixSmartQuotes]\n\t *\n\t * @example\n\t * ```typescript\n\t * // Use only HTML stripping, skip Unicode normalization\n\t * const citations = extractCitations(text, {\n\t *   cleaners: [stripHtmlTags]\n\t * })\n\t * ```\n\t */\n\tcleaners?: Array<(text: string) => string>\n\n\t/**\n\t * Custom regex patterns (overrides defaults).\n\t *\n\t * If provided, these patterns replace the default pattern set:\n\t * [casePatterns, statutePatterns, journalPatterns, neutralPatterns, shortFormPatterns]\n\t *\n\t * @example\n\t * ```typescript\n\t * // Extract only case citations\n\t * const citations = extractCitations(text, {\n\t *   patterns: casePatterns\n\t * })\n\t * ```\n\t */\n\tpatterns?: Pattern[]\n\n\t/**\n\t * Resolve short-form citations to their full antecedents (default: false).\n\t *\n\t * If true, returns ResolvedCitation[] with resolution metadata for short-form citations\n\t * (Id., supra, short-form case). Full citations are unchanged.\n\t *\n\t * @example\n\t * ```typescript\n\t * const text = \"Smith v. Jones, 500 F.2d 100 (1974). Id. at 105.\"\n\t * const citations = extractCitations(text, { resolve: true })\n\t * // citations[1].resolution.resolvedTo === 0 (points to Smith v. Jones)\n\t * ```\n\t */\n\tresolve?: boolean\n\n\t/**\n\t * Options for citation resolution (only used if resolve: true).\n\t *\n\t * @example\n\t * ```typescript\n\t * const citations = extractCitations(text, {\n\t *   resolve: true,\n\t *   resolutionOptions: {\n\t *     scopeStrategy: 'paragraph',\n\t *     fuzzyPartyMatching: true\n\t *   }\n\t * })\n\t * ```\n\t */\n\tresolutionOptions?: ResolutionOptions\n}\n\n/**\n * Extracts legal citations from text using the full parsing pipeline.\n *\n * Pipeline flow:\n * 1. **Clean:** Remove HTML tags, normalize Unicode, fix smart quotes\n * 2. **Tokenize:** Apply regex patterns to find citation candidates\n * 3. **Extract:** Parse metadata (volume, reporter, page, etc.)\n * 4. **Translate:** Map positions from cleaned text back to original text\n *\n * This function is synchronous because all stages (cleaning, tokenization,\n * extraction) are synchronous. For async operations (e.g., future reporters-db\n * lookups), use extractCitationsAsync().\n *\n * Position tracking:\n * - TransformationMap is built during cleaning\n * - Tokens contain positions in cleaned text (cleanStart/cleanEnd)\n * - Extraction translates cleaned positions → original positions\n * - Final citations have originalStart/originalEnd pointing to input text\n *\n * Warnings from cleaning layer are attached to all extracted citations.\n *\n * @param text - Raw text to extract citations from (may contain HTML, Unicode)\n * @param options - Optional customization (cleaners, patterns)\n * @returns Array of citations with parsed metadata and accurate positions\n *\n * @example\n * ```typescript\n * const text = \"See Smith v. Doe, 500 F.2d 123 (9th Cir. 2020)\"\n * const citations = extractCitations(text)\n * // citations[0] = {\n * //   type: \"case\",\n * //   volume: 500,\n * //   reporter: \"F.2d\",\n * //   page: 123,\n * //   court: \"9th Cir.\",\n * //   year: 2020,\n * //   span: { originalStart: 18, originalEnd: 30, ... }\n * // }\n * ```\n *\n * @example\n * ```typescript\n * // Extract from HTML\n * const html = \"<p>In <b>Smith</b>, 500 F.2d 123, the court held...</p>\"\n * const citations = extractCitations(html)\n * // HTML is stripped, positions point to original HTML\n * ```\n *\n * @example\n * ```typescript\n * // Extract multiple citation types\n * const text = \"See 42 U.S.C. § 1983; Smith, 500 F.2d 123; 123 Harv. L. Rev. 456\"\n * const citations = extractCitations(text)\n * // citations[0].type === \"statute\"\n * // citations[1].type === \"case\"\n * // citations[2].type === \"journal\"\n * ```\n */\nexport function extractCitations(text: string, options: ExtractOptions & { resolve: true }): ResolvedCitation[]\nexport function extractCitations(text: string, options?: ExtractOptions): Citation[]\nexport function extractCitations(\n\ttext: string,\n\toptions?: ExtractOptions,\n): Citation[] | ResolvedCitation[] {\n\tconst startTime = performance.now()\n\n\t// Step 1: Clean text\n\tconst { cleaned, transformationMap, warnings } = cleanText(\n\t\ttext,\n\t\toptions?.cleaners,\n\t)\n\n\t// Step 2: Tokenize (synchronous)\n\t// Note: Pattern order matters for deduplication - more specific patterns first\n\tconst allPatterns = options?.patterns || [\n\t\t...neutralPatterns,      // Most specific (year-based format)\n\t\t...shortFormPatterns,    // Short-form (requires \" at \" keyword)\n\t\t...casePatterns,         // Case citations (reporter-specific)\n\t\t...statutePatterns,      // Statutes (code-specific)\n\t\t...journalPatterns,      // Least specific (broad pattern)\n\t]\n\tconst tokens = tokenize(cleaned, allPatterns)\n\n\t// Step 3: Deduplicate overlapping tokens\n\t// Multiple patterns may match the same text (e.g., \"500 F.2d 123\" matches both federal-reporter and state-reporter)\n\t// Keep only the most specific match for each position\n\tconst deduplicatedTokens: typeof tokens = []\n\tconst seenPositions = new Set<string>()\n\n\tfor (const token of tokens) {\n\t\tconst posKey = `${token.span.cleanStart}-${token.span.cleanEnd}`\n\t\tif (!seenPositions.has(posKey)) {\n\t\t\tseenPositions.add(posKey)\n\t\t\tdeduplicatedTokens.push(token)\n\t\t}\n\t}\n\n\t// Step 4: Extract citations from deduplicated tokens\n\tconst citations: Citation[] = []\n\tfor (const token of deduplicatedTokens) {\n\t\tlet citation: Citation\n\n\t\tswitch (token.type) {\n\t\t\tcase 'case':\n\t\t\t\t// Check pattern ID to distinguish short-form from full citations\n\t\t\t\tif (token.patternId === 'id' || token.patternId === 'ibid') {\n\t\t\t\t\tcitation = extractId(token, transformationMap)\n\t\t\t\t} else if (token.patternId === 'supra') {\n\t\t\t\t\tcitation = extractSupra(token, transformationMap)\n\t\t\t\t} else if (token.patternId === 'shortFormCase') {\n\t\t\t\t\tcitation = extractShortFormCase(token, transformationMap)\n\t\t\t\t} else {\n\t\t\t\t\tcitation = extractCase(token, transformationMap)\n\t\t\t\t}\n\t\t\t\tbreak\n\t\t\tcase 'statute':\n\t\t\t\tcitation = extractStatute(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tcase 'journal':\n\t\t\t\tcitation = extractJournal(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tcase 'neutral':\n\t\t\t\tcitation = extractNeutral(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tcase 'publicLaw':\n\t\t\t\tcitation = extractPublicLaw(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tcase 'federalRegister':\n\t\t\t\tcitation = extractFederalRegister(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tdefault:\n\t\t\t\t// Unknown type - skip\n\t\t\t\tcontinue\n\t\t}\n\n\t\t// Attach cleaning warnings to citation if any\n\t\tif (warnings.length > 0) {\n\t\t\tcitation.warnings = [...(citation.warnings || []), ...warnings]\n\t\t}\n\n\t\t// Update processing time\n\t\tcitation.processTimeMs = performance.now() - startTime\n\n\t\tcitations.push(citation)\n\t}\n\n\t// Step 5: Resolve short-form citations if requested\n\tif (options?.resolve) {\n\t\treturn resolveCitations(citations, text, options.resolutionOptions)\n\t}\n\n\treturn citations\n}\n\n/**\n * Asynchronous version of extractCitations().\n *\n * Currently wraps the synchronous extractCitations() function. This API\n * exists for future extensibility when async operations are added:\n * - Async reporters-db lookups (Phase 3)\n * - Async resolution/annotation services\n * - Web Workers for parallel processing\n *\n * For now, this function immediately resolves with the same results as\n * the synchronous version.\n *\n * @param text - Raw text to extract citations from\n * @param options - Optional customization (cleaners, patterns, resolve)\n * @returns Promise resolving to array of citations (or ResolvedCitation[] if resolve: true)\n *\n * @example\n * ```typescript\n * const citations = await extractCitationsAsync(text, { resolve: true })\n * // Returns ResolvedCitation[] with resolution metadata\n * ```\n */\nexport async function extractCitationsAsync(text: string, options: ExtractOptions & { resolve: true }): Promise<ResolvedCitation[]>\nexport async function extractCitationsAsync(text: string, options?: ExtractOptions): Promise<Citation[]>\nexport async function extractCitationsAsync(\n\ttext: string,\n\toptions?: ExtractOptions,\n): Promise<Citation[] | ResolvedCitation[]> {\n\t// Async wrapper for future extensibility (e.g., async reporters-db lookup)\n\t// For MVP, wraps synchronous extractCitations\n\treturn extractCitations(text, options)\n}\n"],"mappings":"mEAKA,SAAgB,EAAe,EAA8C,CAC3E,OAAO,EAAS,OAAS,QACpB,EAAS,OAAS,WAClB,EAAS,OAAS,WAClB,EAAS,OAAS,WAClB,EAAS,OAAS,aAClB,EAAS,OAAS,kBAMzB,SAAgB,EAAoB,EAAmD,CACrF,OAAO,EAAS,OAAS,MACpB,EAAS,OAAS,SAClB,EAAS,OAAS,gBAMzB,SAAgB,EAAe,EAAkD,CAC/E,OAAO,EAAS,OAAS,OAO3B,SAAgB,EACd,EACA,EAC+B,CAC/B,OAAO,EAAS,OAAS,EAmB3B,SAAgB,EAAkB,EAAiB,CACjD,MAAU,MAAM,qBAAqB,IAAI,CC5C3C,SAAgB,EAAc,EAAsB,CACnD,OAAO,EAAK,QAAQ,WAAY,GAAG,CAUpC,SAAgB,EAAoB,EAAsB,CACzD,OAAO,EAAK,QAAQ,aAAc,IAAI,CAAC,QAAQ,SAAU,IAAI,CAU9D,SAAgB,EAAiB,EAAsB,CACtD,OAAO,EAAK,UAAU,OAAO,CAU9B,SAAgB,EAAe,EAAsB,CACpD,OAAO,EACL,QAAQ,kBAAmB,IAAI,CAC/B,QAAQ,kBAAmB,IAAI,CCXlC,SAAgB,EACf,EACA,EAA4C,CAC3C,EACA,EACA,EACA,EACA,CACiB,CAElB,IAAI,EAAc,EACd,EAAkB,IAAI,IACtB,EAAkB,IAAI,IAG1B,IAAK,IAAI,EAAI,EAAG,GAAK,EAAS,OAAQ,IACrC,EAAgB,IAAI,EAAG,EAAE,CACzB,EAAgB,IAAI,EAAG,EAAE,CAI1B,IAAK,IAAM,KAAW,EAAU,CAC/B,IAAM,EAAa,EACb,EAAY,EAAQ,EAAY,CAEtC,GAAI,IAAe,EAAW,CAE7B,GAAM,CAAE,qBAAoB,sBAAuB,EAClD,EACA,EACA,EACA,EACA,CAED,EAAkB,EAClB,EAAkB,EAClB,EAAc,GAShB,MAAO,CACN,QAAS,EACT,kBAP4C,CAC5C,kBACA,kBACA,CAKA,SAAU,EAAE,CACZ,CAeF,SAAS,EACR,EACA,EACA,EACA,EAIC,CACD,IAAM,EAAqB,IAAI,IACzB,EAAqB,IAAI,IAE3B,EAAY,EACZ,EAAW,EAGf,KAAO,GAAa,EAAW,QAAU,GAAY,EAAU,QAAQ,CAEtE,GAAI,GAAa,EAAW,QAAU,GAAY,EAAU,OAAQ,CACnE,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAU,EAAY,CAC7C,EAAmB,IAAI,EAAa,EAAS,CAC7C,MAID,GAAI,GAAa,EAAW,OAAQ,CACnC,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAU,EAAY,CAC7C,IACA,SAID,GAAI,GAAY,EAAU,OAAQ,CACjC,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAa,EAAS,CAC7C,IACA,SAID,GAAI,EAAW,KAAe,EAAU,GAAW,CAClD,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAU,EAAY,CAC7C,EAAmB,IAAI,EAAa,EAAS,CAC7C,IACA,QACM,CAGN,IAAI,EAAa,GAIjB,IAAK,IAAI,EAAY,EAAG,GAAa,IAChC,IAAY,GAAa,EAAW,QADU,IAGlD,GAAI,EAAW,EAAY,KAAe,EAAU,GAAW,CAE9D,IAAK,IAAI,EAAI,EAAG,EAAI,EAAW,IAAK,CACnC,IAAM,EACL,EAAmB,IAAI,EAAY,EAAE,EAAI,EAAY,EACtD,EAAmB,IAAI,EAAa,EAAS,CAE9C,GAAa,EACb,EAAa,GACb,MAIF,GAAI,EAAY,SAGhB,IAAK,IAAI,EAAY,EAAG,GAAa,IAChC,IAAW,GAAa,EAAU,QADY,IAGlD,GAAI,EAAW,KAAe,EAAU,EAAW,GAAY,CAE9D,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,IAAK,IAAI,EAAI,EAAG,EAAI,EAAW,IAC9B,EAAmB,IAAI,EAAW,EAAG,EAAY,CAElD,GAAY,EACZ,EAAa,GACb,MAIF,GAAI,EAAY,SAGhB,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAU,EAAY,CAC7C,EAAmB,IAAI,EAAa,EAAS,CAC7C,IACA,KAIF,MAAO,CAAE,qBAAoB,qBAAoB,CCpLlD,MAAa,EAA0B,CACrC,CACE,GAAI,mBACJ,MAAO,0FACP,YAAa,mDACb,KAAM,OACP,CACD,CACE,GAAI,gBACJ,MAAO,gEACP,YAAa,+BACb,KAAM,OACP,CACD,CACE,GAAI,iBACJ,MAAO,0DACP,YAAa,6EACb,KAAM,OACP,CACF,CC3BY,EAA6B,CACxC,CACE,GAAI,MACJ,MAAO,uCACP,YAAa,iDACb,KAAM,UACP,CACD,CACE,GAAI,aACJ,MAAO,yDACP,YAAa,sEACb,KAAM,UACP,CACF,CCbY,EAA6B,CACxC,CACE,GAAI,aACJ,MAAO,4CACP,YAAa,iGACb,KAAM,UACP,CACF,CCPY,EAA6B,CACxC,CACE,GAAI,UACJ,MAAO,4BACP,YAAa,6CACb,KAAM,UACP,CACD,CACE,GAAI,QACJ,MAAO,wCACP,YAAa,sDACb,KAAM,UACP,CACD,CACE,GAAI,aACJ,MAAO,sCACP,YAAa,oDACb,KAAM,YACP,CACD,CACE,GAAI,mBACJ,MAAO,qCACP,YAAa,0DACb,KAAM,kBACP,CACF,CCIY,EAA+B,CAC1C,CACE,GAAI,KACJ,MAhC8B,+BAiC9B,YAAa,8CACb,KAAM,OACP,CACD,CACE,GAAI,OACJ,MAnCgC,iCAoChC,YAAa,oDACb,KAAM,OACP,CACD,CACE,GAAI,QACJ,MAjCiC,yFAkCjC,YAAa,mEACb,KAAM,OACP,CACD,CACE,GAAI,gBACJ,MAhC2C,6DAiC3C,YAAa,sDACb,KAAM,OACP,CACF,CCcD,SAAgB,EACd,EACA,EAAsB,CACpB,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACJ,CACQ,CACT,IAAM,EAAkB,EAAE,CAE1B,IAAK,IAAM,KAAW,EACpB,GAAI,CAEF,IAAM,EAAU,EAAY,SAAS,EAAQ,MAAM,CAEnD,IAAK,IAAM,KAAS,EAElB,EAAO,KAAK,CACV,KAAM,EAAM,GACZ,KAAM,CACJ,WAAY,EAAM,MAClB,SAAU,EAAM,MAAS,EAAM,GAAG,OACnC,CACD,KAAM,EAAQ,KACd,UAAW,EAAQ,GACpB,CAAC,OAEG,EAAO,CAEd,QAAQ,KACN,WAAW,EAAQ,GAAG,yBACtB,aAAiB,MAAQ,EAAM,QAAU,OAAO,EAAM,CACvD,CACD,SAOJ,OAFA,EAAO,MAAM,EAAG,IAAM,EAAE,KAAK,WAAa,EAAE,KAAK,WAAW,CAErD,ECtDT,SAAgB,EACf,EACA,EACmB,CACnB,GAAM,CAAE,OAAM,QAAS,EAMjB,EAD0B,qCACM,KAAK,EAAK,CAEhD,GAAI,CAAC,EAEJ,MAAU,MAAM,kCAAkC,IAAO,CAG1D,IAAM,EAAS,OAAO,SAAS,EAAM,GAAI,GAAG,CACtC,EAAW,EAAM,GAAG,MAAM,CAC1B,EAAO,OAAO,SAAS,EAAM,GAAI,GAAG,CAKpC,EADe,YACa,KAAK,EAAK,CACtC,EAAU,EAAe,OAAO,SAAS,EAAa,GAAI,GAAG,CAAG,IAAA,GAKhE,EADY,0BACU,KAAK,EAAK,CAChC,EAAO,EAAY,OAAO,SAAS,EAAU,GAAI,GAAG,CAAG,IAAA,GAKvD,EADa,2BACW,KAAK,EAAK,CAClC,EAAQ,EAAa,EAAW,GAAG,MAAM,CAAG,IAAA,GAG5C,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAG1D,EAAa,GA+CjB,MA5CwB,4JA2BvB,CAEmB,KAAM,GAAM,EAAS,SAAS,EAAE,CAAC,GACpD,GAAc,IAIX,IAAS,IAAA,IAER,GADgB,IAAI,MAAM,CAAC,aAAa,GAE3C,GAAc,IAKhB,EAAa,KAAK,IAAI,EAAY,EAAI,CAE/B,CACN,KAAM,OACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,aACA,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,SACA,WACA,OACA,UACA,QACA,OACA,CCtIF,SAAgB,EACf,EACA,EACkB,CAClB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADe,0DACM,KAAK,EAAK,CAErC,GAAI,CAAC,EACJ,MAAU,MAAM,qCAAqC,IAAO,CAG7D,IAAM,EAAQ,EAAM,GAAK,OAAO,SAAS,EAAM,GAAI,GAAG,CAAG,IAAA,GACnD,EAAO,EAAM,GAAG,MAAM,CACtB,EAAU,EAAM,GAGhB,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAG1D,EAAa,GAkBjB,MAfmB,CAClB,SACA,SACA,iBACA,kBACA,0BACA,8BACA,CAEc,KAAM,GAAM,EAAK,SAAS,EAAE,CAAC,GAC3C,GAAc,IAGf,EAAa,KAAK,IAAI,EAAY,EAAI,CAE/B,CACN,KAAM,UACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,aACA,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,QACA,OACA,UACA,CC1DF,SAAgB,EACf,EACA,EACkB,CAClB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADe,mCACM,KAAK,EAAK,CAErC,GAAI,CAAC,EACJ,MAAU,MAAM,qCAAqC,IAAO,CAG7D,IAAM,EAAS,OAAO,SAAS,EAAM,GAAI,GAAG,CACtC,EAAU,EAAM,GAAG,MAAM,CACzB,EAAO,OAAO,SAAS,EAAM,GAAI,GAAG,CAIpC,EADe,YACa,KAAK,EAAK,CACtC,EAAU,EAAe,OAAO,SAAS,EAAa,GAAI,GAAG,CAAG,IAAA,GAGhE,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,UACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,SACA,UACA,aAAc,EACd,OACA,UACA,CCvDF,SAAgB,EACf,EACA,EACkB,CAClB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADe,+CACM,KAAK,EAAK,CAErC,GAAI,CAAC,EACJ,MAAU,MAAM,qCAAqC,IAAO,CAG7D,IAAM,EAAO,OAAO,SAAS,EAAM,GAAI,GAAG,CACpC,EAAQ,EAAM,GACd,EAAiB,EAAM,GAGvB,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,UACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,EAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,OACA,QACA,iBACA,CC3CF,SAAgB,EACf,EACA,EACoB,CACpB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADiB,wCACM,KAAK,EAAK,CAEvC,GAAI,CAAC,EACJ,MAAU,MAAM,wCAAwC,IAAO,CAGhE,IAAM,EAAW,OAAO,SAAS,EAAM,GAAI,GAAG,CACxC,EAAY,OAAO,SAAS,EAAM,GAAI,GAAG,CAGzC,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,YACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,WACA,YACA,CC5CF,SAAgB,EACf,EACA,EAC0B,CAC1B,GAAM,CAAE,OAAM,QAAS,EAKjB,EADuB,iCACM,KAAK,EAAK,CAE7C,GAAI,CAAC,EACJ,MAAU,MAAM,8CAA8C,IAAO,CAGtE,IAAM,EAAS,OAAO,SAAS,EAAM,GAAI,GAAG,CACtC,EAAO,OAAO,SAAS,EAAM,GAAI,GAAG,CAKpC,EADY,wBACU,KAAK,EAAK,CAChC,EAAO,EAAY,OAAO,SAAS,EAAU,GAAI,GAAG,CAAG,IAAA,GAGvD,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,kBACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,SACA,OACA,OACA,CCpDF,SAAgB,EACf,EACA,EACa,CACb,GAAM,CAAE,OAAM,QAAS,EAKjB,EADU,oCACM,KAAK,EAAK,CAEhC,GAAI,CAAC,EACJ,MAAU,MAAM,iCAAiC,IAAO,CAIzD,IAAM,EAAU,EAAM,GAAK,OAAO,SAAS,EAAM,GAAI,GAAG,CAAG,IAAA,GAGrD,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,KACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,EAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,UACA,CAmCF,SAAgB,EACf,EACA,EACgB,CAChB,GAAM,CAAE,OAAM,QAAS,EAMjB,EADa,wFACM,KAAK,EAAK,CAEnC,GAAI,CAAC,EACJ,MAAU,MAAM,mCAAmC,IAAO,CAG3D,IAAM,EAAY,EAAM,GAClB,EAAU,EAAM,GAAK,OAAO,SAAS,EAAM,GAAI,GAAG,CAAG,IAAA,GAGrD,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,QACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,YACA,UACA,CAqCF,SAAgB,EACf,EACA,EACwB,CACxB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADiB,wDACM,KAAK,EAAK,CAEvC,GAAI,CAAC,EACJ,MAAU,MAAM,6CAA6C,IAAO,CAGrE,IAAM,EAAS,OAAO,SAAS,EAAM,GAAI,GAAG,CACtC,EAAW,EAAM,GAAG,MAAM,CAC1B,EAAU,OAAO,SAAS,EAAM,GAAI,GAAG,CAGvC,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,gBACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,SACA,WACA,UACA,CCjOF,SAAgB,EACd,EACA,EACA,EAA0B,SACL,CACrB,IAAM,EAAe,IAAI,IAGnB,EAAuB,CAAC,EAAE,CAC5B,EAEJ,MAAQ,EAAQ,EAAgB,KAAK,EAAK,IAAM,MAE9C,EAAW,KAAK,EAAM,MAAQ,EAAM,GAAG,OAAO,CAGhD,EAAW,KAAK,EAAK,OAAO,CAG5B,IAAK,IAAI,EAAI,EAAG,EAAI,EAAU,OAAQ,IAAK,CAEzC,IAAM,EADW,EAAU,GACI,KAAK,cAGhC,EAAe,EACnB,IAAK,IAAI,EAAI,EAAG,EAAI,EAAW,OAAS,EAAG,IACzC,GAAI,GAAiB,EAAW,IAAM,EAAgB,EAAW,EAAI,GAAI,CACvE,EAAe,EACf,MAIJ,EAAa,IAAI,EAAG,EAAa,CAGnC,OAAO,EAYT,SAAgB,EACd,EACA,EACA,EACA,EACS,CACT,GAAI,IAAa,OAEf,MAAO,GAIT,IAAM,EAAsB,EAAa,IAAI,EAAgB,CACvD,EAAmB,EAAa,IAAI,EAAa,CASvD,OANI,IAAwB,IAAA,IAAa,IAAqB,IAAA,GACrD,GAKF,IAAwB,ECpEjC,SAAgB,EAAoB,EAAW,EAAmB,CAEhE,GAAI,EAAE,SAAW,EAAG,OAAO,EAAE,OAC7B,GAAI,EAAE,SAAW,EAAG,OAAO,EAAE,OAI7B,IAAM,EAAiB,MAAM,KAAK,CAAE,OAAQ,EAAE,OAAS,EAAG,KACxD,MAAM,EAAE,OAAS,EAAE,CAAC,KAAK,EAAE,CAC5B,CAGD,IAAK,IAAI,EAAI,EAAG,GAAK,EAAE,OAAQ,IAC7B,EAAG,GAAG,GAAK,EAEb,IAAK,IAAI,EAAI,EAAG,GAAK,EAAE,OAAQ,IAC7B,EAAG,GAAG,GAAK,EAIb,IAAK,IAAI,EAAI,EAAG,GAAK,EAAE,OAAQ,IAC7B,IAAK,IAAI,EAAI,EAAG,GAAK,EAAE,OAAQ,IACzB,EAAE,EAAI,KAAO,EAAE,EAAI,GAErB,EAAG,GAAG,GAAK,EAAG,EAAI,GAAG,EAAI,GAMzB,EAAG,GAAG,GAAK,EAAI,KAAK,IAClB,EAAG,EAAI,GAAG,GACV,EAAG,GAAG,EAAI,GACV,EAAG,EAAI,GAAG,EAAI,GACf,CAKP,OAAO,EAAG,EAAE,QAAQ,EAAE,QAgBxB,SAAgB,EAA8B,EAAW,EAAmB,CAE1E,IAAM,EAAS,EAAE,aAAa,CACxB,EAAS,EAAE,aAAa,CAGxB,EAAW,EAAoB,EAAQ,EAAO,CAG9C,EAAY,KAAK,IAAI,EAAO,OAAQ,EAAO,OAAO,CAIxD,OAHI,IAAc,EAAU,EAGrB,EAAI,EAAW,ECtDxB,IAAa,EAAb,KAA8B,CAa5B,YACE,EACA,EACA,EAA6B,EAAE,CAC/B,CACA,KAAK,UAAY,EACjB,KAAK,KAAO,EAGZ,KAAK,QAAU,CACb,cAAe,EAAQ,eAAiB,YACxC,qBAAsB,EAAQ,sBAAwB,GACtD,yBAA0B,EAAQ,0BAA4B,SAC9D,mBAAoB,EAAQ,oBAAsB,GAClD,oBAAqB,EAAQ,qBAAuB,GACpD,sBAAuB,EAAQ,uBAAyB,GACxD,iBAAkB,EAAQ,kBAAoB,GAC/C,CAGD,KAAK,QAAU,CACb,cAAe,EACf,aAAc,EACd,iBAAkB,IAAA,GAClB,oBAAqB,IAAI,IACzB,aAAc,IAAI,IACnB,CAGG,KAAK,QAAQ,uBACf,KAAK,QAAQ,aAAe,EAC1B,EACA,EACA,KAAK,QAAQ,yBACd,EASL,SAA8B,CAC5B,IAAM,EAA+B,EAAE,CAEvC,IAAK,IAAI,EAAI,EAAG,EAAI,KAAK,UAAU,OAAQ,IAAK,CAC9C,KAAK,QAAQ,cAAgB,EAC7B,IAAM,EAAW,KAAK,UAAU,GAG5B,EAEJ,OAAQ,EAAS,KAAjB,CACE,IAAK,KACH,EAAa,KAAK,UAAU,EAAS,CACrC,MACF,IAAK,QACH,EAAa,KAAK,aAAa,EAAS,CACxC,MACF,IAAK,gBACH,EAAa,KAAK,qBAAqB,EAAS,CAChD,MACF,QAEM,EAAe,EAAS,GAC1B,KAAK,QAAQ,iBAAmB,EAChC,KAAK,kBAAkB,EAAU,EAAE,EAErC,MAKJ,EAAS,KAAK,CACZ,GAAG,EACH,aACD,CAAqB,CAGxB,OAAO,EAMT,UAAkB,EAAoD,CACpE,IAAM,EAAe,KAAK,QAAQ,cAG9B,EACJ,IAAK,IAAI,EAAI,EAAe,EAAG,GAAK,EAAG,IAErC,GADkB,KAAK,UAAU,GACnB,OAAS,OAAQ,CAC7B,EAAkB,EAClB,MAcJ,OATI,IAAoB,IAAA,GACf,KAAK,oBAAoB,wCAAwC,CAIrE,KAAK,cAAc,EAAiB,EAAa,CAI/C,CACL,WAAY,EACZ,WAAY,EACb,CANQ,KAAK,oBAAoB,6CAA6C,CAYjF,aAAqB,EAAuD,CAC1E,IAAM,EAAe,KAAK,QAAQ,cAC5B,EAAkB,KAAK,mBAAmB,EAAS,UAAU,CAG/D,EAEJ,IAAK,GAAM,CAAC,EAAW,KAAkB,KAAK,QAAQ,oBAAqB,CAEzE,GAAI,CAAC,KAAK,cAAc,EAAe,EAAa,CAClD,SAIF,IAAM,EAAa,EAA8B,EAAiB,EAAU,EAGxE,CAAC,GAAa,EAAa,EAAU,cACvC,EAAY,CAAE,MAAO,EAAe,aAAY,EAKpD,GAAI,CAAC,EACH,OAAO,KAAK,oBAAoB,kCAAkC,CAGpE,GAAI,EAAU,WAAa,KAAK,QAAQ,oBACtC,OAAO,KAAK,oBACV,yBAAyB,EAAU,WAAW,QAAQ,EAAE,CAAC,mBAAmB,KAAK,QAAQ,sBAC1F,CAIH,IAAM,EAAqB,EAAE,CAK7B,OAJI,EAAU,WAAa,GACzB,EAAS,KAAK,2BAA2B,EAAU,WAAW,QAAQ,EAAE,GAAG,CAGtE,CACL,WAAY,EAAU,MACtB,WAAY,EAAU,WACtB,SAAU,EAAS,OAAS,EAAI,EAAW,IAAA,GAC5C,CAMH,qBAA6B,EAA+D,CAC1F,IAAM,EAAe,KAAK,QAAQ,cAGlC,IAAK,IAAI,EAAI,EAAe,EAAG,GAAK,EAAG,IAAK,CAC1C,IAAM,EAAY,KAAK,UAAU,GAG7B,KAAU,OAAS,QAMrB,EAAU,SAAW,EAAS,QAC9B,KAAK,kBAAkB,EAAU,SAAS,GAAK,KAAK,kBAAkB,EAAS,SAAS,CAQxF,OALK,KAAK,cAAc,EAAG,EAAa,CAKjC,CACL,WAAY,EACZ,WAAY,IACb,CAPQ,KAAK,oBAAoB,2CAA2C,CAWjF,OAAO,KAAK,oBAAoB,uCAAuC,CAOzE,kBAA0B,EAAoB,EAAqB,CAEjE,GAAI,EAAS,OAAS,OAAQ,CAC5B,IAAM,EAAY,KAAK,iBAAiB,EAAS,CACjD,GAAI,EAAW,CACb,IAAM,EAAa,KAAK,mBAAmB,EAAU,CACrD,KAAK,QAAQ,oBAAoB,IAAI,EAAY,EAAM,GAS7D,iBAAyB,EAAgD,CAKvE,IAAM,EAAgB,EAAS,KAAK,cAE9B,EAAgB,KAAK,IAAI,EAAG,EAAgB,IAAI,CAChD,EAAa,KAAK,KAAK,UAAU,EAAe,EAAc,CAI9D,EAAS,EAAW,MAAM,4FAA4F,CAO5H,OANI,EACK,EAAO,GAAG,MAAM,CAIL,EAAW,MAAM,8CAA8C,GAC9D,GAAG,MAAM,CAMhC,mBAA2B,EAAsB,CAC/C,OAAO,EACJ,aAAa,CACb,QAAQ,OAAQ,IAAI,CACpB,MAAM,CAMX,kBAA0B,EAA0B,CAClD,OAAO,EACJ,aAAa,CACb,QAAQ,OAAQ,GAAG,CACnB,QAAQ,MAAO,GAAG,CAMvB,cAAsB,EAAyB,EAA+B,CAC5E,OAAO,EACL,EACA,EACA,KAAK,QAAQ,aACb,KAAK,QAAQ,cACd,CAMH,oBAA4B,EAA8C,CACxE,GAAI,KAAK,QAAQ,iBACf,MAAO,CACL,WAAY,IAAA,GACZ,cAAe,EACf,WAAY,EACb,GCvSP,SAAgB,EACd,EACA,EACA,EACoB,CAEpB,OADiB,IAAI,EAAiB,EAAW,EAAM,EAAQ,CAC/C,SAAS,CC6H3B,SAAgB,EACf,EACA,EACkC,CAClC,IAAM,EAAY,YAAY,KAAK,CAG7B,CAAE,UAAS,oBAAmB,YAAa,EAChD,EACA,GAAS,SACT,CAWK,EAAS,EAAS,EAPJ,GAAS,UAAY,CACxC,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACH,CAC4C,CAKvC,EAAoC,EAAE,CACtC,EAAgB,IAAI,IAE1B,IAAK,IAAM,KAAS,EAAQ,CAC3B,IAAM,EAAS,GAAG,EAAM,KAAK,WAAW,GAAG,EAAM,KAAK,WACjD,EAAc,IAAI,EAAO,GAC7B,EAAc,IAAI,EAAO,CACzB,EAAmB,KAAK,EAAM,EAKhC,IAAM,EAAwB,EAAE,CAChC,IAAK,IAAM,KAAS,EAAoB,CACvC,IAAI,EAEJ,OAAQ,EAAM,KAAd,CACC,IAAK,OAEJ,AAOC,EAPG,EAAM,YAAc,MAAQ,EAAM,YAAc,OACxC,EAAU,EAAO,EAAkB,CACpC,EAAM,YAAc,QACnB,EAAa,EAAO,EAAkB,CACvC,EAAM,YAAc,gBACnB,EAAqB,EAAO,EAAkB,CAE9C,EAAY,EAAO,EAAkB,CAEjD,MACD,IAAK,UACJ,EAAW,EAAe,EAAO,EAAkB,CACnD,MACD,IAAK,UACJ,EAAW,EAAe,EAAO,EAAkB,CACnD,MACD,IAAK,UACJ,EAAW,EAAe,EAAO,EAAkB,CACnD,MACD,IAAK,YACJ,EAAW,EAAiB,EAAO,EAAkB,CACrD,MACD,IAAK,kBACJ,EAAW,EAAuB,EAAO,EAAkB,CAC3D,MACD,QAEC,SAIE,EAAS,OAAS,IACrB,EAAS,SAAW,CAAC,GAAI,EAAS,UAAY,EAAE,CAAG,GAAG,EAAS,EAIhE,EAAS,cAAgB,YAAY,KAAK,CAAG,EAE7C,EAAU,KAAK,EAAS,CAQzB,OAJI,GAAS,QACL,EAAiB,EAAW,EAAM,EAAQ,kBAAkB,CAG7D,EA2BR,eAAsB,EACrB,EACA,EAC2C,CAG3C,OAAO,EAAiB,EAAM,EAAQ"}
+{"version":3,"file":"index.cjs","names":[],"sources":["../src/types/guards.ts","../src/clean/cleaners.ts","../src/clean/cleanText.ts","../src/patterns/casePatterns.ts","../src/patterns/statutePatterns.ts","../src/patterns/journalPatterns.ts","../src/patterns/neutralPatterns.ts","../src/patterns/shortForm.ts","../src/tokenize/tokenizer.ts","../src/extract/dates.ts","../src/extract/extractCase.ts","../src/extract/extractStatute.ts","../src/extract/extractJournal.ts","../src/extract/extractNeutral.ts","../src/extract/extractPublicLaw.ts","../src/extract/extractFederalRegister.ts","../src/extract/extractStatutesAtLarge.ts","../src/extract/extractShortForms.ts","../src/resolve/scopeBoundary.ts","../src/resolve/levenshtein.ts","../src/resolve/DocumentResolver.ts","../src/resolve/index.ts","../src/extract/detectParallel.ts","../src/extract/extractCitations.ts"],"sourcesContent":["import type { Citation, CitationType, CitationOfType, FullCitation, ShortFormCitation, FullCaseCitation } from \"./citation\"\n\n/**\n * Type guard: narrows Citation to a full citation (case, statute, journal, neutral, publicLaw, federalRegister).\n */\nexport function isFullCitation(citation: Citation): citation is FullCitation {\n  return citation.type === 'case'\n    || citation.type === 'statute'\n    || citation.type === 'journal'\n    || citation.type === 'neutral'\n    || citation.type === 'publicLaw'\n    || citation.type === 'federalRegister'\n    || citation.type === 'statutesAtLarge'\n}\n\n/**\n * Type guard: narrows Citation to a short-form citation (id, supra, shortFormCase).\n */\nexport function isShortFormCitation(citation: Citation): citation is ShortFormCitation {\n  return citation.type === 'id'\n    || citation.type === 'supra'\n    || citation.type === 'shortFormCase'\n}\n\n/**\n * Type guard: narrows Citation to a full case citation.\n */\nexport function isCaseCitation(citation: Citation): citation is FullCaseCitation {\n  return citation.type === 'case'\n}\n\n/**\n * Generic type guard that narrows a Citation to a specific type.\n * Useful when the target type is dynamic or generic.\n */\nexport function isCitationType<T extends CitationType>(\n  citation: Citation,\n  type: T\n): citation is CitationOfType<T> {\n  return citation.type === type\n}\n\n/**\n * Exhaustiveness helper for switch statements on discriminated unions.\n *\n * Place in the `default` branch to get a compile-time error if a new\n * variant is added but not handled.\n *\n * @example\n * ```typescript\n * switch (citation.type) {\n *   case 'case': ...\n *   case 'statute': ...\n *   // If you forget a variant, TypeScript errors here:\n *   default: assertUnreachable(citation.type)\n * }\n * ```\n */\nexport function assertUnreachable(x: never): never {\n  throw new Error(`Unexpected value: ${x}`)\n}\n","/**\n * Built-in text cleaner functions for preprocessing legal documents.\n *\n * Each cleaner is a simple transformation: (text: string) => string\n * Cleaners can be composed via the cleanText() pipeline.\n */\n\n/**\n * Remove all HTML tags from text.\n *\n * @example\n * stripHtmlTags(\"Smith v. <b>Doe</b>, 500 F.2d 123\")\n * // => \"Smith v. Doe, 500 F.2d 123\"\n */\nexport function stripHtmlTags(text: string): string {\n\treturn text.replace(/<[^>]+>/g, \"\")\n}\n\n/**\n * Normalize whitespace: convert tabs/newlines to spaces, collapse multiple spaces.\n *\n * @example\n * normalizeWhitespace(\"Smith  v.  Doe,   500 F.2d 123\")\n * // => \"Smith v. Doe, 500 F.2d 123\"\n */\nexport function normalizeWhitespace(text: string): string {\n\treturn text.replace(/[\\t\\n\\r]+/g, \" \").replace(/ {2,}/g, \" \")\n}\n\n/**\n * Apply Unicode NFKC normalization (ligatures → separate chars).\n *\n * @example\n * normalizeUnicode(\"Smith v. Doe, 500 F.2d 123\") // with ligature \"fi\"\n * // => \"Smith v. Doe, 500 F.2d 123\" // normalized\n */\nexport function normalizeUnicode(text: string): string {\n\treturn text.normalize(\"NFKC\")\n}\n\n/**\n * Replace curly quotes and apostrophes with straight quotes.\n *\n * @example\n * fixSmartQuotes(\"\"Smith\" v. 'Doe', 500 F.2d 123\")\n * // => \"\\\"Smith\\\" v. 'Doe', 500 F.2d 123\"\n */\nexport function fixSmartQuotes(text: string): string {\n\treturn text\n\t\t.replace(/[\\u201C\\u201D]/g, '\"') // curly double quotes\n\t\t.replace(/[\\u2018\\u2019]/g, \"'\") // curly single quotes/apostrophes\n}\n\n/**\n * Remove underscore OCR artifacts (common in scanned documents).\n *\n * @example\n * removeOcrArtifacts(\"Smith v. Doe, 500 F._2d 123\")\n * // => \"Smith v. Doe, 500 F.2d 123\"\n */\nexport function removeOcrArtifacts(text: string): string {\n\treturn text.replace(/_/g, \"\")\n}\n","import type { TransformationMap } from \"../types/span\"\nimport type { Warning } from \"../types/citation\"\nimport {\n\tfixSmartQuotes,\n\tnormalizeUnicode,\n\tnormalizeWhitespace,\n\tstripHtmlTags,\n} from \"./cleaners\"\n\n/**\n * Result of text cleaning operation.\n */\nexport interface CleanTextResult {\n\t/** Cleaned text after all transformations */\n\tcleaned: string\n\n\t/** Position mappings between cleaned and original text */\n\ttransformationMap: TransformationMap\n\n\t/** Warnings generated during cleaning (currently unused) */\n\twarnings: Warning[]\n}\n\n/**\n * Clean text using a pipeline of transformation functions.\n *\n * Applies cleaners sequentially while maintaining accurate position mappings\n * between the original and cleaned text. This enables citation extraction from\n * cleaned text while reporting positions in the original text.\n *\n * @param original - Original input text\n * @param cleaners - Array of cleaner functions to apply (default: stripHtmlTags, normalizeWhitespace, normalizeUnicode, fixSmartQuotes)\n * @returns Cleaned text with position mappings and warnings\n *\n * @example\n * const result = cleanText(\"Smith v. <b>Doe</b>, 500 F.2d 123\")\n * // result.cleaned: \"Smith v. Doe, 500 F.2d 123\"\n * // result.transformationMap tracks position shifts from HTML removal\n */\nexport function cleanText(\n\toriginal: string,\n\tcleaners: Array<(text: string) => string> = [\n\t\tstripHtmlTags,\n\t\tnormalizeWhitespace,\n\t\tnormalizeUnicode,\n\t\tfixSmartQuotes,\n\t],\n): CleanTextResult {\n\t// Initialize 1:1 position mapping\n\tlet currentText = original\n\tlet cleanToOriginal = new Map<number, number>()\n\tlet originalToClean = new Map<number, number>()\n\n\t// Identity mapping: cleanToOriginal[i] = i, originalToClean[i] = i\n\tfor (let i = 0; i <= original.length; i++) {\n\t\tcleanToOriginal.set(i, i)\n\t\toriginalToClean.set(i, i)\n\t}\n\n\t// Apply each cleaner sequentially, rebuilding position maps\n\tfor (const cleaner of cleaners) {\n\t\tconst beforeText = currentText\n\t\tconst afterText = cleaner(currentText)\n\n\t\tif (beforeText !== afterText) {\n\t\t\t// Text changed - rebuild position maps\n\t\t\tconst { newCleanToOriginal, newOriginalToClean } = rebuildPositionMaps(\n\t\t\t\tbeforeText,\n\t\t\t\tafterText,\n\t\t\t\tcleanToOriginal,\n\t\t\t\toriginalToClean,\n\t\t\t)\n\n\t\t\tcleanToOriginal = newCleanToOriginal\n\t\t\toriginalToClean = newOriginalToClean\n\t\t\tcurrentText = afterText\n\t\t}\n\t}\n\n\tconst transformationMap: TransformationMap = {\n\t\tcleanToOriginal,\n\t\toriginalToClean,\n\t}\n\n\treturn {\n\t\tcleaned: currentText,\n\t\ttransformationMap,\n\t\twarnings: [],\n\t}\n}\n\n/**\n * Rebuild position maps after a text transformation.\n *\n * Uses a simplified algorithm that scans through both strings, matching\n * characters where possible and tracking the offset accumulation.\n *\n * @param beforeText - Text before transformation\n * @param afterText - Text after transformation\n * @param oldCleanToOriginal - Previous clean-to-original mapping\n * @param oldOriginalToClean - Previous original-to-clean mapping\n * @returns New position maps\n */\nfunction rebuildPositionMaps(\n\tbeforeText: string,\n\tafterText: string,\n\toldCleanToOriginal: Map<number, number>,\n\t_oldOriginalToClean: Map<number, number>,\n): {\n\tnewCleanToOriginal: Map<number, number>\n\tnewOriginalToClean: Map<number, number>\n} {\n\tconst newCleanToOriginal = new Map<number, number>()\n\tconst newOriginalToClean = new Map<number, number>()\n\n\tlet beforeIdx = 0\n\tlet afterIdx = 0\n\n\t// Scan through both strings, matching characters where possible\n\twhile (beforeIdx <= beforeText.length || afterIdx <= afterText.length) {\n\t\t// Both at end\n\t\tif (beforeIdx >= beforeText.length && afterIdx >= afterText.length) {\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewCleanToOriginal.set(afterIdx, originalPos)\n\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\tbreak\n\t\t}\n\n\t\t// Before text exhausted (expansion case)\n\t\tif (beforeIdx >= beforeText.length) {\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewCleanToOriginal.set(afterIdx, originalPos)\n\t\t\tafterIdx++\n\t\t\tcontinue\n\t\t}\n\n\t\t// After text exhausted (removal case)\n\t\tif (afterIdx >= afterText.length) {\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\tbeforeIdx++\n\t\t\tcontinue\n\t\t}\n\n\t\t// Characters match - carry forward the mapping\n\t\tif (beforeText[beforeIdx] === afterText[afterIdx]) {\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewCleanToOriginal.set(afterIdx, originalPos)\n\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\tbeforeIdx++\n\t\t\tafterIdx++\n\t\t} else {\n\t\t\t// Characters differ - need to determine if this is insertion/deletion/replacement\n\t\t\t// Look ahead to find next match\n\t\t\tlet foundMatch = false\n\t\t\tconst maxLookAhead = 20 // Limit lookahead to avoid performance issues\n\n\t\t\t// Check if something was deleted from before text\n\t\t\tfor (let lookAhead = 1; lookAhead <= maxLookAhead; lookAhead++) {\n\t\t\t\tif (beforeIdx + lookAhead >= beforeText.length) break\n\n\t\t\t\tif (beforeText[beforeIdx + lookAhead] === afterText[afterIdx]) {\n\t\t\t\t\t// Found a match - characters were deleted from before text\n\t\t\t\t\tfor (let i = 0; i < lookAhead; i++) {\n\t\t\t\t\t\tconst originalPos =\n\t\t\t\t\t\t\toldCleanToOriginal.get(beforeIdx + i) ?? beforeIdx + i\n\t\t\t\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\t\t\t}\n\t\t\t\t\tbeforeIdx += lookAhead\n\t\t\t\t\tfoundMatch = true\n\t\t\t\t\tbreak\n\t\t\t\t}\n\t\t\t}\n\n\t\t\tif (foundMatch) continue\n\n\t\t\t// Check if something was inserted into after text\n\t\t\tfor (let lookAhead = 1; lookAhead <= maxLookAhead; lookAhead++) {\n\t\t\t\tif (afterIdx + lookAhead >= afterText.length) break\n\n\t\t\t\tif (beforeText[beforeIdx] === afterText[afterIdx + lookAhead]) {\n\t\t\t\t\t// Found a match - characters were inserted into after text\n\t\t\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\t\t\tfor (let i = 0; i < lookAhead; i++) {\n\t\t\t\t\t\tnewCleanToOriginal.set(afterIdx + i, originalPos)\n\t\t\t\t\t}\n\t\t\t\t\tafterIdx += lookAhead\n\t\t\t\t\tfoundMatch = true\n\t\t\t\t\tbreak\n\t\t\t\t}\n\t\t\t}\n\n\t\t\tif (foundMatch) continue\n\n\t\t\t// No match found within lookahead - treat as replacement\n\t\t\tconst originalPos = oldCleanToOriginal.get(beforeIdx) ?? beforeIdx\n\t\t\tnewCleanToOriginal.set(afterIdx, originalPos)\n\t\t\tnewOriginalToClean.set(originalPos, afterIdx)\n\t\t\tbeforeIdx++\n\t\t\tafterIdx++\n\t\t}\n\t}\n\n\treturn { newCleanToOriginal, newOriginalToClean }\n}\n","/**\n * Case Citation Regex Patterns\n *\n * These patterns are designed for tokenization (broad matching) not extraction.\n * They identify potential case citations in text for the tokenizer (Plan 3).\n * Metadata parsing and validation against reporters-db happens in Phase 2 Plan 5 (extraction layer).\n *\n * Pattern Design Principles (from RESEARCH.md):\n * - Use \\b word boundaries to avoid matching \"F.\" in \"F.B.I.\"\n * - Avoid nested quantifiers: (a+)+ causes ReDoS\n * - Keep patterns simple: tokenization only needs to find candidates\n * - Use global flag /g for matchAll()\n */\n\nimport type { FullCitationType } from '@/types/citation'\n\nexport interface Pattern {\n  id: string\n  regex: RegExp\n  description: string\n  type: FullCitationType\n}\n\nexport const casePatterns: Pattern[] = [\n  {\n    id: 'federal-reporter',\n    regex: /\\b(\\d+(?:-\\d+)?)\\s+(F\\.|F\\.2d|F\\.3d|F\\.4th|F\\.\\s?Supp\\.|F\\.\\s?Supp\\.\\s?2d|F\\.\\s?Supp\\.\\s?3d|F\\.\\s?Supp\\.\\s?4th)\\s+(\\d+|_{3,}|-{3,})(?=\\s|$|\\(|,|;|\\.)/g,\n    description: 'Federal Reporter (F., F.2d, F.3d, F.4th, F.Supp., etc.)',\n    type: 'case',\n  },\n  {\n    id: 'supreme-court',\n    regex: /\\b(\\d+(?:-\\d+)?)\\s+(U\\.\\s?S\\.|S\\.\\s?Ct\\.|L\\.\\s?Ed\\.(?:\\s?2d)?)\\s+(\\d+|_{3,}|-{3,})(?=\\s|$|\\(|,|;|\\.)/g,\n    description: 'U.S. Supreme Court reporters',\n    type: 'case',\n  },\n  {\n    id: 'state-reporter',\n    regex: /\\b(\\d+(?:-\\d+)?)\\s+([A-Z][A-Za-z.]+(?:\\s?2d|\\s?3d|\\s?4th|\\s?5th)?)\\s+(\\d+|_{3,}|-{3,})(?=\\s|$|\\(|,|;|\\.)/g,\n    description: 'State reporters (broad pattern, validated against reporters-db in Phase 3)',\n    type: 'case',\n  },\n]\n","/**\n * Statute Citation Regex Patterns\n *\n * Patterns for U.S. Code and state code citations.\n * These are intentionally broad for tokenization - validation against\n * actual code databases happens in Phase 2 Plan 5 (extraction layer).\n *\n * Pattern Design:\n * - Simple structure to avoid ReDoS\n * - Matches both \"§\" and \"Section\" formats\n * - State codes use broad pattern (validated later)\n */\n\nimport type { Pattern } from './casePatterns'\n\nexport const statutePatterns: Pattern[] = [\n  {\n    id: 'usc',\n    regex: /\\b(\\d+)\\s+U\\.S\\.C\\.?\\s+§+\\s*(\\d+[A-Za-z]*)\\b/g,\n    description: 'U.S. Code citations (e.g., \"42 U.S.C. § 1983\")',\n    type: 'statute',\n  },\n  {\n    id: 'state-code',\n    regex: /\\b([A-Z][a-z]+\\.?\\s+[A-Za-z.]+\\s+Code)\\s+§\\s*(\\d+[A-Za-z]*)\\b/g,\n    description: 'State code citations (broad pattern, e.g., \"Cal. Penal Code § 187\")',\n    type: 'statute',\n  },\n]\n","/**\n * Journal Citation Regex Patterns\n *\n * Patterns for law review and journal citations.\n * These are intentionally broad for tokenization - validation against\n * journals-db happens in Phase 3 (extraction layer).\n *\n * Pattern Design:\n * - Matches volume-journal-page format\n * - Broad journal name matching (validated later)\n * - Simple structure to avoid ReDoS\n */\n\nimport type { Pattern } from './casePatterns'\n\nexport const journalPatterns: Pattern[] = [\n  {\n    id: 'law-review',\n    regex: /\\b(\\d+(?:-\\d+)?)\\s+([A-Z][A-Za-z.\\s]+)\\s+(\\d+)\\b/g,\n    description: 'Law review citations (e.g., \"120 Harv. L. Rev. 500\"), validated against journals-db in Phase 3',\n    type: 'journal',\n  },\n]\n","/**\n * Neutral and Online Citation Regex Patterns\n *\n * Patterns for WestLaw, LexisNexis, public laws, and Federal Register citations.\n * These have predictable formats and don't require external validation.\n *\n * Pattern Design:\n * - Matches year-database-number format for online citations\n * - Matches Pub. L. No. format for public laws\n * - Matches volume-Fed. Reg.-page for Federal Register\n * - Simple structure to avoid ReDoS\n */\n\nimport type { Pattern } from './casePatterns'\n\nexport const neutralPatterns: Pattern[] = [\n  {\n    id: 'westlaw',\n    regex: /\\b(\\d{4})\\s+WL\\s+(\\d+)\\b/g,\n    description: 'WestLaw citations (e.g., \"2021 WL 123456\")',\n    type: 'neutral',\n  },\n  {\n    id: 'lexis',\n    regex: /\\b(\\d{4})\\s+U\\.S\\.\\s+LEXIS\\s+(\\d+)\\b/g,\n    description: 'LexisNexis citations (e.g., \"2021 U.S. LEXIS 5000\")',\n    type: 'neutral',\n  },\n  {\n    id: 'public-law',\n    regex: /\\bPub\\.\\s?L\\.\\s?No\\.\\s?(\\d+-\\d+)\\b/g,\n    description: 'Public Law citations (e.g., \"Pub. L. No. 117-58\")',\n    type: 'publicLaw',\n  },\n  {\n    id: 'federal-register',\n    regex: /\\b(\\d+(?:-\\d+)?)\\s+Fed\\.\\s?Reg\\.\\s+(\\d+)\\b/g,\n    description: 'Federal Register citations (e.g., \"86 Fed. Reg. 12345\")',\n    type: 'federalRegister',\n  },\n  {\n    id: 'statutes-at-large',\n    regex: /\\b(\\d+(?:-\\d+)?)\\s+Stat\\.\\s+(\\d+)\\b/g,\n    description: 'Statutes at Large citations (e.g., \"124 Stat. 119\")',\n    type: 'statutesAtLarge',\n  },\n  {\n    id: 'compact-law-review',\n    regex: /\\b(\\d+(?:-\\d+)?)\\s+([A-Z][A-Za-z.]+L\\.(?:Rev|J|Q)\\.)\\s+(\\d+)\\b/g,\n    description: 'Compact law review citations without spaces (e.g., \"93 Harv.L.Rev. 752\")',\n    type: 'journal',\n  },\n]\n","/**\n * Short-form Citation Regex Patterns\n *\n * Patterns for Id., Ibid., supra, and short-form case citations.\n * These refer to earlier citations in the document.\n *\n * Pattern Design:\n * - Simple structure to avoid ReDoS (no nested quantifiers)\n * - Broad matching for tokenization; validation happens in extraction layer\n * - Word boundaries to prevent false positives (e.g., \"Idaho\" vs \"Id.\")\n */\n\nimport type { Pattern } from './casePatterns'\n\n/** Id. with optional pincite: \"Id.\" or \"Id. at 253\" */\nexport const ID_PATTERN: RegExp = /\\b[Ii]d\\.(?:\\s+at\\s+(\\d+))?/g\n\n/** Ibid. with optional pincite (less common variant) */\nexport const IBID_PATTERN: RegExp = /\\b[Ii]bid\\.(?:\\s+at\\s+(\\d+))?/g\n\n/**\n * Supra with party name and optional pincite.\n * Pattern: word(s), supra [, at page]\n * Captures: (1) party name, (2) pincite\n * Note: Matches party names including \"v.\" (e.g., \"Smith v. Jones, supra\")\n */\nexport const SUPRA_PATTERN: RegExp = /\\b([A-Z][a-zA-Z]+(?:(?:\\s+v\\.?\\s+|\\s+)[A-Z][a-zA-Z]+)*)\\s*,?\\s+supra(?:,?\\s+at\\s+(\\d+))?/g\n\n/**\n * Short-form case: volume reporter at page\n * Pattern: number space abbreviation space \"at\" space number\n * Simplified detection; full parsing in extraction layer\n */\nexport const SHORT_FORM_CASE_PATTERN: RegExp = /\\b(\\d+(?:-\\d+)?)\\s+([A-Z][A-Za-z.\\s]+?(?:\\d[a-z])?)\\s+at\\s+(\\d+)\\b/g\n\n/** All short-form patterns for tokenization */\nexport const SHORT_FORM_PATTERNS: readonly RegExp[] = [\n  ID_PATTERN,\n  IBID_PATTERN,\n  SUPRA_PATTERN,\n  SHORT_FORM_CASE_PATTERN,\n] as const\n\n/** Pattern objects for consistency with other pattern modules */\nexport const shortFormPatterns: Pattern[] = [\n  {\n    id: 'id',\n    regex: ID_PATTERN,\n    description: 'Id. citations (e.g., \"Id.\" or \"Id. at 253\")',\n    type: 'case', // Will be typed as 'id' in extraction layer\n  },\n  {\n    id: 'ibid',\n    regex: IBID_PATTERN,\n    description: 'Ibid. citations (e.g., \"Ibid.\" or \"Ibid. at 125\")',\n    type: 'case', // Will be typed as 'id' in extraction layer\n  },\n  {\n    id: 'supra',\n    regex: SUPRA_PATTERN,\n    description: 'Supra citations (e.g., \"Smith, supra\" or \"Smith, supra, at 460\")',\n    type: 'case', // Will be typed as 'supra' in extraction layer\n  },\n  {\n    id: 'shortFormCase',\n    regex: SHORT_FORM_CASE_PATTERN,\n    description: 'Short-form case citations (e.g., \"500 F.2d at 125\")',\n    type: 'case', // Will be typed as 'shortFormCase' in extraction layer\n  },\n]\n","/**\n * Tokenization Layer for Citation Extraction\n *\n * Applies regex patterns to cleaned text to produce citation candidate tokens.\n * This is the second stage of the parsing pipeline:\n *   1. Clean text (remove HTML, normalize Unicode)\n *   2. Tokenize (apply patterns to find candidates) ← THIS MODULE\n *   3. Extract (parse metadata, validate against reporters-db)\n *\n * Tokenization is intentionally broad - it finds potential citations without\n * validating them. The extraction layer (Plan 5) validates tokens against\n * reporters-db and parses metadata.\n *\n * @module tokenize\n */\n\nimport type { Span } from '@/types/span'\nimport type { Pattern } from '@/patterns'\nimport {\n  casePatterns,\n  statutePatterns,\n  journalPatterns,\n  neutralPatterns,\n} from '@/patterns'\nimport { shortFormPatterns } from '@/patterns/shortForm'\n\n/**\n * A token representing a potential citation found in cleaned text.\n *\n * Tokens are produced by applying regex patterns to cleaned text.\n * They include matched text, position in cleaned text, and pattern metadata\n * for use in the extraction layer.\n */\nexport interface Token {\n  /** Matched text from input */\n  text: string\n\n  /** Position in cleaned text (cleanStart/cleanEnd only, no original positions yet) */\n  span: Pick<Span, 'cleanStart' | 'cleanEnd'>\n\n  /** Pattern type that matched this token */\n  type: Pattern['type']\n\n  /** Pattern ID that matched this token */\n  patternId: string\n}\n\n/**\n * Tokenizes cleaned text by applying regex patterns to find citation candidates.\n *\n * For each pattern in the patterns array:\n *   1. Apply pattern.regex.matchAll(cleanedText)\n *   2. Create Token for each match with position, text, and pattern metadata\n *   3. Collect all tokens from all patterns\n *   4. Sort by cleanStart position (ascending)\n *\n * Timeout protection: If a pattern throws (e.g., ReDoS), skip it and continue\n * with remaining patterns. Logs warning to console.\n *\n * Note: This function is synchronous because regex matching is inherently\n * synchronous. This enables both sync (extractCitations) and async\n * (extractCitationsAsync) APIs in Plan 6.\n *\n * @param cleanedText - Text that has been cleaned by cleanText() from Plan 1\n * @param patterns - Regex patterns to apply (defaults to all patterns from Plan 2)\n * @returns Array of tokens sorted by position (cleanStart ascending)\n *\n * @example\n * ```typescript\n * import { tokenize } from '@/tokenize'\n * import { cleanText } from '@/clean'\n *\n * const original = \"See Smith v. Doe, 500 F.2d 123 (9th Cir. 2020)\"\n * const { cleanedText } = cleanText(original)\n * const tokens = tokenize(cleanedText)\n * // tokens[0] = {\n * //   text: \"500 F.2d 123\",\n * //   span: { cleanStart: 18, cleanEnd: 30 },\n * //   type: \"case\",\n * //   patternId: \"federal-reporter\"\n * // }\n * ```\n */\nexport function tokenize(\n  cleanedText: string,\n  patterns: Pattern[] = [\n    ...casePatterns,\n    ...statutePatterns,\n    ...journalPatterns,\n    ...neutralPatterns,\n    ...shortFormPatterns,\n  ]\n): Token[] {\n  const tokens: Token[] = []\n\n  for (const pattern of patterns) {\n    try {\n      // Apply pattern to cleaned text\n      const matches = cleanedText.matchAll(pattern.regex)\n\n      for (const match of matches) {\n        // Create token from match\n        tokens.push({\n          text: match[0],\n          span: {\n            cleanStart: match.index!,\n            cleanEnd: match.index! + match[0].length,\n          },\n          type: pattern.type,\n          patternId: pattern.id,\n        })\n      }\n    } catch (error) {\n      // Timeout protection: If pattern throws (ReDoS, etc.), skip it\n      console.warn(\n        `Pattern ${pattern.id} threw error, skipping:`,\n        error instanceof Error ? error.message : String(error)\n      )\n    }\n  }\n\n  // Sort tokens by position (cleanStart ascending)\n  tokens.sort((a, b) => a.span.cleanStart - b.span.cleanStart)\n\n  return tokens\n}\n","/**\n * Date Parsing Utilities for Legal Citations\n *\n * Parses dates from parentheticals in legal citations. Supports three formats:\n * 1. Abbreviated month: \"Jan. 15, 2020\"\n * 2. Full month: \"January 15, 2020\"\n * 3. Numeric US: \"1/15/2020\"\n * 4. Year-only: \"2020\"\n *\n * @module extract/dates\n */\n\n/**\n * Structured date components.\n * Month and day are optional to support year-only dates.\n */\nexport interface ParsedDate {\n  year: number\n  month?: number\n  day?: number\n}\n\n/**\n * Date in both ISO string and structured format.\n */\nexport interface StructuredDate {\n  /** ISO 8601 format: YYYY-MM-DD, YYYY-MM, or YYYY */\n  iso: string\n  /** Structured date components */\n  parsed: ParsedDate\n}\n\n/**\n * Month name/abbreviation to numeric value (1-12).\n * Includes both 3-letter and 4-letter (Sept) abbreviations.\n */\nconst MONTH_MAP: Record<string, number> = {\n  jan: 1,\n  january: 1,\n  feb: 2,\n  february: 2,\n  mar: 3,\n  march: 3,\n  apr: 4,\n  april: 4,\n  may: 5,\n  jun: 6,\n  june: 6,\n  jul: 7,\n  july: 7,\n  aug: 8,\n  august: 8,\n  sep: 9,\n  sept: 9,\n  september: 9,\n  oct: 10,\n  october: 10,\n  nov: 11,\n  november: 11,\n  dec: 12,\n  december: 12,\n}\n\n/**\n * Parse a month name or abbreviation to numeric value (1-12).\n *\n * @param monthStr - Month name or abbreviation (e.g., \"Jan\", \"January\", \"Sept.\")\n * @returns Numeric month (1-12)\n * @throws Error if month name is not recognized\n *\n * @example\n * ```typescript\n * parseMonth(\"Jan\") // 1\n * parseMonth(\"Sept.\") // 9\n * parseMonth(\"December\") // 12\n * ```\n */\nexport function parseMonth(monthStr: string): number {\n  // Normalize: lowercase, strip trailing period\n  const normalized = monthStr.toLowerCase().replace(/\\.$/, '')\n  const month = MONTH_MAP[normalized]\n\n  if (month === undefined) {\n    throw new Error(`Invalid month name: ${monthStr}`)\n  }\n\n  return month\n}\n\n/**\n * Convert structured date components to ISO 8601 string.\n * Handles full dates, month+year, and year-only formats.\n *\n * @param parsed - Structured date components\n * @returns ISO 8601 string (YYYY-MM-DD, YYYY-MM, or YYYY)\n *\n * @example\n * ```typescript\n * toIsoDate({ year: 2020, month: 1, day: 15 }) // \"2020-01-15\"\n * toIsoDate({ year: 2020, month: 1 }) // \"2020-01\"\n * toIsoDate({ year: 2020 }) // \"2020\"\n * ```\n */\nexport function toIsoDate(parsed: ParsedDate): string {\n  const { year, month, day } = parsed\n\n  if (month !== undefined && day !== undefined) {\n    // Full date: YYYY-MM-DD with zero-padding\n    const monthStr = String(month).padStart(2, '0')\n    const dayStr = String(day).padStart(2, '0')\n    return `${year}-${monthStr}-${dayStr}`\n  }\n\n  if (month !== undefined) {\n    // Month+year: YYYY-MM with zero-padding\n    const monthStr = String(month).padStart(2, '0')\n    return `${year}-${monthStr}`\n  }\n\n  // Year-only: YYYY\n  return String(year)\n}\n\n/**\n * Parse a date string into structured format.\n * Tries multiple formats in order:\n * 1. Abbreviated month (Jan. 15, 2020)\n * 2. Full month (January 15, 2020)\n * 3. Numeric US format (1/15/2020)\n * 4. Year-only (2020)\n *\n * @param dateStr - Date string in any supported format\n * @returns Structured date with ISO string, or undefined if no match\n *\n * @example\n * ```typescript\n * parseDate(\"Jan. 15, 2020\") // { iso: \"2020-01-15\", parsed: { year: 2020, month: 1, day: 15 } }\n * parseDate(\"January 15, 2020\") // { iso: \"2020-01-15\", parsed: { year: 2020, month: 1, day: 15 } }\n * parseDate(\"1/15/2020\") // { iso: \"2020-01-15\", parsed: { year: 2020, month: 1, day: 15 } }\n * parseDate(\"2020\") // { iso: \"2020\", parsed: { year: 2020 } }\n * parseDate(\"no date\") // undefined\n * ```\n */\nexport function parseDate(dateStr: string): StructuredDate | undefined {\n  // Try abbreviated month format: Jan. 15, 2020 or Feb 9, 2015\n  const abbrMatch = dateStr.match(/\\b(Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Sept|Oct|Nov|Dec)\\.?\\s+(\\d{1,2}),?\\s+(\\d{4})\\b/i)\n  if (abbrMatch) {\n    const month = parseMonth(abbrMatch[1])\n    const day = Number.parseInt(abbrMatch[2], 10)\n    const year = Number.parseInt(abbrMatch[3], 10)\n    const parsed = { year, month, day }\n    return { iso: toIsoDate(parsed), parsed }\n  }\n\n  // Try full month format: January 15, 2020\n  const fullMatch = dateStr.match(/\\b(January|February|March|April|May|June|July|August|September|October|November|December)\\s+(\\d{1,2}),?\\s+(\\d{4})\\b/i)\n  if (fullMatch) {\n    const month = parseMonth(fullMatch[1])\n    const day = Number.parseInt(fullMatch[2], 10)\n    const year = Number.parseInt(fullMatch[3], 10)\n    const parsed = { year, month, day }\n    return { iso: toIsoDate(parsed), parsed }\n  }\n\n  // Try numeric US format: 1/15/2020\n  const numericMatch = dateStr.match(/\\b(\\d{1,2})\\/(\\d{1,2})\\/(\\d{4})\\b/)\n  if (numericMatch) {\n    const month = Number.parseInt(numericMatch[1], 10)\n    const day = Number.parseInt(numericMatch[2], 10)\n    const year = Number.parseInt(numericMatch[3], 10)\n    const parsed = { year, month, day }\n    return { iso: toIsoDate(parsed), parsed }\n  }\n\n  // Try year-only: 2020\n  const yearMatch = dateStr.match(/\\b(\\d{4})\\b/)\n  if (yearMatch) {\n    const year = Number.parseInt(yearMatch[1], 10)\n    const parsed = { year }\n    return { iso: toIsoDate(parsed), parsed }\n  }\n\n  // No match\n  return undefined\n}\n","/**\n * Case Citation Extraction\n *\n * Parses tokenized case citations to extract volume, reporter, page, and\n * optional metadata (pincite, court, year). This is the third stage of\n * the parsing pipeline:\n *   1. Clean text (remove HTML, normalize Unicode)\n *   2. Tokenize (apply patterns to find candidates)\n *   3. Extract (parse metadata, validate) ← THIS MODULE\n *\n * Extraction parses structured data from token text. Validation against\n * reporters-db happens in Phase 3 (resolution layer).\n *\n * @module extract/extractCase\n */\n\nimport type { Token } from '@/tokenize'\nimport type { FullCaseCitation } from '@/types/citation'\nimport type { TransformationMap, Span } from '@/types/span'\nimport { parseDate, type StructuredDate } from './dates'\n\n/** Parse a volume string as number when purely numeric, string when hyphenated */\nfunction parseVolume(raw: string): number | string {\n\tconst num = Number.parseInt(raw, 10)\n\treturn String(num) === raw ? num : raw\n}\n\n/** Month abbreviations and full names found in legal citation parentheticals */\nconst MONTH_PATTERN = /(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Sept|Oct|Nov|Dec|January|February|March|April|May|June|July|August|September|October|November|December)\\.?/\n\n// ============================================================================\n// Compiled regex patterns for performance (hoisted to module level)\n// ============================================================================\n\n/** Matches volume-reporter-page format in citation core */\nconst VOLUME_REPORTER_PAGE_REGEX = /^(\\d+(?:-\\d+)?)\\s+([A-Za-z0-9.\\s]+)\\s+(\\d+|_{3,}|-{3,})/\n\n/** Detects blank page placeholders (3+ underscores or dashes) */\nconst BLANK_PAGE_REGEX = /^[_-]{3,}$/\n\n/** Extracts pincite (page reference after comma) */\nconst PINCITE_REGEX = /,\\s*(\\d+)/\n\n/** Matches parenthetical content */\nconst PAREN_REGEX = /\\(([^)]+)\\)/\n\n/** Look-ahead pattern for parenthetical after token */\nconst LOOKAHEAD_PAREN_REGEX = /^(?:,\\s*\\d+)*\\s*\\(([^)]+)\\)/\n\n/** Extracts pincite from look-ahead text */\nconst LOOKAHEAD_PINCITE_REGEX = /^,\\s*(\\d+)/\n\n/** Matches chained parentheticals with disposition */\nconst CHAINED_DISPOSITION_REGEX = /\\([^)]+\\)\\s*\\((en banc|per curiam)\\)/i\n\n/** Identifies Supreme Court reporters for court inference */\nconst SCOTUS_REPORTER_REGEX = /^(?:U\\.?\\s?S\\.|S\\.?\\s?Ct\\.|L\\.?\\s?Ed\\.)/\n\n/** Citation boundary pattern (digit-period-space) */\nconst CITATION_BOUNDARY_REGEX = /\\d\\.\\s+/g\n\n/** Standard \"v.\" or \"vs.\" case name format */\nconst V_CASE_NAME_REGEX = /([A-Z][A-Za-z0-9\\s.,'&()/-]+?)\\s+v(?:s)?\\.?\\s+([A-Za-z0-9\\s.,'&()/-]+?)\\s*,\\s*$/\n\n/** Procedural prefix case name format */\nconst PROCEDURAL_PREFIX_REGEX = /\\b(In re|Ex parte|Matter of|Estate of|State ex rel\\.|United States ex rel\\.|Application of|Petition of)\\s+([A-Za-z0-9\\s.,'&()/-]+?)\\s*,\\s*$/i\n\n/**\n * Strips date components (month, day, year) from parenthetical content\n * to isolate the court abbreviation.\n * E.g., \"2d Cir. Jan. 15, 2020\" → \"2d Cir.\"\n *        \"C.D. Cal. Feb. 9, 2015\" → \"C.D. Cal.\"\n *        \"D. Mass. Mar. 2020\" → \"D. Mass.\"\n *        \"D. Mass. 1/15/2020\" → \"D. Mass.\"\n */\nfunction stripDateFromCourt(content: string): string | undefined {\n\t// Strip trailing numeric date format first (1/15/2020)\n\tlet court = content.replace(/\\s*\\d{1,2}\\/\\d{1,2}\\/\\d{4}\\s*$/, '').trim()\n\t// Strip trailing year\n\tcourt = court.replace(/\\s*\\d{4}\\s*$/, '').trim()\n\t// Strip trailing date components: optional day+comma, month abbreviation or full name\n\tcourt = court.replace(/\\s*,?\\s*\\d{1,2}\\s*,?\\s*$/, '').trim()\n\tcourt = court.replace(new RegExp(`\\\\s*${MONTH_PATTERN.source}\\\\s*$`, 'i'), '').trim()\n\t// Strip any trailing commas left over\n\tcourt = court.replace(/,\\s*$/, '').trim()\n\treturn court && /[A-Za-z]/.test(court) ? court : undefined\n}\n\n/**\n * Extract case name via backward search from citation core.\n * Looks for \"v.\" pattern or procedural prefixes (In re, Ex parte, Matter of).\n *\n * @param cleanedText - Full cleaned text\n * @param coreStart - Position where citation core begins (volume start)\n * @param maxLookback - Maximum characters to search backward (default 150)\n * @returns Case name and start position, or undefined if not found\n *\n * @example\n * ```typescript\n * extractCaseName(text, 20, 150)\n * // Returns: { caseName: \"Smith v. Jones\", nameStart: 0 }\n * ```\n */\nfunction extractCaseName(\n\tcleanedText: string,\n\tcoreStart: number,\n\tmaxLookback = 150,\n): { caseName: string; nameStart: number } | undefined {\n\tconst searchStart = Math.max(0, coreStart - maxLookback)\n\tlet precedingText = cleanedText.substring(searchStart, coreStart)\n\tlet adjustedSearchStart = searchStart\n\n\t// Split at last sentence boundary to avoid crossing citation boundaries\n\t// Find last occurrence of digit-period-space pattern (end of reporter page like \"10. Jones\")\n\t// This is more specific than generic \". [A-Z]\" which would match \"v.\" or \"United States v.\"\n\tlet lastBoundaryIndex = -1\n\tlet match: RegExpExecArray | null\n\twhile ((match = CITATION_BOUNDARY_REGEX.exec(precedingText)) !== null) {\n\t\tlastBoundaryIndex = match.index + match[0].length\n\t}\n\n\tif (lastBoundaryIndex !== -1) {\n\t\tprecedingText = precedingText.substring(lastBoundaryIndex)\n\t\tadjustedSearchStart = searchStart + lastBoundaryIndex\n\t}\n\n\t// Priority 1: Standard \"v.\" or \"vs.\" format with comma before citation\n\t// Match party names with letters, numbers (for \"Doe No. 2\"), periods, apostrophes, ampersands, hyphens, slashes\n\tconst vMatch = V_CASE_NAME_REGEX.exec(precedingText)\n\tif (vMatch) {\n\t\t// Check for semicolon in matched text (multi-citation separator)\n\t\tif (!vMatch[0].includes(';')) {\n\t\t\tconst caseName = `${vMatch[1].trim()} v. ${vMatch[2].trim()}`\n\t\t\tconst nameStart = adjustedSearchStart + vMatch.index\n\t\t\treturn { caseName, nameStart }\n\t\t}\n\t}\n\n\t// Priority 2: Procedural prefixes (including Estate of)\n\tconst procMatch = PROCEDURAL_PREFIX_REGEX.exec(precedingText)\n\tif (procMatch) {\n\t\t// Check for semicolon in matched text (multi-citation separator)\n\t\tif (!procMatch[0].includes(';')) {\n\t\t\tconst caseName = `${procMatch[1]} ${procMatch[2].trim()}`\n\t\t\tconst nameStart = adjustedSearchStart + procMatch.index\n\t\t\treturn { caseName, nameStart }\n\t\t}\n\t}\n\n\treturn undefined\n}\n\n/**\n * Find the end of parenthetical content, including chained parentheticals and subsequent history.\n * Tracks paren depth to handle nested parens, and continues scanning for chained parens.\n *\n * @param cleanedText - Full cleaned text\n * @param searchStart - Position to start searching from (after citation core)\n * @param maxLookahead - Maximum characters to search forward (default 200)\n * @returns Position after final closing paren (exclusive), or searchStart if no parens\n *\n * @example\n * ```typescript\n * findParentheticalEnd(text, 20, 200)\n * // For \"(2020) (en banc)\" returns position after final \")\"\n * ```\n */\nfunction findParentheticalEnd(\n\tcleanedText: string,\n\tsearchStart: number,\n\tmaxLookahead = 200,\n): number {\n\tlet pos = searchStart\n\tconst endLimit = Math.min(cleanedText.length, searchStart + maxLookahead)\n\tlet depth = 0\n\tlet foundAnyParen = false\n\n\twhile (pos < endLimit) {\n\t\tconst char = cleanedText[pos]\n\n\t\tif (char === '(') {\n\t\t\tdepth++\n\t\t\tfoundAnyParen = true\n\t\t\tpos++\n\t\t} else if (char === ')') {\n\t\t\tdepth--\n\t\t\tpos++\n\n\t\t\t// When depth returns to 0, check for chained paren or subsequent history\n\t\t\tif (depth === 0) {\n\t\t\t\t// Skip whitespace\n\t\t\t\tlet nextPos = pos\n\t\t\t\twhile (nextPos < endLimit && /\\s/.test(cleanedText[nextPos])) {\n\t\t\t\t\tnextPos++\n\t\t\t\t}\n\n\t\t\t\t// Check for chained parenthetical\n\t\t\t\tif (cleanedText[nextPos] === '(') {\n\t\t\t\t\tpos = nextPos\n\t\t\t\t\tcontinue\n\t\t\t\t}\n\n\t\t\t\t// Check for subsequent history signals\n\t\t\t\tconst remainingText = cleanedText.substring(nextPos, endLimit)\n\t\t\t\tconst historyRegex =\n\t\t\t\t\t/^,\\s*(aff'd|rev'd|cert\\.\\s*denied|overruled\\s+by|vacated\\s+by)/i\n\t\t\t\tif (historyRegex.test(remainingText)) {\n\t\t\t\t\t// Continue scanning - subsequent history has its own paren\n\t\t\t\t\tpos = nextPos\n\t\t\t\t\tcontinue\n\t\t\t\t}\n\n\t\t\t\t// No chained paren or subsequent history - we're done\n\t\t\t\treturn pos\n\t\t\t}\n\t\t} else {\n\t\t\tpos++\n\t\t}\n\t}\n\n\t// If we found parens but didn't close them all, return where we stopped\n\t// If we never found parens, return searchStart\n\treturn foundAnyParen ? pos : searchStart\n}\n\n/**\n * Parse parenthetical content to extract court, year, date, and disposition.\n * Unified parser replacing the old year-only logic.\n *\n * @param content - Parenthetical content (without the parens themselves)\n * @returns Structured parenthetical data\n *\n * @example\n * ```typescript\n * parseParenthetical(\"9th Cir. 2020\")\n * // Returns: { court: \"9th Cir.\", year: 2020, date: { iso: \"2020\", parsed: { year: 2020 } } }\n *\n * parseParenthetical(\"2d Cir. Jan. 15, 2020\")\n * // Returns: { court: \"2d Cir.\", year: 2020, date: { iso: \"2020-01-15\", parsed: { year: 2020, month: 1, day: 15 } } }\n *\n * parseParenthetical(\"en banc\")\n * // Returns: { disposition: \"en banc\" }\n * ```\n */\nfunction parseParenthetical(content: string): {\n\tcourt?: string\n\tyear?: number\n\tdate?: StructuredDate\n\tdisposition?: string\n} {\n\tconst result: {\n\t\tcourt?: string\n\t\tyear?: number\n\t\tdate?: StructuredDate\n\t\tdisposition?: string\n\t} = {}\n\n\t// Parse structured date using dates.ts\n\tconst dateResult = parseDate(content)\n\tif (dateResult) {\n\t\tresult.date = dateResult\n\t\tresult.year = dateResult.parsed.year\n\t}\n\n\t// Extract court (strips date components)\n\tconst courtResult = stripDateFromCourt(content)\n\tif (courtResult) {\n\t\tresult.court = courtResult\n\t}\n\n\t// Check for disposition\n\tif (/\\ben banc\\b/i.test(content)) {\n\t\tresult.disposition = 'en banc'\n\t} else if (/\\bper curiam\\b/i.test(content)) {\n\t\tresult.disposition = 'per curiam'\n\t}\n\n\treturn result\n}\n\n/**\n * Normalize party name for matching by removing legal noise.\n * Normalization pipeline:\n * 1. Strip \"et al.\" (case-insensitive)\n * 2. Strip \"d/b/a\" and everything after (case-insensitive)\n * 3. Strip \"aka\" and everything after (case-insensitive, word boundary)\n * 4. Strip trailing corporate suffixes (Inc., LLC, Corp., Ltd., Co., LLP, LP, P.C.) - iterative\n * 5. Strip leading articles (The, A, An)\n * 6. Normalize whitespace\n * 7. Trim and lowercase\n *\n * @param name - Raw party name\n * @returns Normalized party name\n *\n * @example\n * ```typescript\n * normalizePartyName(\"The Smith Corp., Inc.\") // \"smith\"\n * normalizePartyName(\"Doe et al.\") // \"doe\"\n * normalizePartyName(\"United States\") // \"united states\" (not stripped)\n * ```\n */\nfunction normalizePartyName(name: string): string {\n\tlet normalized = name\n\n\t// Strip \"et al.\" (with or without period, case-insensitive)\n\tnormalized = normalized.replace(/\\bet\\s+al\\.?/gi, '')\n\n\t// Strip \"d/b/a\" and everything after it (case-insensitive)\n\tnormalized = normalized.replace(/\\s+d\\/b\\/a\\b.*/gi, '')\n\n\t// Strip \"aka\" and everything after it (case-insensitive, word boundary)\n\tnormalized = normalized.replace(/\\s+aka\\b.*/gi, '')\n\n\t// Strip trailing corporate suffixes (with or without trailing period, handle comma)\n\t// Repeat to handle multiple suffixes like \"Corp., Inc.\"\n\tlet prev = ''\n\twhile (prev !== normalized) {\n\t\tprev = normalized\n\t\tnormalized = normalized.replace(/,?\\s*(Inc|LLC|Corp|Ltd|Co|LLP|LP|P\\.C)\\.?$/gi, '')\n\t}\n\n\t// Strip leading articles (only at start)\n\tnormalized = normalized.replace(/^(The|A|An)\\s+/i, '')\n\n\t// Normalize whitespace (collapse multiple spaces)\n\tnormalized = normalized.replace(/\\s+/g, ' ')\n\n\t// Trim and lowercase\n\treturn normalized.trim().toLowerCase()\n}\n\n/**\n * Extract plaintiff and defendant party names from case name.\n * Handles adversarial cases (v.) and procedural prefixes (In re, Ex parte, etc.).\n *\n * @param caseName - Case name string\n * @returns Party name data with raw and normalized fields\n *\n * @example\n * ```typescript\n * extractPartyNames(\"Smith v. Jones\")\n * // Returns: { plaintiff: \"Smith\", plaintiffNormalized: \"smith\", defendant: \"Jones\", defendantNormalized: \"jones\" }\n *\n * extractPartyNames(\"In re Smith\")\n * // Returns: { plaintiff: \"In re Smith\", plaintiffNormalized: \"smith\", proceduralPrefix: \"In re\" }\n *\n * extractPartyNames(\"People v. Smith\")\n * // Returns: { plaintiff: \"People\", plaintiffNormalized: \"people\", defendant: \"Smith\", defendantNormalized: \"smith\" }\n * ```\n */\nfunction extractPartyNames(caseName: string): {\n\tplaintiff?: string\n\tplaintiffNormalized?: string\n\tdefendant?: string\n\tdefendantNormalized?: string\n\tproceduralPrefix?: string\n} {\n\t// Procedural prefix patterns (anchored to start, case-insensitive)\n\tconst proceduralPrefixes = [\n\t\t'In re',\n\t\t'Ex parte',\n\t\t'Matter of',\n\t\t'State ex rel.',\n\t\t'United States ex rel.',\n\t\t'Application of',\n\t\t'Petition of',\n\t\t'Estate of',\n\t]\n\n\t// Check for procedural prefix first\n\tfor (const prefix of proceduralPrefixes) {\n\t\tconst prefixRegex = new RegExp(`^(${prefix})\\\\s+(.+)$`, 'i')\n\t\tconst match = prefixRegex.exec(caseName)\n\t\tif (match) {\n\t\t\tconst matchedPrefix = match[1]\n\t\t\tconst subject = match[2]\n\n\t\t\t// Check if there's a \"v.\" after the prefix (adversarial case)\n\t\t\tif (/\\s+v\\.?\\s+/i.test(subject)) {\n\t\t\t\t// Adversarial case with procedural-looking plaintiff (e.g., \"Estate of X v. Y\")\n\t\t\t\t// Split on \"v.\"\n\t\t\t\tconst vMatch = /^(.+?)\\s+v\\.?\\s+(.+)$/i.exec(caseName)\n\t\t\t\tif (vMatch) {\n\t\t\t\t\tconst plaintiff = vMatch[1].trim()\n\t\t\t\t\tconst defendant = vMatch[2].trim()\n\t\t\t\t\treturn {\n\t\t\t\t\t\tplaintiff,\n\t\t\t\t\t\tplaintiffNormalized: normalizePartyName(plaintiff),\n\t\t\t\t\t\tdefendant,\n\t\t\t\t\t\tdefendantNormalized: normalizePartyName(defendant),\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t} else {\n\t\t\t\t// Pure procedural (no \"v.\")\n\t\t\t\treturn {\n\t\t\t\t\tplaintiff: caseName,\n\t\t\t\t\tplaintiffNormalized: normalizePartyName(subject),\n\t\t\t\t\tproceduralPrefix: matchedPrefix,\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\n\t// Split on \"v.\" for adversarial cases\n\tconst vRegex = /^(.+?)\\s+v\\.?\\s+(.+)$/i\n\tconst vMatch = vRegex.exec(caseName)\n\tif (vMatch) {\n\t\tlet plaintiff = vMatch[1].trim()\n\t\tconst defendant = vMatch[2].trim()\n\n\t\t// Strip signal words from plaintiff (e.g., \"In Smith\" → \"Smith\", \"See Jones\" → \"Jones\")\n\t\t// Preserve \"In re\" which is a procedural prefix, not a signal word\n\t\tplaintiff = plaintiff.replace(/^(?:In(?!\\s+re\\b)|See(?:\\s+[Aa]lso)?|Compare|But(?:\\s+[Ss]ee)?|Cf\\.?|Also)\\s+/i, '').trim()\n\n\t\treturn {\n\t\t\tplaintiff: plaintiff || vMatch[1].trim(), // Fallback to original if strip leaves nothing\n\t\t\tplaintiffNormalized: normalizePartyName(plaintiff || vMatch[1].trim()),\n\t\t\tdefendant,\n\t\t\tdefendantNormalized: normalizePartyName(defendant),\n\t\t}\n\t}\n\n\t// No \"v.\" and no procedural prefix - no parties extracted\n\treturn {}\n}\n\n/**\n * Extracts case citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Volume: Leading digits (e.g., \"500\" from \"500 F.2d 123\")\n * - Reporter: Alphabetic abbreviation (e.g., \"F.2d\")\n * - Page: Trailing digits after reporter (e.g., \"123\")\n * - Pincite: Optional page reference after comma (e.g., \", 125\")\n * - Court: Optional court abbreviation in parentheses (e.g., \"(9th Cir.)\")\n * - Year: Optional year in parentheses (e.g., \"(2020)\")\n *\n * Confidence scoring:\n * - Base: 0.5\n * - Common reporter pattern (F., U.S., etc.): +0.3\n * - Valid year (not future): +0.2\n * - Capped at 1.0\n *\n * Position translation:\n * - Uses TransformationMap to convert clean positions → original positions\n * - cleanStart/cleanEnd from token span\n * - originalStart/originalEnd via transformationMap.cleanToOriginal\n *\n * Note: This function does NOT validate against reporters-db. That happens\n * in Phase 3 (resolution layer). Phase 2 extraction only parses structure.\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns FullCaseCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"500 F.2d 123, 125\",\n *   span: { cleanStart: 10, cleanEnd: 27 },\n *   type: \"case\",\n *   patternId: \"federal-reporter\"\n * }\n * const citation = extractCase(token, transformationMap)\n * // citation = {\n * //   type: \"case\",\n * //   text: \"500 F.2d 123, 125\",\n * //   volume: 500,\n * //   reporter: \"F.2d\",\n * //   page: 123,\n * //   pincite: 125,\n * //   span: { cleanStart: 10, cleanEnd: 27, originalStart: 10, originalEnd: 27 },\n * //   confidence: 0.8,\n * //   ...\n * // }\n * ```\n */\nexport function extractCase(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n\tcleanedText?: string,\n): FullCaseCitation {\n\tconst { text, span } = token\n\n\t// Parse volume-reporter-page using regex\n\t// Pattern: volume (digits) + reporter (letters/periods/spaces/numbers) + page (digits or blank placeholder)\n\t// Use greedy matching for reporter to capture full abbreviation including spaces\n\tconst match = VOLUME_REPORTER_PAGE_REGEX.exec(text)\n\n\tif (!match) {\n\t\t// Fallback if pattern doesn't match (shouldn't happen if tokenizer is correct)\n\t\tthrow new Error(`Failed to parse case citation: ${text}`)\n\t}\n\n\tconst volume = parseVolume(match[1])\n\tconst reporter = match[2].trim()\n\n\t// Check if page is a blank placeholder\n\tconst pageStr = match[3]\n\tconst isBlankPage = BLANK_PAGE_REGEX.test(pageStr)\n\tconst page = isBlankPage ? undefined : Number.parseInt(pageStr, 10)\n\tconst hasBlankPage = isBlankPage ? true : undefined\n\n\t// Extract optional pincite (page reference after comma)\n\t// Pattern: \", digits\" (e.g., \", 125\")\n\tconst pinciteMatch = PINCITE_REGEX.exec(text)\n\tlet pincite = pinciteMatch ? Number.parseInt(pinciteMatch[1], 10) : undefined\n\n\t// Initialize Phase 6 fields\n\tlet year: number | undefined\n\tlet court: string | undefined\n\tlet date: StructuredDate | undefined\n\tlet disposition: string | undefined\n\tlet caseName: string | undefined\n\tlet fullSpan: Span | undefined\n\n\t// Extract parenthetical from token text\n\tlet parentheticalContent: string | undefined\n\t// Match any parenthetical (with or without letters)\n\tconst parenMatch = PAREN_REGEX.exec(text)\n\tif (parenMatch) {\n\t\tparentheticalContent = parenMatch[1]\n\t\t// Parse parenthetical using unified parser\n\t\tconst parenResult = parseParenthetical(parentheticalContent)\n\t\tyear = parenResult.year\n\t\tcourt = parenResult.court\n\t\tdate = parenResult.date\n\t\tdisposition = parenResult.disposition\n\t}\n\n\t// Look ahead in cleaned text for parenthetical after the token\n\t// Tokenization patterns only capture volume-reporter-page, so parentheticals\n\t// like \"(1989)\" or \"(9th Cir. 2020)\" are not in the token text.\n\tif (cleanedText && !parentheticalContent) {\n\t\tconst afterToken = cleanedText.substring(span.cleanEnd)\n\t\tconst lookAheadMatch = LOOKAHEAD_PAREN_REGEX.exec(afterToken)\n\t\tif (lookAheadMatch) {\n\t\t\tparentheticalContent = lookAheadMatch[1]\n\t\t\t// Parse parenthetical using unified parser\n\t\t\tconst parenResult = parseParenthetical(parentheticalContent)\n\t\t\tyear = parenResult.year\n\t\t\tcourt = parenResult.court\n\t\t\tdate = parenResult.date\n\t\t\tdisposition = parenResult.disposition\n\n\t\t\t// Extract pincite from look-ahead if not already found in token text\n\t\t\tif (pincite === undefined) {\n\t\t\t\tconst laPinciteMatch = LOOKAHEAD_PINCITE_REGEX.exec(afterToken)\n\t\t\t\tif (laPinciteMatch) {\n\t\t\t\t\tpincite = Number.parseInt(laPinciteMatch[1], 10)\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\n\t// Check for chained parentheticals with disposition (e.g., \"(2020) (en banc)\")\n\tif (cleanedText && !disposition) {\n\t\tconst afterToken = cleanedText.substring(span.cleanEnd)\n\t\t// Look for second parenthetical after first one\n\t\tconst chainedMatch = CHAINED_DISPOSITION_REGEX.exec(afterToken)\n\t\tif (chainedMatch) {\n\t\t\tdisposition = chainedMatch[1].toLowerCase()\n\t\t}\n\t}\n\n\t// Infer court from reporter for known Supreme Court reporters\n\tif (!court && SCOTUS_REPORTER_REGEX.test(reporter)) {\n\t\tcourt = 'scotus'\n\t}\n\n\t// Phase 6: Extract case name via backward search\n\tif (cleanedText) {\n\t\tconst caseNameResult = extractCaseName(cleanedText, span.cleanStart)\n\t\tif (caseNameResult) {\n\t\t\tcaseName = caseNameResult.caseName\n\n\t\t\t// Calculate fullSpan: case name start through parenthetical end\n\t\t\tconst parenEnd = findParentheticalEnd(cleanedText, span.cleanEnd)\n\t\t\tconst fullCleanStart = caseNameResult.nameStart\n\t\t\tconst fullCleanEnd = parenEnd > span.cleanEnd ? parenEnd : span.cleanEnd\n\n\t\t\t// Translate to original positions\n\t\t\tconst fullOriginalStart =\n\t\t\t\ttransformationMap.cleanToOriginal.get(fullCleanStart) ?? fullCleanStart\n\t\t\tconst fullOriginalEnd =\n\t\t\t\ttransformationMap.cleanToOriginal.get(fullCleanEnd) ?? fullCleanEnd\n\n\t\t\tfullSpan = {\n\t\t\t\tcleanStart: fullCleanStart,\n\t\t\t\tcleanEnd: fullCleanEnd,\n\t\t\t\toriginalStart: fullOriginalStart,\n\t\t\t\toriginalEnd: fullOriginalEnd,\n\t\t\t}\n\t\t}\n\t}\n\n\t// Phase 7: Extract party names from case name\n\tlet plaintiff: string | undefined\n\tlet plaintiffNormalized: string | undefined\n\tlet defendant: string | undefined\n\tlet defendantNormalized: string | undefined\n\tlet proceduralPrefix: string | undefined\n\n\tif (caseName) {\n\t\tconst partyResult = extractPartyNames(caseName)\n\t\tplaintiff = partyResult.plaintiff\n\t\tplaintiffNormalized = partyResult.plaintiffNormalized\n\t\tdefendant = partyResult.defendant\n\t\tdefendantNormalized = partyResult.defendantNormalized\n\t\tproceduralPrefix = partyResult.proceduralPrefix\n\t}\n\n\t// Translate positions from clean → original (citation core only - span unchanged)\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Calculate confidence score\n\tlet confidence = 0.5 // Base confidence\n\n\t// Common reporter patterns (F., U.S., S. Ct., etc.)\n\tconst commonReporters = [\n\t\t'F.',\n\t\t'F.2d',\n\t\t'F.3d',\n\t\t'F.4th',\n\t\t'U.S.',\n\t\t'S. Ct.',\n\t\t'L. Ed.',\n\t\t'P.',\n\t\t'P.2d',\n\t\t'P.3d',\n\t\t'A.',\n\t\t'A.2d',\n\t\t'A.3d',\n\t\t'N.E.',\n\t\t'N.E.2d',\n\t\t'N.E.3d',\n\t\t'N.W.',\n\t\t'N.W.2d',\n\t\t'S.E.',\n\t\t'S.E.2d',\n\t\t'S.W.',\n\t\t'S.W.2d',\n\t\t'S.W.3d',\n\t\t'So.',\n\t\t'So. 2d',\n\t\t'So. 3d',\n\t]\n\n\tif (commonReporters.some((r) => reporter.includes(r))) {\n\t\tconfidence += 0.3\n\t}\n\n\t// Valid year check (not in future)\n\tif (year !== undefined) {\n\t\tconst currentYear = new Date().getFullYear()\n\t\tif (year <= currentYear) {\n\t\t\tconfidence += 0.2\n\t\t}\n\t}\n\n\t// Cap at 1.0\n\tconfidence = Math.min(confidence, 1.0)\n\n\t// Override confidence for blank page citations\n\tif (hasBlankPage) {\n\t\tconfidence = 0.8\n\t}\n\n\treturn {\n\t\ttype: 'case',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0, // Placeholder - timing handled by orchestration layer\n\t\tpatternsChecked: 1, // Single token processed\n\t\tvolume,\n\t\treporter,\n\t\tpage,\n\t\tpincite,\n\t\tcourt,\n\t\tyear,\n\t\thasBlankPage,\n\t\tdate,\n\t\tfullSpan,\n\t\tcaseName,\n\t\tdisposition,\n\t\tplaintiff,\n\t\tplaintiffNormalized,\n\t\tdefendant,\n\t\tdefendantNormalized,\n\t\tproceduralPrefix,\n\t}\n}\n","/**\n * Statute Citation Extraction\n *\n * Parses tokenized statute citations to extract title, code, section, and\n * optional subsections. Examples: \"42 U.S.C. § 1983\", \"Cal. Civ. Code § 1234(a)(1)\"\n *\n * @module extract/extractStatute\n */\n\nimport type { Token } from '@/tokenize'\nimport type { StatuteCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts statute citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Title: Optional leading digits (e.g., \"42\" from \"42 U.S.C. § 1983\")\n * - Code: Statutory code abbreviation (e.g., \"U.S.C.\", \"Cal. Civ. Code\")\n * - Section: Section number after § symbol (e.g., \"1983\")\n * - Subsections: Optional parenthetical subdivisions (e.g., \"(a)(1)\")\n *\n * Confidence scoring:\n * - Base: 0.5\n * - Known code pattern (U.S.C., C.F.R., state codes): +0.3\n * - Capped at 1.0\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns StatuteCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"42 U.S.C. § 1983\",\n *   span: { cleanStart: 10, cleanEnd: 26 },\n *   type: \"statute\",\n *   patternId: \"usc\"\n * }\n * const citation = extractStatute(token, transformationMap)\n * // citation = {\n * //   type: \"statute\",\n * //   title: 42,\n * //   code: \"U.S.C.\",\n * //   section: \"1983\",\n * //   ...\n * // }\n * ```\n */\nexport function extractStatute(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): StatuteCitation {\n\tconst { text, span } = token\n\n\t// Parse title-code-section using regex\n\t// Pattern: optional title (digits) + code (letters/periods/spaces) + § + section\n\tconst statuteRegex = /^(?:(\\d+)\\s+)?([A-Za-z.\\s]+?)\\s*§\\s*(\\d+[A-Za-z0-9-]*)/\n\tconst match = statuteRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse statute citation: ${text}`)\n\t}\n\n\tconst title = match[1] ? Number.parseInt(match[1], 10) : undefined\n\tconst code = match[2].trim()\n\tconst section = match[3]\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Calculate confidence score\n\tlet confidence = 0.5 // Base confidence\n\n\t// Known statutory code patterns\n\tconst knownCodes = [\n\t\t'U.S.C.',\n\t\t'C.F.R.',\n\t\t'Cal. Civ. Code',\n\t\t'Cal. Penal Code',\n\t\t'N.Y. Civ. Prac. L. & R.',\n\t\t'Tex. Civ. Prac. & Rem. Code',\n\t]\n\n\tif (knownCodes.some((c) => code.includes(c))) {\n\t\tconfidence += 0.3\n\t}\n\n\tconfidence = Math.min(confidence, 1.0)\n\n\treturn {\n\t\ttype: 'statute',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\ttitle,\n\t\tcode,\n\t\tsection,\n\t}\n}\n","/**\n * Journal Citation Extraction\n *\n * Parses tokenized journal citations to extract volume, journal name, page,\n * and optional metadata. Examples: \"123 Harv. L. Rev. 456\", \"75 Yale L.J. 789, 791\"\n *\n * @module extract/extractJournal\n */\n\nimport type { Token } from '@/tokenize'\nimport type { JournalCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts journal citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Volume: Leading digits (e.g., \"123\" from \"123 Harv. L. Rev. 456\")\n * - Journal: Journal abbreviation (e.g., \"Harv. L. Rev.\")\n * - Page: Starting page number (e.g., \"456\")\n * - Pincite: Optional specific page reference after comma (e.g., \", 458\")\n *\n * Confidence scoring:\n * - Base: 0.6 (journal validation happens in Phase 3)\n *\n * Note: Author and title extraction from preceding text is not implemented\n * in Phase 2. That requires context analysis in Phase 3.\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns JournalCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"123 Harv. L. Rev. 456\",\n *   span: { cleanStart: 10, cleanEnd: 31 },\n *   type: \"journal\",\n *   patternId: \"journal-standard\"\n * }\n * const citation = extractJournal(token, transformationMap)\n * // citation = {\n * //   type: \"journal\",\n * //   volume: 123,\n * //   journal: \"Harv. L. Rev.\",\n * //   abbreviation: \"Harv. L. Rev.\",\n * //   page: 456,\n * //   ...\n * // }\n * ```\n */\nexport function extractJournal(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): JournalCitation {\n\tconst { text, span } = token\n\n\t// Parse volume-journal-page using regex\n\t// Pattern: volume (digits) + journal (letters/periods/spaces) + page (digits)\n\tconst journalRegex = /^(\\d+(?:-\\d+)?)\\s+([A-Za-z.\\s]+?)\\s+(\\d+)/\n\tconst match = journalRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse journal citation: ${text}`)\n\t}\n\n\tconst rawVolume = match[1]\n\tconst volume = /^\\d+$/.test(rawVolume) ? Number.parseInt(rawVolume, 10) : rawVolume\n\tconst journal = match[2].trim()\n\tconst page = Number.parseInt(match[3], 10)\n\n\t// Extract optional pincite (page reference after comma)\n\tconst pinciteRegex = /,\\s*(\\d+)/\n\tconst pinciteMatch = pinciteRegex.exec(text)\n\tconst pincite = pinciteMatch ? Number.parseInt(pinciteMatch[1], 10) : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.6 base (journal validation against database happens in Phase 3)\n\tconst confidence = 0.6\n\n\treturn {\n\t\ttype: 'journal',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tvolume,\n\t\tjournal,\n\t\tabbreviation: journal, // For Phase 2, abbreviation = journal name\n\t\tpage,\n\t\tpincite,\n\t}\n}\n","/**\n * Neutral Citation Extraction\n *\n * Parses tokenized neutral (vendor-neutral) citations to extract year, court,\n * and document number. Examples: \"2020 WL 123456\", \"2020 U.S. LEXIS 456\"\n *\n * @module extract/extractNeutral\n */\n\nimport type { Token } from '@/tokenize'\nimport type { NeutralCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts neutral citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Year: 4-digit year (e.g., \"2020\")\n * - Court: Vendor identifier (e.g., \"WL\", \"U.S. LEXIS\")\n * - Document number: Unique document identifier (e.g., \"123456\")\n *\n * Confidence scoring:\n * - 1.0 (neutral format is unambiguous and standardized)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns NeutralCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"2020 WL 123456\",\n *   span: { cleanStart: 10, cleanEnd: 24 },\n *   type: \"neutral\",\n *   patternId: \"westlaw-neutral\"\n * }\n * const citation = extractNeutral(token, transformationMap)\n * // citation = {\n * //   type: \"neutral\",\n * //   year: 2020,\n * //   court: \"WL\",\n * //   documentNumber: \"123456\",\n * //   confidence: 1.0,\n * //   ...\n * // }\n * ```\n */\nexport function extractNeutral(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): NeutralCitation {\n\tconst { text, span } = token\n\n\t// Parse year-court-documentNumber using regex\n\t// Pattern: 4-digit year + court identifier (WL, LEXIS, etc.) + document number\n\tconst neutralRegex = /^(\\d{4})\\s+(WL|LEXIS|U\\.S\\.\\s+LEXIS)\\s+(\\d+)/\n\tconst match = neutralRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse neutral citation: ${text}`)\n\t}\n\n\tconst year = Number.parseInt(match[1], 10)\n\tconst court = match[2]\n\tconst documentNumber = match[3]\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 1.0 (neutral format is unambiguous)\n\tconst confidence = 1.0\n\n\treturn {\n\t\ttype: 'neutral',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tyear,\n\t\tcourt,\n\t\tdocumentNumber,\n\t}\n}\n","/**\n * Public Law Citation Extraction\n *\n * Parses tokenized public law citations to extract congress number and law number.\n * Examples: \"Pub. L. No. 116-283\", \"Pub. L. 117-58\"\n *\n * @module extract/extractPublicLaw\n */\n\nimport type { Token } from '@/tokenize'\nimport type { PublicLawCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts public law citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Congress: Congress number (e.g., \"116\" from \"Pub. L. No. 116-283\")\n * - Law number: Law number within that Congress (e.g., \"283\")\n *\n * Confidence scoring:\n * - 0.9 (public law format is fairly standard)\n *\n * Note: Bill title extraction from nearby text is not implemented in Phase 2.\n * That requires context analysis in Phase 3.\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns PublicLawCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"Pub. L. No. 116-283\",\n *   span: { cleanStart: 10, cleanEnd: 29 },\n *   type: \"publicLaw\",\n *   patternId: \"public-law\"\n * }\n * const citation = extractPublicLaw(token, transformationMap)\n * // citation = {\n * //   type: \"publicLaw\",\n * //   congress: 116,\n * //   lawNumber: 283,\n * //   confidence: 0.9,\n * //   ...\n * // }\n * ```\n */\nexport function extractPublicLaw(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): PublicLawCitation {\n\tconst { text, span } = token\n\n\t// Parse congress-lawNumber using regex\n\t// Pattern: \"Pub. L.\" (with optional \"No.\") + congress number + \"-\" + law number\n\tconst publicLawRegex = /Pub\\.\\s?L\\.(?:\\s?No\\.)?\\s?(\\d+)-(\\d+)/\n\tconst match = publicLawRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse public law citation: ${text}`)\n\t}\n\n\tconst congress = Number.parseInt(match[1], 10)\n\tconst lawNumber = Number.parseInt(match[2], 10)\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.9 (public law format is fairly standard)\n\tconst confidence = 0.9\n\n\treturn {\n\t\ttype: 'publicLaw',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tcongress,\n\t\tlawNumber,\n\t}\n}\n","/**\n * Federal Register Citation Extraction\n *\n * Parses tokenized Federal Register citations to extract volume, page, and\n * optional year. Examples: \"85 Fed. Reg. 12345\", \"86 Fed. Reg. 56789 (Jan. 15, 2021)\"\n *\n * @module extract/extractFederalRegister\n */\n\nimport type { Token } from '@/tokenize'\nimport type { FederalRegisterCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts Federal Register citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Volume: Federal Register volume number (e.g., \"85\")\n * - Page: Page number (e.g., \"12345\")\n * - Year: Optional publication year in parentheses (e.g., \"(2021)\")\n *\n * Confidence scoring:\n * - 0.9 (Federal Register format is standardized)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns FederalRegisterCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"85 Fed. Reg. 12345\",\n *   span: { cleanStart: 10, cleanEnd: 28 },\n *   type: \"federalRegister\",\n *   patternId: \"federal-register\"\n * }\n * const citation = extractFederalRegister(token, transformationMap)\n * // citation = {\n * //   type: \"federalRegister\",\n * //   volume: 85,\n * //   page: 12345,\n * //   confidence: 0.9,\n * //   ...\n * // }\n * ```\n */\nexport function extractFederalRegister(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): FederalRegisterCitation {\n\tconst { text, span } = token\n\n\t// Parse volume-page using regex\n\t// Pattern: volume (digits) + \"Fed. Reg.\" + page (digits)\n\tconst federalRegisterRegex = /^(\\d+(?:-\\d+)?)\\s+Fed\\.\\s?Reg\\.\\s+(\\d+)/\n\tconst match = federalRegisterRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse Federal Register citation: ${text}`)\n\t}\n\n\tconst rawVolume = match[1]\n\tconst volume = /^\\d+$/.test(rawVolume) ? Number.parseInt(rawVolume, 10) : rawVolume\n\tconst page = Number.parseInt(match[2], 10)\n\n\t// Extract optional year in parentheses\n\t// Pattern: \"(year)\" or \"(month day, year)\"\n\tconst yearRegex = /\\((?:.*?\\s)?(\\d{4})\\)/\n\tconst yearMatch = yearRegex.exec(text)\n\tconst year = yearMatch ? Number.parseInt(yearMatch[1], 10) : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.9 (Federal Register format is standardized)\n\tconst confidence = 0.9\n\n\treturn {\n\t\ttype: 'federalRegister',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tvolume,\n\t\tpage,\n\t\tyear,\n\t}\n}\n","/**\n * Statutes at Large Citation Extractor\n *\n * Extracts session law citations from the Statutes at Large (e.g., \"124 Stat. 119\").\n * These are chronological compilations of federal laws, distinct from both\n * codified statutes (U.S.C.) and case reporters.\n *\n * Format: volume Stat. page [(year)]\n *\n * @module extract/extractStatutesAtLarge\n */\n\nimport type { StatutesAtLargeCitation } from \"@/types/citation\"\nimport type { TransformationMap } from \"@/types/span\"\nimport type { Token } from \"@/tokenize/tokenizer\"\n\nexport function extractStatutesAtLarge(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): StatutesAtLargeCitation {\n\tconst { text, span } = token\n\n\t// Parse volume-Stat.-page\n\tconst statRegex = /^(\\d+(?:-\\d+)?)\\s+Stat\\.\\s+(\\d+)/\n\tconst match = statRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse Statutes at Large citation: ${text}`)\n\t}\n\n\tconst rawVolume = match[1]\n\tconst volume = /^\\d+$/.test(rawVolume) ? Number.parseInt(rawVolume, 10) : rawVolume\n\tconst page = Number.parseInt(match[2], 10)\n\n\t// Extract optional year in parentheses\n\tconst yearRegex = /\\((?:.*?\\s)?(\\d{4})\\)/\n\tconst yearMatch = yearRegex.exec(text)\n\tconst year = yearMatch ? Number.parseInt(yearMatch[1], 10) : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.9 (Statutes at Large format is standardized)\n\tconst confidence = 0.9\n\n\treturn {\n\t\ttype: 'statutesAtLarge',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tvolume,\n\t\tpage,\n\t\tyear,\n\t}\n}\n","/**\n * Short-form Citation Extraction\n *\n * Parses tokenized short-form citations (Id., supra, short-form case) to extract\n * metadata. Short-form citations refer to earlier citations in the document.\n *\n * @module extract/extractShortForms\n */\n\nimport type { Token } from '@/tokenize'\nimport type { IdCitation, SupraCitation, ShortFormCaseCitation } from '@/types/citation'\nimport type { TransformationMap } from '@/types/span'\n\n/**\n * Extracts Id. citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Pincite: Optional page reference (e.g., \"253\" from \"Id. at 253\")\n *\n * Confidence scoring:\n * - 1.0 (Id. format is unambiguous and standardized)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns IdCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"Id. at 253\",\n *   span: { cleanStart: 10, cleanEnd: 20 },\n *   type: \"case\",\n *   patternId: \"id\"\n * }\n * const citation = extractId(token, transformationMap)\n * // citation = {\n * //   type: \"id\",\n * //   pincite: 253,\n * //   confidence: 1.0,\n * //   ...\n * // }\n * ```\n */\nexport function extractId(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): IdCitation {\n\tconst { text, span } = token\n\n\t// Parse Id. with optional pincite\n\t// Pattern: Id. or Ibid. with optional \"at [page]\"\n\tconst idRegex = /[Ii](?:d|bid)\\.(?:\\s+at\\s+(\\d+))?/\n\tconst match = idRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse Id. citation: ${text}`)\n\t}\n\n\t// Extract pincite if present\n\tconst pincite = match[1] ? Number.parseInt(match[1], 10) : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 1.0 (Id. format is unambiguous)\n\tconst confidence = 1.0\n\n\treturn {\n\t\ttype: 'id',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tpincite,\n\t}\n}\n\n/**\n * Extracts supra citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Party name: Name preceding \"supra\" (e.g., \"Smith\" from \"Smith, supra\")\n * - Pincite: Optional page reference (e.g., \"460\" from \"Smith, supra, at 460\")\n *\n * Confidence scoring:\n * - 0.9 (supra format is fairly standard but party name extraction can vary)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns SupraCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"Smith, supra, at 460\",\n *   span: { cleanStart: 10, cleanEnd: 30 },\n *   type: \"case\",\n *   patternId: \"supra\"\n * }\n * const citation = extractSupra(token, transformationMap)\n * // citation = {\n * //   type: \"supra\",\n * //   partyName: \"Smith\",\n * //   pincite: 460,\n * //   confidence: 0.9,\n * //   ...\n * // }\n * ```\n */\nexport function extractSupra(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): SupraCitation {\n\tconst { text, span } = token\n\n\t// Parse party name and optional pincite\n\t// Pattern: word(s), supra [, at page]\n\t// Note: Matches party names including \"v.\" (e.g., \"Smith v. Jones\")\n\tconst supraRegex = /\\b([A-Z][a-zA-Z]+(?:(?:\\s+v\\.?\\s+|\\s+)[A-Z][a-zA-Z]+)*)\\s*,?\\s+supra(?:,?\\s+at\\s+(\\d+))?/\n\tconst match = supraRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse supra citation: ${text}`)\n\t}\n\n\tconst partyName = match[1]\n\tconst pincite = match[2] ? Number.parseInt(match[2], 10) : undefined\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.9 (supra format is fairly standard)\n\tconst confidence = 0.9\n\n\treturn {\n\t\ttype: 'supra',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tpartyName,\n\t\tpincite,\n\t}\n}\n\n/**\n * Extracts short-form case citation metadata from a tokenized citation.\n *\n * Parses token text to extract:\n * - Volume: Volume number\n * - Reporter: Reporter abbreviation\n * - Pincite: Page reference (from \"at [page]\" pattern)\n *\n * Confidence scoring:\n * - 0.7 (short-form case citations are more ambiguous than full citations)\n *\n * @param token - Token from tokenizer containing matched text and clean positions\n * @param transformationMap - Position mapping from clean → original text\n * @returns ShortFormCaseCitation with parsed metadata and translated positions\n *\n * @example\n * ```typescript\n * const token = {\n *   text: \"500 F.2d at 125\",\n *   span: { cleanStart: 10, cleanEnd: 25 },\n *   type: \"case\",\n *   patternId: \"short-form-case\"\n * }\n * const citation = extractShortFormCase(token, transformationMap)\n * // citation = {\n * //   type: \"shortFormCase\",\n * //   volume: 500,\n * //   reporter: \"F.2d\",\n * //   pincite: 125,\n * //   confidence: 0.7,\n * //   ...\n * // }\n * ```\n */\nexport function extractShortFormCase(\n\ttoken: Token,\n\ttransformationMap: TransformationMap,\n): ShortFormCaseCitation {\n\tconst { text, span } = token\n\n\t// Parse volume-reporter-at-page\n\t// Pattern: number space abbreviation space \"at\" space number\n\tconst shortFormRegex = /(\\d+(?:-\\d+)?)\\s+([A-Z][A-Za-z.\\s]+?(?:\\d[a-z])?)\\s+at\\s+(\\d+)/\n\tconst match = shortFormRegex.exec(text)\n\n\tif (!match) {\n\t\tthrow new Error(`Failed to parse short-form case citation: ${text}`)\n\t}\n\n\tconst rawVolume = match[1]\n\tconst volume = /^\\d+$/.test(rawVolume) ? Number.parseInt(rawVolume, 10) : rawVolume\n\tconst reporter = match[2].trim() // Remove trailing spaces\n\tconst pincite = Number.parseInt(match[3], 10)\n\n\t// Translate positions from clean → original\n\tconst originalStart =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanStart) ?? span.cleanStart\n\tconst originalEnd =\n\t\ttransformationMap.cleanToOriginal.get(span.cleanEnd) ?? span.cleanEnd\n\n\t// Confidence: 0.7 (short-form citations are more ambiguous)\n\tconst confidence = 0.7\n\n\treturn {\n\t\ttype: 'shortFormCase',\n\t\ttext,\n\t\tspan: {\n\t\t\tcleanStart: span.cleanStart,\n\t\t\tcleanEnd: span.cleanEnd,\n\t\t\toriginalStart,\n\t\t\toriginalEnd,\n\t\t},\n\t\tconfidence,\n\t\tmatchedText: text,\n\t\tprocessTimeMs: 0,\n\t\tpatternsChecked: 1,\n\t\tvolume,\n\t\treporter,\n\t\tpincite,\n\t}\n}\n","/**\n * Scope Boundary Detection\n *\n * Detects paragraph/section boundaries in text and validates whether\n * an antecedent citation is within the resolution scope.\n */\n\nimport type { Citation } from '../types/citation'\nimport type { ScopeStrategy } from './types'\n\n/**\n * Detects paragraph boundaries from text and assigns each citation to a paragraph.\n *\n * @param text - Original document text\n * @param citations - Extracted citations with position spans\n * @param boundaryPattern - Regex pattern to detect boundaries (default: /\\n\\n+/)\n * @returns Map of citation index to paragraph number (0-based)\n */\nexport function detectParagraphBoundaries(\n  text: string,\n  citations: Citation[],\n  boundaryPattern: RegExp = /\\n\\n+/g\n): Map<number, number> {\n  const paragraphMap = new Map<number, number>()\n\n  // Find all paragraph boundaries (positions in text)\n  const boundaries: number[] = [0] // Start of document is first boundary\n  let match: RegExpExecArray | null\n\n  while ((match = boundaryPattern.exec(text)) !== null) {\n    // Boundary is at end of match (start of next paragraph)\n    boundaries.push(match.index + match[0].length)\n  }\n\n  boundaries.push(text.length) // End of document\n\n  // Assign each citation to a paragraph\n  for (let i = 0; i < citations.length; i++) {\n    const citation = citations[i]\n    const citationStart = citation.span.originalStart\n\n    // Find which paragraph this citation belongs to\n    let paragraphNum = 0\n    for (let j = 0; j < boundaries.length - 1; j++) {\n      if (citationStart >= boundaries[j] && citationStart < boundaries[j + 1]) {\n        paragraphNum = j\n        break\n      }\n    }\n\n    paragraphMap.set(i, paragraphNum)\n  }\n\n  return paragraphMap\n}\n\n/**\n * Checks if an antecedent citation is within resolution scope.\n *\n * @param antecedentIndex - Index of the antecedent citation\n * @param currentIndex - Index of current citation being resolved\n * @param paragraphMap - Map of citation index to paragraph number\n * @param strategy - Scope boundary strategy\n * @returns true if antecedent is within scope, false otherwise\n */\nexport function isWithinBoundary(\n  antecedentIndex: number,\n  currentIndex: number,\n  paragraphMap: Map<number, number>,\n  strategy: ScopeStrategy\n): boolean {\n  if (strategy === 'none') {\n    // No boundary restriction - can resolve across entire document\n    return true\n  }\n\n  // Get paragraph numbers for both citations\n  const antecedentParagraph = paragraphMap.get(antecedentIndex)\n  const currentParagraph = paragraphMap.get(currentIndex)\n\n  // If either is undefined, default to allowing resolution\n  if (antecedentParagraph === undefined || currentParagraph === undefined) {\n    return true\n  }\n\n  // For paragraph/section/footnote strategies, citations must be in same boundary\n  // (In this MVP, section and footnote behave same as paragraph - future enhancement)\n  return antecedentParagraph === currentParagraph\n}\n","/**\n * Levenshtein Distance\n *\n * Calculates edit distance between strings for fuzzy party name matching\n * in supra citation resolution.\n *\n * Uses dynamic programming for O(m*n) time complexity.\n */\n\n/**\n * Calculates Levenshtein distance (edit distance) between two strings.\n *\n * The edit distance is the minimum number of single-character edits\n * (insertions, deletions, substitutions) needed to change one string into the other.\n *\n * @param a - First string\n * @param b - Second string\n * @returns Number of edits required (0 = identical)\n */\nexport function levenshteinDistance(a: string, b: string): number {\n  // Handle empty strings\n  if (a.length === 0) return b.length\n  if (b.length === 0) return a.length\n\n  // Create 2D array for dynamic programming\n  // dp[i][j] = edit distance between a[0...i-1] and b[0...j-1]\n  const dp: number[][] = Array.from({ length: a.length + 1 }, () =>\n    Array(b.length + 1).fill(0)\n  )\n\n  // Initialize base cases\n  for (let i = 0; i <= a.length; i++) {\n    dp[i][0] = i // Distance from a[0...i-1] to empty string\n  }\n  for (let j = 0; j <= b.length; j++) {\n    dp[0][j] = j // Distance from empty string to b[0...j-1]\n  }\n\n  // Fill the DP table\n  for (let i = 1; i <= a.length; i++) {\n    for (let j = 1; j <= b.length; j++) {\n      if (a[i - 1] === b[j - 1]) {\n        // Characters match - no edit needed\n        dp[i][j] = dp[i - 1][j - 1]\n      } else {\n        // Characters differ - take minimum of:\n        // 1. Insert: dp[i][j-1] + 1\n        // 2. Delete: dp[i-1][j] + 1\n        // 3. Substitute: dp[i-1][j-1] + 1\n        dp[i][j] = 1 + Math.min(\n          dp[i - 1][j],     // Delete from a\n          dp[i][j - 1],     // Insert into a\n          dp[i - 1][j - 1]  // Substitute\n        )\n      }\n    }\n  }\n\n  return dp[a.length][b.length]\n}\n\n/**\n * Calculates normalized Levenshtein similarity (0-1 scale).\n *\n * Returns similarity score where:\n * - 1.0 = identical strings\n * - 0.0 = completely different\n *\n * Comparison is case-insensitive.\n *\n * @param a - First string\n * @param b - Second string\n * @returns Similarity score from 0 to 1\n */\nexport function normalizedLevenshteinDistance(a: string, b: string): number {\n  // Normalize to lowercase for case-insensitive comparison\n  const lowerA = a.toLowerCase()\n  const lowerB = b.toLowerCase()\n\n  // Calculate raw edit distance\n  const distance = levenshteinDistance(lowerA, lowerB)\n\n  // Normalize by max length\n  const maxLength = Math.max(lowerA.length, lowerB.length)\n  if (maxLength === 0) return 1.0 // Both empty strings\n\n  // Convert distance to similarity: 1 - (distance / maxLength)\n  return 1 - distance / maxLength\n}\n","/**\n * Document-Scoped Citation Resolver\n *\n * Resolves short-form citations (Id./supra/short-form case) to their full antecedent citations\n * by maintaining resolution context and enforcing scope boundaries.\n *\n * Resolution rules:\n * - Id. resolves to immediately preceding full citation (within scope)\n * - Supra resolves to full citation with matching party name (within scope)\n * - Short-form case resolves to full case with matching volume/reporter (within scope)\n */\n\nimport type {\n  Citation,\n  FullCaseCitation,\n  IdCitation,\n  SupraCitation,\n  ShortFormCaseCitation,\n} from '../types/citation'\nimport { isFullCitation } from '../types/guards'\nimport type {\n  ResolutionOptions,\n  ResolutionResult,\n  ResolvedCitation,\n  ResolutionContext,\n} from './types'\nimport { detectParagraphBoundaries, isWithinBoundary } from './scopeBoundary'\nimport { normalizedLevenshteinDistance } from './levenshtein'\n\n/**\n * Document-scoped resolver that processes citations sequentially\n * and resolves short-form citations to their antecedents.\n */\nexport class DocumentResolver {\n  private readonly citations: Citation[]\n  private readonly text: string\n  private readonly options: Required<ResolutionOptions>\n  private readonly context: ResolutionContext\n\n  /**\n   * Creates a new DocumentResolver.\n   *\n   * @param citations - All citations in document (in order of appearance)\n   * @param text - Original document text\n   * @param options - Resolution options\n   */\n  constructor(\n    citations: Citation[],\n    text: string,\n    options: ResolutionOptions = {}\n  ) {\n    this.citations = citations\n    this.text = text\n\n    // Apply defaults to options\n    this.options = {\n      scopeStrategy: options.scopeStrategy ?? 'paragraph',\n      autoDetectParagraphs: options.autoDetectParagraphs ?? true,\n      paragraphBoundaryPattern: options.paragraphBoundaryPattern ?? /\\n\\n+/g,\n      fuzzyPartyMatching: options.fuzzyPartyMatching ?? true,\n      partyMatchThreshold: options.partyMatchThreshold ?? 0.8,\n      allowNestedResolution: options.allowNestedResolution ?? false,\n      reportUnresolved: options.reportUnresolved ?? true,\n    }\n\n    // Initialize resolution context\n    this.context = {\n      citationIndex: 0,\n      allCitations: citations,\n      lastFullCitation: undefined,\n      fullCitationHistory: new Map(),\n      paragraphMap: new Map(),\n    }\n\n    // Detect paragraph boundaries if enabled\n    if (this.options.autoDetectParagraphs) {\n      this.context.paragraphMap = detectParagraphBoundaries(\n        text,\n        citations,\n        this.options.paragraphBoundaryPattern\n      )\n    }\n  }\n\n  /**\n   * Resolves all citations in the document.\n   *\n   * @returns Array of citations with resolution metadata\n   */\n  resolve(): ResolvedCitation[] {\n    const resolved: ResolvedCitation[] = []\n\n    for (let i = 0; i < this.citations.length; i++) {\n      this.context.citationIndex = i\n      const citation = this.citations[i]\n\n      // Resolve based on citation type\n      let resolution: ResolutionResult | undefined\n\n      switch (citation.type) {\n        case 'id':\n          resolution = this.resolveId(citation)\n          break\n        case 'supra':\n          resolution = this.resolveSupra(citation)\n          break\n        case 'shortFormCase':\n          resolution = this.resolveShortFormCase(citation)\n          break\n        default:\n          // Full citation - update context for future resolutions\n          if (isFullCitation(citation)) {\n            this.context.lastFullCitation = i\n            this.trackFullCitation(citation, i)\n          }\n          break\n      }\n\n      // Add citation with resolution metadata\n      // Type assertion is safe: runtime logic only sets resolution on short-form citations\n      resolved.push({\n        ...citation,\n        resolution,\n      } as ResolvedCitation)\n    }\n\n    return resolved\n  }\n\n  /**\n   * Resolves Id. citation to immediately preceding full case citation.\n   */\n  private resolveId(_citation: IdCitation): ResolutionResult | undefined {\n    const currentIndex = this.context.citationIndex\n\n    // Find most recent full case citation (Id. only resolves to case citations, not statutes/journals)\n    let antecedentIndex: number | undefined\n    for (let i = currentIndex - 1; i >= 0; i--) {\n      const candidate = this.citations[i]\n      if (candidate.type === 'case') {\n        antecedentIndex = i\n        break\n      }\n    }\n\n    // Check if we have a previous case citation\n    if (antecedentIndex === undefined) {\n      return this.createFailureResult('No preceding full case citation found')\n    }\n\n    // Check scope boundary\n    if (!this.isWithinScope(antecedentIndex, currentIndex)) {\n      return this.createFailureResult('Antecedent citation outside scope boundary')\n    }\n\n    return {\n      resolvedTo: antecedentIndex,\n      confidence: 1.0, // Id. resolution is unambiguous when successful\n    }\n  }\n\n  /**\n   * Resolves supra citation by matching party name.\n   */\n  private resolveSupra(citation: SupraCitation): ResolutionResult | undefined {\n    const currentIndex = this.context.citationIndex\n    const targetPartyName = this.normalizePartyName(citation.partyName)\n\n    // Search full citation history for matching party name\n    let bestMatch: { index: number; similarity: number } | undefined\n\n    for (const [partyName, citationIndex] of this.context.fullCitationHistory) {\n      // Check scope boundary\n      if (!this.isWithinScope(citationIndex, currentIndex)) {\n        continue\n      }\n\n      // Calculate similarity\n      const similarity = normalizedLevenshteinDistance(targetPartyName, partyName)\n\n      // Update best match if this is better\n      if (!bestMatch || similarity > bestMatch.similarity) {\n        bestMatch = { index: citationIndex, similarity }\n      }\n    }\n\n    // Check if we found a match above threshold\n    if (!bestMatch) {\n      return this.createFailureResult('No full citation found in scope')\n    }\n\n    if (bestMatch.similarity < this.options.partyMatchThreshold) {\n      return this.createFailureResult(\n        `Party name similarity ${bestMatch.similarity.toFixed(2)} below threshold ${this.options.partyMatchThreshold}`\n      )\n    }\n\n    // Return successful resolution with confidence based on similarity\n    const warnings: string[] = []\n    if (bestMatch.similarity < 1.0) {\n      warnings.push(`Fuzzy match: similarity ${bestMatch.similarity.toFixed(2)}`)\n    }\n\n    return {\n      resolvedTo: bestMatch.index,\n      confidence: bestMatch.similarity,\n      warnings: warnings.length > 0 ? warnings : undefined,\n    }\n  }\n\n  /**\n   * Resolves short-form case citation by matching volume/reporter.\n   */\n  private resolveShortFormCase(citation: ShortFormCaseCitation): ResolutionResult | undefined {\n    const currentIndex = this.context.citationIndex\n\n    // Search backwards for matching full case citation\n    for (let i = currentIndex - 1; i >= 0; i--) {\n      const candidate = this.citations[i]\n\n      // Only match against full case citations\n      if (candidate.type !== 'case') {\n        continue\n      }\n\n      // Check if volume and reporter match\n      if (\n        candidate.volume === citation.volume &&\n        this.normalizeReporter(candidate.reporter) === this.normalizeReporter(citation.reporter)\n      ) {\n        // Check scope boundary\n        if (!this.isWithinScope(i, currentIndex)) {\n          return this.createFailureResult('Matching citation outside scope boundary')\n        }\n\n        // Found a match\n        return {\n          resolvedTo: i,\n          confidence: 0.95, // High confidence but not perfect (multiple cases could have same volume/reporter)\n        }\n      }\n    }\n\n    return this.createFailureResult('No matching full case citation found')\n  }\n\n  /**\n   * Tracks a full citation in the resolution history.\n   * Extracts party name for supra resolution.\n   * Uses extracted party names (Phase 7) when available, falls back to backward search.\n   */\n  private trackFullCitation(citation: Citation, index: number): void {\n    // Only case citations have party names for supra resolution\n    if (citation.type === 'case') {\n      // Phase 7: Use extracted party names when available\n      // Defendant name stored first (preferred for Bluebook-style supra matching)\n      if (citation.defendantNormalized) {\n        this.context.fullCitationHistory.set(citation.defendantNormalized, index)\n      }\n      if (citation.plaintiffNormalized) {\n        this.context.fullCitationHistory.set(citation.plaintiffNormalized, index)\n      }\n\n      // Fallback: backward search from text (pre-Phase 7 compatibility)\n      if (!citation.plaintiffNormalized && !citation.defendantNormalized) {\n        const partyName = this.extractPartyName(citation)\n        if (partyName) {\n          const normalized = this.normalizePartyName(partyName)\n          this.context.fullCitationHistory.set(normalized, index)\n        }\n      }\n    }\n  }\n\n  /**\n   * Extracts party name from full case citation text.\n   * Handles \"Party v. Party\" format by looking at text before citation span.\n   */\n  private extractPartyName(citation: FullCaseCitation): string | undefined {\n    // Look at text before citation span to find party names\n    // Case citations typically appear as: \"Smith v. Jones, 100 F.2d 10\"\n    // But tokenizer only captures \"100 F.2d 10\" - we need to look backwards in text\n\n    const citationStart = citation.span.originalStart\n    // Look backwards up to 100 characters for party name\n    const lookbackStart = Math.max(0, citationStart - 100)\n    const beforeText = this.text.substring(lookbackStart, citationStart)\n\n    // Match pattern: \"FirstParty v. SecondParty, \" before the citation\n    // Capture the first party name (handles single-letter party names like \"A\" or \"B\")\n    const vMatch = beforeText.match(/([A-Z][a-zA-Z]*(?:\\s+[A-Z][a-zA-Z]*)*)\\s+v\\.?\\s+[A-Z][a-zA-Z]*(?:\\s+[A-Z][a-zA-Z]*)*,\\s*$/)\n    if (vMatch) {\n      return this.stripSignalWords(vMatch[1].trim())\n    }\n\n    // Fallback: try to find any capitalized word(s) before comma\n    const beforeComma = beforeText.match(/([A-Z][a-zA-Z]*(?:\\s+[A-Z][a-zA-Z]*)*),\\s*$/)\n    if (beforeComma) {\n      return this.stripSignalWords(beforeComma[1].trim())\n    }\n    return undefined\n  }\n\n  /**\n   * Strips citation signal words that may precede party names.\n   * E.g., \"In Smith\" → \"Smith\", \"See Also Jones\" → \"Jones\"\n   * Preserves \"In re\" which is a case name format, not a signal word.\n   */\n  private stripSignalWords(name: string): string {\n    const stripped = name.replace(/^(?:In(?!\\s+re\\b)|See(?:\\s+[Aa]lso)?|Compare|But(?:\\s+[Ss]ee)?|Cf\\.?|Also)\\s+/i, '').trim()\n    // Only return stripped version if something remains\n    return stripped.length > 0 ? stripped : name\n  }\n\n  /**\n   * Normalizes party name for matching.\n   */\n  private normalizePartyName(name: string): string {\n    return name\n      .toLowerCase()\n      .replace(/\\s+/g, ' ') // Normalize whitespace\n      .trim()\n  }\n\n  /**\n   * Normalizes reporter abbreviation for matching.\n   */\n  private normalizeReporter(reporter: string): string {\n    return reporter\n      .toLowerCase()\n      .replace(/\\s+/g, '') // Remove spaces (F.2d vs F. 2d)\n      .replace(/\\./g, '')   // Remove periods\n  }\n\n  /**\n   * Checks if antecedent citation is within scope boundary.\n   */\n  private isWithinScope(antecedentIndex: number, currentIndex: number): boolean {\n    return isWithinBoundary(\n      antecedentIndex,\n      currentIndex,\n      this.context.paragraphMap,\n      this.options.scopeStrategy\n    )\n  }\n\n  /**\n   * Creates a failure result for unresolved citations.\n   */\n  private createFailureResult(reason: string): ResolutionResult | undefined {\n    if (this.options.reportUnresolved) {\n      return {\n        resolvedTo: undefined,\n        failureReason: reason,\n        confidence: 0.0,\n      }\n    }\n    return undefined\n  }\n}\n","/**\n * Citation Resolution\n *\n * Resolves short-form citations (Id./supra/short-form case) to their full antecedents.\n *\n * @example\n * ```ts\n * import { resolveCitations } from 'eyecite-ts/resolve'\n * import { extractCitations } from 'eyecite-ts'\n *\n * const text = 'See Smith v. Jones, 500 F.2d 100 (1974). Id. at 105.'\n * const citations = extractCitations(text)\n * const resolved = resolveCitations(citations, text)\n *\n * // resolved[1] is Id. citation with resolution.resolvedTo = 0\n * console.log(resolved[1].resolution?.resolvedTo) // 0 (points to Smith v. Jones)\n * ```\n */\n\nimport type { Citation } from '../types/citation'\nimport type { ResolutionOptions, ResolvedCitation } from './types'\nimport { DocumentResolver } from './DocumentResolver'\n\n/**\n * Resolves short-form citations to their full antecedents.\n *\n * Convenience wrapper around DocumentResolver that handles common use cases.\n *\n * @param citations - Extracted citations in order of appearance\n * @param text - Original document text\n * @param options - Resolution options\n * @returns Citations with resolution metadata\n */\nexport function resolveCitations(\n  citations: Citation[],\n  text: string,\n  options?: ResolutionOptions\n): ResolvedCitation[] {\n  const resolver = new DocumentResolver(citations, text, options)\n  return resolver.resolve()\n}\n\n// Re-export core types and classes\nexport { DocumentResolver } from './DocumentResolver'\nexport type {\n  ResolutionOptions,\n  ResolutionResult,\n  ResolvedCitation,\n  ScopeStrategy,\n} from './types'\n","/**\n * Parallel Citation Detection\n *\n * Detects parallel citation groups (same case in multiple reporters) using\n * comma-separated case citations sharing a closing parenthetical.\n *\n * Detection happens after tokenization and deduplication, before extraction\n * in the main extractCitations pipeline.\n *\n * @module extract/detectParallel\n */\n\nimport type { Token } from '@/tokenize/tokenizer'\n\n/**\n * Maximum characters allowed between end of comma and start of next citation.\n * Bluebook standard uses tight spacing: \"500 F.2d 123, 200 F. Supp. 456\"\n */\nconst MAX_PROXIMITY = 5\n\n/**\n * Maximum total gap (chars) between end of one citation and start of next\n * to even consider them as parallel candidates. Beyond this distance, we can\n * skip all other checks (comma, parenthetical, etc.) for performance.\n * Includes comma, spaces, and potential pincite: \", 125, \" = ~10 chars\n */\nconst MAX_GAP_FOR_PARALLEL = 20\n\n/**\n * Detect parallel citation groups from tokenized citations.\n *\n * Returns a map of primary citation index to array of secondary citation indices.\n * Parallel citations are comma-separated case citations sharing a parenthetical.\n *\n * Detection algorithm:\n * 1. Iterate tokens with lookahead (i, i+1, i+2...)\n * 2. Check if token[i] and token[i+1] are both case citations\n * 3. Check if comma separates them (within MAX_PROXIMITY chars)\n * 4. Check if both citations share a closing parenthetical (via cleaned text)\n * 5. If all conditions met, add to parallel group\n * 6. Continue for chain (i+1, i+2, i+3...) until no more matches\n *\n * @param tokens - Tokenized citations (after deduplication)\n * @param cleanedText - Cleaned text to check for commas and parentheticals\n * @returns Map of primary index to array of secondary indices\n *\n * @example\n * ```typescript\n * const tokens = [\n *   { text: \"410 U.S. 113\", span: { cleanStart: 0, cleanEnd: 12 }, type: \"case\" },\n *   { text: \"93 S. Ct. 705\", span: { cleanStart: 14, cleanEnd: 27 }, type: \"case\" }\n * ]\n * const cleaned = \"410 U.S. 113, 93 S. Ct. 705 (1973)\"\n * const result = detectParallelCitations(tokens, cleaned)\n * // result = Map { 0 => [1] }\n * ```\n */\nexport function detectParallelCitations(\n\ttokens: Token[],\n\tcleanedText = '',\n): Map<number, number[]> {\n\tconst parallelGroups = new Map<number, number[]>()\n\n\t// Edge cases: empty array or no text\n\tif (tokens.length === 0 || cleanedText === '') {\n\t\treturn parallelGroups\n\t}\n\n\t// Track which tokens are already in a parallel group (as secondary)\n\tconst usedAsSecondary = new Set<number>()\n\n\tfor (let i = 0; i < tokens.length; i++) {\n\t\tconst primary = tokens[i]\n\n\t\t// Skip if not a case citation\n\t\tif (primary.type !== 'case') {\n\t\t\tcontinue\n\t\t}\n\n\t\t// Skip if already used as secondary in another group\n\t\tif (usedAsSecondary.has(i)) {\n\t\t\tcontinue\n\t\t}\n\n\t\tconst secondaryIndices: number[] = []\n\n\t\t// Look ahead for potential secondary citations\n\t\t// Chain detection: \"A, B, C (year)\" where A is primary, B and C are secondaries\n\t\tfor (let j = i + 1; j < tokens.length; j++) {\n\t\t\tconst secondary = tokens[j]\n\n\t\t\t// Only case citations can be parallel\n\t\t\tif (secondary.type !== 'case') {\n\t\t\t\tbreak // Stop looking once we hit non-case citation\n\t\t\t}\n\n\t\t\t// Check proximity: comma should be right after primary (or previous secondary in chain)\n\t\t\tconst prevToken = j === i + 1 ? primary : tokens[j - 1]\n\t\t\tconst gapStart = prevToken.span.cleanEnd\n\t\t\tconst gapEnd = secondary.span.cleanStart\n\n\t\t\t// Early exit: If gap is too large, no need to check comma/parenthetical\n\t\t\t// This optimization reduces O(n²) to O(n×k) where k is avg tokens within MAX_GAP\n\t\t\tconst gapSize = gapEnd - gapStart\n\t\t\tif (gapSize > MAX_GAP_FOR_PARALLEL) {\n\t\t\t\tbreak // Too far apart to be parallel, stop looking\n\t\t\t}\n\n\t\t\t// Extract the gap text between citations\n\t\t\tconst gapText = cleanedText.substring(gapStart, gapEnd)\n\n\t\t\t// Bluebook requires comma separator for parallel citations\n\t\t\tif (!gapText.includes(',')) {\n\t\t\t\tbreak // No comma = not parallel, stop looking\n\t\t\t}\n\n\t\t\t// Check proximity: distance from comma to next citation start\n\t\t\t// MAX_PROXIMITY enforces tight spacing: \"A, B\" not \"A,      B\"\n\t\t\tconst commaIndex = gapText.indexOf(',')\n\t\t\tconst distanceAfterComma = gapText.length - commaIndex - 1\n\n\t\t\tif (distanceAfterComma > MAX_PROXIMITY) {\n\t\t\t\tbreak // Too far apart, stop looking\n\t\t\t}\n\n\t\t\t// Check for shared parenthetical\n\t\t\t// Both citations must share the SAME closing parenthetical\n\t\t\t// Reject: \"A (1970), B (1971)\" - separate parens = different cases\n\t\t\t// Accept: \"A, B (1970)\" - shared paren = parallel citations\n\t\t\tconst textBetween = cleanedText.substring(primary.span.cleanEnd, secondary.span.cleanEnd)\n\t\t\tif (textBetween.includes(')')) {\n\t\t\t\tbreak // Separate parentheticals = not parallel, stop looking\n\t\t\t}\n\n\t\t\t// Check that there IS a parenthetical after the secondary citation\n\t\t\tif (!hasSharedParenthetical(cleanedText, secondary.span.cleanEnd)) {\n\t\t\t\tbreak // No shared parenthetical, stop looking\n\t\t\t}\n\n\t\t\t// All conditions met - this is a parallel citation\n\t\t\tsecondaryIndices.push(j)\n\t\t\tusedAsSecondary.add(j)\n\t\t}\n\n\t\t// If we found any secondary citations, record the group\n\t\tif (secondaryIndices.length > 0) {\n\t\t\tparallelGroups.set(i, secondaryIndices)\n\t\t}\n\t}\n\n\treturn parallelGroups\n}\n\n/**\n * Check if there's a closing parenthetical after the given position.\n *\n * This is a simple heuristic: look for \"(...)\" pattern within reasonable distance.\n * Full parenthetical parsing happens in extractCase, this just validates presence.\n *\n * @param cleanedText - Cleaned text\n * @param position - Position to start searching from\n * @returns true if closing parenthetical found\n */\nfunction hasSharedParenthetical(cleanedText: string, position: number): boolean {\n\t// Look ahead up to 200 characters for opening parenthesis\n\tconst searchText = cleanedText.substring(position, position + 200)\n\n\t// Find opening parenthesis\n\tconst openIndex = searchText.indexOf('(')\n\tif (openIndex === -1) {\n\t\treturn false\n\t}\n\n\t// Find matching closing parenthesis (simple depth tracking)\n\tlet depth = 0\n\tfor (let i = openIndex; i < searchText.length; i++) {\n\t\tif (searchText[i] === '(') {\n\t\t\tdepth++\n\t\t} else if (searchText[i] === ')') {\n\t\t\tdepth--\n\t\t\tif (depth === 0) {\n\t\t\t\t// Found matching closing parenthesis\n\t\t\t\treturn true\n\t\t\t}\n\t\t}\n\t}\n\n\treturn false\n}\n","/**\n * Main Citation Extraction Pipeline\n *\n * Orchestrates the complete citation extraction flow:\n *   1. Clean text (remove HTML, normalize Unicode)\n *   2. Tokenize (apply patterns to find candidates)\n *   3. Extract (parse metadata from tokens)\n *\n * This is the primary public API for citation extraction.\n *\n * @module extract/extractCitations\n */\n\nimport { cleanText } from '@/clean'\nimport { tokenize } from '@/tokenize'\nimport {\n\textractCase,\n\textractStatute,\n\textractJournal,\n\textractNeutral,\n\textractPublicLaw,\n\textractFederalRegister,\n\textractStatutesAtLarge,\n} from '@/extract'\nimport { extractId, extractSupra, extractShortFormCase } from './extractShortForms'\nimport {\n\tcasePatterns,\n\tstatutePatterns,\n\tjournalPatterns,\n\tneutralPatterns,\n\tshortFormPatterns,\n} from '@/patterns'\nimport { resolveCitations } from '../resolve'\nimport { detectParallelCitations } from './detectParallel'\nimport type { Citation } from '@/types/citation'\nimport type { Pattern } from '@/patterns'\nimport type { ResolutionOptions, ResolvedCitation } from '../resolve/types'\n\n/**\n * Options for customizing citation extraction behavior.\n */\nexport interface ExtractOptions {\n\t/**\n\t * Custom text cleaners (overrides defaults).\n\t *\n\t * If provided, these cleaners replace the default pipeline:\n\t * [stripHtmlTags, normalizeWhitespace, normalizeUnicode, fixSmartQuotes]\n\t *\n\t * @example\n\t * ```typescript\n\t * // Use only HTML stripping, skip Unicode normalization\n\t * const citations = extractCitations(text, {\n\t *   cleaners: [stripHtmlTags]\n\t * })\n\t * ```\n\t */\n\tcleaners?: Array<(text: string) => string>\n\n\t/**\n\t * Custom regex patterns (overrides defaults).\n\t *\n\t * If provided, these patterns replace the default pattern set:\n\t * [casePatterns, statutePatterns, journalPatterns, neutralPatterns, shortFormPatterns]\n\t *\n\t * @example\n\t * ```typescript\n\t * // Extract only case citations\n\t * const citations = extractCitations(text, {\n\t *   patterns: casePatterns\n\t * })\n\t * ```\n\t */\n\tpatterns?: Pattern[]\n\n\t/**\n\t * Resolve short-form citations to their full antecedents (default: false).\n\t *\n\t * If true, returns ResolvedCitation[] with resolution metadata for short-form citations\n\t * (Id., supra, short-form case). Full citations are unchanged.\n\t *\n\t * @example\n\t * ```typescript\n\t * const text = \"Smith v. Jones, 500 F.2d 100 (1974). Id. at 105.\"\n\t * const citations = extractCitations(text, { resolve: true })\n\t * // citations[1].resolution.resolvedTo === 0 (points to Smith v. Jones)\n\t * ```\n\t */\n\tresolve?: boolean\n\n\t/**\n\t * Options for citation resolution (only used if resolve: true).\n\t *\n\t * @example\n\t * ```typescript\n\t * const citations = extractCitations(text, {\n\t *   resolve: true,\n\t *   resolutionOptions: {\n\t *     scopeStrategy: 'paragraph',\n\t *     fuzzyPartyMatching: true\n\t *   }\n\t * })\n\t * ```\n\t */\n\tresolutionOptions?: ResolutionOptions\n}\n\n/**\n * Extracts legal citations from text using the full parsing pipeline.\n *\n * Pipeline flow:\n * 1. **Clean:** Remove HTML tags, normalize Unicode, fix smart quotes\n * 2. **Tokenize:** Apply regex patterns to find citation candidates\n * 3. **Extract:** Parse metadata (volume, reporter, page, etc.)\n * 4. **Translate:** Map positions from cleaned text back to original text\n *\n * This function is synchronous because all stages (cleaning, tokenization,\n * extraction) are synchronous. For async operations (e.g., future reporters-db\n * lookups), use extractCitationsAsync().\n *\n * Position tracking:\n * - TransformationMap is built during cleaning\n * - Tokens contain positions in cleaned text (cleanStart/cleanEnd)\n * - Extraction translates cleaned positions → original positions\n * - Final citations have originalStart/originalEnd pointing to input text\n *\n * Warnings from cleaning layer are attached to all extracted citations.\n *\n * @param text - Raw text to extract citations from (may contain HTML, Unicode)\n * @param options - Optional customization (cleaners, patterns)\n * @returns Array of citations with parsed metadata and accurate positions\n *\n * @example\n * ```typescript\n * const text = \"See Smith v. Doe, 500 F.2d 123 (9th Cir. 2020)\"\n * const citations = extractCitations(text)\n * // citations[0] = {\n * //   type: \"case\",\n * //   volume: 500,\n * //   reporter: \"F.2d\",\n * //   page: 123,\n * //   court: \"9th Cir.\",\n * //   year: 2020,\n * //   span: { originalStart: 18, originalEnd: 30, ... }\n * // }\n * ```\n *\n * @example\n * ```typescript\n * // Extract from HTML\n * const html = \"<p>In <b>Smith</b>, 500 F.2d 123, the court held...</p>\"\n * const citations = extractCitations(html)\n * // HTML is stripped, positions point to original HTML\n * ```\n *\n * @example\n * ```typescript\n * // Extract multiple citation types\n * const text = \"See 42 U.S.C. § 1983; Smith, 500 F.2d 123; 123 Harv. L. Rev. 456\"\n * const citations = extractCitations(text)\n * // citations[0].type === \"statute\"\n * // citations[1].type === \"case\"\n * // citations[2].type === \"journal\"\n * ```\n */\nexport function extractCitations(text: string, options: ExtractOptions & { resolve: true }): ResolvedCitation[]\nexport function extractCitations(text: string, options?: ExtractOptions): Citation[]\nexport function extractCitations(\n\ttext: string,\n\toptions?: ExtractOptions,\n): Citation[] | ResolvedCitation[] {\n\tconst startTime = performance.now()\n\n\t// Step 1: Clean text\n\tconst { cleaned, transformationMap, warnings } = cleanText(\n\t\ttext,\n\t\toptions?.cleaners,\n\t)\n\n\t// Step 2: Tokenize (synchronous)\n\t// Note: Pattern order matters for deduplication - more specific patterns first\n\tconst allPatterns = options?.patterns || [\n\t\t...neutralPatterns,      // Most specific (year-based format)\n\t\t...shortFormPatterns,    // Short-form (requires \" at \" keyword)\n\t\t...casePatterns,         // Case citations (reporter-specific)\n\t\t...statutePatterns,      // Statutes (code-specific)\n\t\t...journalPatterns,      // Least specific (broad pattern)\n\t]\n\tconst tokens = tokenize(cleaned, allPatterns)\n\n\t// Step 3: Deduplicate overlapping tokens\n\t// Multiple patterns may match the same text (e.g., \"500 F.2d 123\" matches both federal-reporter and state-reporter)\n\t// Keep only the most specific match for each position\n\tconst deduplicatedTokens: typeof tokens = []\n\tconst seenPositions = new Set<number | string>()\n\n\t// Performance optimization: Use bitpacking for typical documents (<65K chars)\n\t// For larger documents, fall back to string keys\n\tconst useBitpacking = cleaned.length < 65536\n\n\tfor (const token of tokens) {\n\t\tconst posKey = useBitpacking\n\t\t\t? (token.span.cleanStart << 16) | token.span.cleanEnd\n\t\t\t: `${token.span.cleanStart}-${token.span.cleanEnd}`\n\t\tif (!seenPositions.has(posKey)) {\n\t\t\tseenPositions.add(posKey)\n\t\t\tdeduplicatedTokens.push(token)\n\t\t}\n\t}\n\n\t// Step 3.5: Detect parallel citation groups\n\t// Map of primary token index -> array of secondary token indices\n\tconst parallelGroups = detectParallelCitations(deduplicatedTokens, cleaned)\n\n\t// Step 4: Extract citations from deduplicated tokens\n\tconst citations: Citation[] = []\n\tfor (let i = 0; i < deduplicatedTokens.length; i++) {\n\t\tconst token = deduplicatedTokens[i]\n\t\tlet citation: Citation\n\n\t\tswitch (token.type) {\n\t\t\tcase 'case':\n\t\t\t\t// Check pattern ID to distinguish short-form from full citations\n\t\t\t\tif (token.patternId === 'id' || token.patternId === 'ibid') {\n\t\t\t\t\tcitation = extractId(token, transformationMap)\n\t\t\t\t} else if (token.patternId === 'supra') {\n\t\t\t\t\tcitation = extractSupra(token, transformationMap)\n\t\t\t\t} else if (token.patternId === 'shortFormCase') {\n\t\t\t\t\tcitation = extractShortFormCase(token, transformationMap)\n\t\t\t\t} else {\n\t\t\t\t\tcitation = extractCase(token, transformationMap, cleaned)\n\t\t\t\t}\n\t\t\t\tbreak\n\t\t\tcase 'statute':\n\t\t\t\tcitation = extractStatute(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tcase 'journal':\n\t\t\t\tcitation = extractJournal(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tcase 'neutral':\n\t\t\t\tcitation = extractNeutral(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tcase 'publicLaw':\n\t\t\t\tcitation = extractPublicLaw(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tcase 'federalRegister':\n\t\t\t\tcitation = extractFederalRegister(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tcase 'statutesAtLarge':\n\t\t\t\tcitation = extractStatutesAtLarge(token, transformationMap)\n\t\t\t\tbreak\n\t\t\tdefault:\n\t\t\t\t// Unknown type - skip\n\t\t\t\tcontinue\n\t\t}\n\n\t\t// Attach cleaning warnings to citation if any\n\t\tif (warnings.length > 0) {\n\t\t\tcitation.warnings = [...(citation.warnings || []), ...warnings]\n\t\t}\n\n\t\t// Update processing time\n\t\tcitation.processTimeMs = performance.now() - startTime\n\n\t\t// Populate parallel citation metadata (Phase 8)\n\t\tif (citation.type === 'case') {\n\t\t\t// Check if this citation is part of a parallel group\n\t\t\tconst isPrimary = parallelGroups.has(i)\n\t\t\tconst isSecondary = Array.from(parallelGroups.values()).some(secondaries => secondaries.includes(i))\n\n\t\t\tif (isPrimary || isSecondary) {\n\t\t\t\t// Find the primary citation for this group\n\t\t\t\tlet primaryIndex = i\n\t\t\t\tif (isSecondary) {\n\t\t\t\t\t// Find which group this is a secondary of\n\t\t\t\t\tfor (const [primary, secondaries] of parallelGroups.entries()) {\n\t\t\t\t\t\tif (secondaries.includes(i)) {\n\t\t\t\t\t\t\tprimaryIndex = primary\n\t\t\t\t\t\t\tbreak\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t}\n\n\t\t\t\t// Get the primary token to build groupId\n\t\t\t\tconst primaryToken = deduplicatedTokens[primaryIndex]\n\t\t\t\t// Extract volume, reporter, page from primary token text\n\t\t\t\t// Match: \"volume reporter page\" where reporter is everything except last number\n\t\t\t\tconst match = /^(\\S+)\\s+(.+)\\s+(\\d+)$/.exec(primaryToken.text)\n\t\t\t\tif (match) {\n\t\t\t\t\tconst [, volume, reporter, page] = match\n\t\t\t\t\tcitation.groupId = `${volume}-${reporter.replace(/\\s+/g, '.')}-${page}`\n\n\t\t\t\t\t// Only primary citation gets parallelCitations array\n\t\t\t\t\tif (isPrimary) {\n\t\t\t\t\t\tconst secondaryIndices = parallelGroups.get(i)!\n\t\t\t\t\t\tcitation.parallelCitations = secondaryIndices.map(secIdx => {\n\t\t\t\t\t\t\tconst secToken = deduplicatedTokens[secIdx]\n\t\t\t\t\t\t\tconst secMatch = /^(\\S+)\\s+(.+)\\s+(\\d+)$/.exec(secToken.text)\n\t\t\t\t\t\t\tif (secMatch) {\n\t\t\t\t\t\t\t\tconst [, secVol, secRep, secPage] = secMatch\n\t\t\t\t\t\t\t\treturn {\n\t\t\t\t\t\t\t\t\tvolume: /^\\d+$/.test(secVol) ? Number.parseInt(secVol, 10) : secVol,\n\t\t\t\t\t\t\t\t\treporter: secRep,\n\t\t\t\t\t\t\t\t\tpage: Number.parseInt(secPage, 10),\n\t\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\treturn { volume: 0, reporter: '', page: 0 } // Fallback (shouldn't happen)\n\t\t\t\t\t\t})\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\n\t\tcitations.push(citation)\n\t}\n\n\t// Step 5: Resolve short-form citations if requested\n\tif (options?.resolve) {\n\t\treturn resolveCitations(citations, text, options.resolutionOptions)\n\t}\n\n\treturn citations\n}\n\n/**\n * Asynchronous version of extractCitations().\n *\n * Currently wraps the synchronous extractCitations() function. This API\n * exists for future extensibility when async operations are added:\n * - Async reporters-db lookups (Phase 3)\n * - Async resolution/annotation services\n * - Web Workers for parallel processing\n *\n * For now, this function immediately resolves with the same results as\n * the synchronous version.\n *\n * @param text - Raw text to extract citations from\n * @param options - Optional customization (cleaners, patterns, resolve)\n * @returns Promise resolving to array of citations (or ResolvedCitation[] if resolve: true)\n *\n * @example\n * ```typescript\n * const citations = await extractCitationsAsync(text, { resolve: true })\n * // Returns ResolvedCitation[] with resolution metadata\n * ```\n */\nexport async function extractCitationsAsync(text: string, options: ExtractOptions & { resolve: true }): Promise<ResolvedCitation[]>\nexport async function extractCitationsAsync(text: string, options?: ExtractOptions): Promise<Citation[]>\nexport async function extractCitationsAsync(\n\ttext: string,\n\toptions?: ExtractOptions,\n): Promise<Citation[] | ResolvedCitation[]> {\n\t// Async wrapper for future extensibility (e.g., async reporters-db lookup)\n\t// For MVP, wraps synchronous extractCitations\n\treturn extractCitations(text, options)\n}\n"],"mappings":"mEAKA,SAAgB,EAAe,EAA8C,CAC3E,OAAO,EAAS,OAAS,QACpB,EAAS,OAAS,WAClB,EAAS,OAAS,WAClB,EAAS,OAAS,WAClB,EAAS,OAAS,aAClB,EAAS,OAAS,mBAClB,EAAS,OAAS,kBAMzB,SAAgB,EAAoB,EAAmD,CACrF,OAAO,EAAS,OAAS,MACpB,EAAS,OAAS,SAClB,EAAS,OAAS,gBAMzB,SAAgB,EAAe,EAAkD,CAC/E,OAAO,EAAS,OAAS,OAO3B,SAAgB,EACd,EACA,EAC+B,CAC/B,OAAO,EAAS,OAAS,EAmB3B,SAAgB,EAAkB,EAAiB,CACjD,MAAU,MAAM,qBAAqB,IAAI,CC7C3C,SAAgB,EAAc,EAAsB,CACnD,OAAO,EAAK,QAAQ,WAAY,GAAG,CAUpC,SAAgB,EAAoB,EAAsB,CACzD,OAAO,EAAK,QAAQ,aAAc,IAAI,CAAC,QAAQ,SAAU,IAAI,CAU9D,SAAgB,EAAiB,EAAsB,CACtD,OAAO,EAAK,UAAU,OAAO,CAU9B,SAAgB,EAAe,EAAsB,CACpD,OAAO,EACL,QAAQ,kBAAmB,IAAI,CAC/B,QAAQ,kBAAmB,IAAI,CCXlC,SAAgB,EACf,EACA,EAA4C,CAC3C,EACA,EACA,EACA,EACA,CACiB,CAElB,IAAI,EAAc,EACd,EAAkB,IAAI,IACtB,EAAkB,IAAI,IAG1B,IAAK,IAAI,EAAI,EAAG,GAAK,EAAS,OAAQ,IACrC,EAAgB,IAAI,EAAG,EAAE,CACzB,EAAgB,IAAI,EAAG,EAAE,CAI1B,IAAK,IAAM,KAAW,EAAU,CAC/B,IAAM,EAAa,EACb,EAAY,EAAQ,EAAY,CAEtC,GAAI,IAAe,EAAW,CAE7B,GAAM,CAAE,qBAAoB,sBAAuB,EAClD,EACA,EACA,EACA,EACA,CAED,EAAkB,EAClB,EAAkB,EAClB,EAAc,GAShB,MAAO,CACN,QAAS,EACT,kBAP4C,CAC5C,kBACA,kBACA,CAKA,SAAU,EAAE,CACZ,CAeF,SAAS,EACR,EACA,EACA,EACA,EAIC,CACD,IAAM,EAAqB,IAAI,IACzB,EAAqB,IAAI,IAE3B,EAAY,EACZ,EAAW,EAGf,KAAO,GAAa,EAAW,QAAU,GAAY,EAAU,QAAQ,CAEtE,GAAI,GAAa,EAAW,QAAU,GAAY,EAAU,OAAQ,CACnE,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAU,EAAY,CAC7C,EAAmB,IAAI,EAAa,EAAS,CAC7C,MAID,GAAI,GAAa,EAAW,OAAQ,CACnC,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAU,EAAY,CAC7C,IACA,SAID,GAAI,GAAY,EAAU,OAAQ,CACjC,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAa,EAAS,CAC7C,IACA,SAID,GAAI,EAAW,KAAe,EAAU,GAAW,CAClD,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAU,EAAY,CAC7C,EAAmB,IAAI,EAAa,EAAS,CAC7C,IACA,QACM,CAGN,IAAI,EAAa,GAIjB,IAAK,IAAI,EAAY,EAAG,GAAa,IAChC,IAAY,GAAa,EAAW,QADU,IAGlD,GAAI,EAAW,EAAY,KAAe,EAAU,GAAW,CAE9D,IAAK,IAAI,EAAI,EAAG,EAAI,EAAW,IAAK,CACnC,IAAM,EACL,EAAmB,IAAI,EAAY,EAAE,EAAI,EAAY,EACtD,EAAmB,IAAI,EAAa,EAAS,CAE9C,GAAa,EACb,EAAa,GACb,MAIF,GAAI,EAAY,SAGhB,IAAK,IAAI,EAAY,EAAG,GAAa,IAChC,IAAW,GAAa,EAAU,QADY,IAGlD,GAAI,EAAW,KAAe,EAAU,EAAW,GAAY,CAE9D,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,IAAK,IAAI,EAAI,EAAG,EAAI,EAAW,IAC9B,EAAmB,IAAI,EAAW,EAAG,EAAY,CAElD,GAAY,EACZ,EAAa,GACb,MAIF,GAAI,EAAY,SAGhB,IAAM,EAAc,EAAmB,IAAI,EAAU,EAAI,EACzD,EAAmB,IAAI,EAAU,EAAY,CAC7C,EAAmB,IAAI,EAAa,EAAS,CAC7C,IACA,KAIF,MAAO,CAAE,qBAAoB,qBAAoB,CCpLlD,MAAa,EAA0B,CACrC,CACE,GAAI,mBACJ,MAAO,yJACP,YAAa,0DACb,KAAM,OACP,CACD,CACE,GAAI,gBACJ,MAAO,wGACP,YAAa,+BACb,KAAM,OACP,CACD,CACE,GAAI,iBACJ,MAAO,4GACP,YAAa,6EACb,KAAM,OACP,CACF,CC3BY,EAA6B,CACxC,CACE,GAAI,MACJ,MAAO,gDACP,YAAa,iDACb,KAAM,UACP,CACD,CACE,GAAI,aACJ,MAAO,iEACP,YAAa,sEACb,KAAM,UACP,CACF,CCbY,EAA6B,CACxC,CACE,GAAI,aACJ,MAAO,oDACP,YAAa,iGACb,KAAM,UACP,CACF,CCPY,EAA6B,CACxC,CACE,GAAI,UACJ,MAAO,4BACP,YAAa,6CACb,KAAM,UACP,CACD,CACE,GAAI,QACJ,MAAO,wCACP,YAAa,sDACb,KAAM,UACP,CACD,CACE,GAAI,aACJ,MAAO,sCACP,YAAa,oDACb,KAAM,YACP,CACD,CACE,GAAI,mBACJ,MAAO,8CACP,YAAa,0DACb,KAAM,kBACP,CACD,CACE,GAAI,oBACJ,MAAO,uCACP,YAAa,sDACb,KAAM,kBACP,CACD,CACE,GAAI,qBACJ,MAAO,kEACP,YAAa,2EACb,KAAM,UACP,CACF,CCRY,EAA+B,CAC1C,CACE,GAAI,KACJ,MAhC8B,+BAiC9B,YAAa,8CACb,KAAM,OACP,CACD,CACE,GAAI,OACJ,MAnCgC,iCAoChC,YAAa,oDACb,KAAM,OACP,CACD,CACE,GAAI,QACJ,MAjCiC,4FAkCjC,YAAa,mEACb,KAAM,OACP,CACD,CACE,GAAI,gBACJ,MAhC2C,sEAiC3C,YAAa,sDACb,KAAM,OACP,CACF,CCcD,SAAgB,EACd,EACA,EAAsB,CACpB,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACJ,CACQ,CACT,IAAM,EAAkB,EAAE,CAE1B,IAAK,IAAM,KAAW,EACpB,GAAI,CAEF,IAAM,EAAU,EAAY,SAAS,EAAQ,MAAM,CAEnD,IAAK,IAAM,KAAS,EAElB,EAAO,KAAK,CACV,KAAM,EAAM,GACZ,KAAM,CACJ,WAAY,EAAM,MAClB,SAAU,EAAM,MAAS,EAAM,GAAG,OACnC,CACD,KAAM,EAAQ,KACd,UAAW,EAAQ,GACpB,CAAC,OAEG,EAAO,CAEd,QAAQ,KACN,WAAW,EAAQ,GAAG,yBACtB,aAAiB,MAAQ,EAAM,QAAU,OAAO,EAAM,CACvD,CAOL,OAFA,EAAO,MAAM,EAAG,IAAM,EAAE,KAAK,WAAa,EAAE,KAAK,WAAW,CAErD,ECxFT,MAAM,EAAoC,CACxC,IAAK,EACL,QAAS,EACT,IAAK,EACL,SAAU,EACV,IAAK,EACL,MAAO,EACP,IAAK,EACL,MAAO,EACP,IAAK,EACL,IAAK,EACL,KAAM,EACN,IAAK,EACL,KAAM,EACN,IAAK,EACL,OAAQ,EACR,IAAK,EACL,KAAM,EACN,UAAW,EACX,IAAK,GACL,QAAS,GACT,IAAK,GACL,SAAU,GACV,IAAK,GACL,SAAU,GACX,CAgBD,SAAgB,EAAW,EAA0B,CAGnD,IAAM,EAAQ,EADK,EAAS,aAAa,CAAC,QAAQ,MAAO,GAAG,EAG5D,GAAI,IAAU,IAAA,GACZ,MAAU,MAAM,uBAAuB,IAAW,CAGpD,OAAO,EAiBT,SAAgB,EAAU,EAA4B,CACpD,GAAM,CAAE,OAAM,QAAO,OAAQ,EAgB7B,OAdI,IAAU,IAAA,IAAa,IAAQ,IAAA,GAI1B,GAAG,EAAK,GAFE,OAAO,EAAM,CAAC,SAAS,EAAG,IAAI,CAEpB,GADZ,OAAO,EAAI,CAAC,SAAS,EAAG,IAAI,GAIzC,IAAU,IAAA,GAOP,OAAO,EAAK,CAJV,GAAG,EAAK,GADE,OAAO,EAAM,CAAC,SAAS,EAAG,IAAI,GA4BnD,SAAgB,EAAU,EAA6C,CAErE,IAAM,EAAY,EAAQ,MAAM,yFAAyF,CACzH,GAAI,EAAW,CACb,IAAM,EAAQ,EAAW,EAAU,GAAG,CAChC,EAAM,OAAO,SAAS,EAAU,GAAI,GAAG,CAEvC,EAAS,CAAE,KADJ,OAAO,SAAS,EAAU,GAAI,GAAG,CACvB,QAAO,MAAK,CACnC,MAAO,CAAE,IAAK,EAAU,EAAO,CAAE,SAAQ,CAI3C,IAAM,EAAY,EAAQ,MAAM,uHAAuH,CACvJ,GAAI,EAAW,CACb,IAAM,EAAQ,EAAW,EAAU,GAAG,CAChC,EAAM,OAAO,SAAS,EAAU,GAAI,GAAG,CAEvC,EAAS,CAAE,KADJ,OAAO,SAAS,EAAU,GAAI,GAAG,CACvB,QAAO,MAAK,CACnC,MAAO,CAAE,IAAK,EAAU,EAAO,CAAE,SAAQ,CAI3C,IAAM,EAAe,EAAQ,MAAM,oCAAoC,CACvE,GAAI,EAAc,CAChB,IAAM,EAAQ,OAAO,SAAS,EAAa,GAAI,GAAG,CAC5C,EAAM,OAAO,SAAS,EAAa,GAAI,GAAG,CAE1C,EAAS,CAAE,KADJ,OAAO,SAAS,EAAa,GAAI,GAAG,CAC1B,QAAO,MAAK,CACnC,MAAO,CAAE,IAAK,EAAU,EAAO,CAAE,SAAQ,CAI3C,IAAM,EAAY,EAAQ,MAAM,cAAc,CAC9C,GAAI,EAAW,CAEb,IAAM,EAAS,CAAE,KADJ,OAAO,SAAS,EAAU,GAAI,GAAG,CACvB,CACvB,MAAO,CAAE,IAAK,EAAU,EAAO,CAAE,SAAQ,EC7J7C,SAAS,EAAY,EAA8B,CAClD,IAAM,EAAM,OAAO,SAAS,EAAK,GAAG,CACpC,OAAO,OAAO,EAAI,GAAK,EAAM,EAAM,EAIpC,MAAM,EAAgB,oJAOhB,EAA6B,0DAG7B,EAAmB,aAGnB,EAAgB,YAGhB,GAAc,cAGd,GAAwB,8BAGxB,EAA0B,aAG1B,EAA4B,wCAG5B,EAAwB,0CAGxB,EAA0B,WAG1B,EAAoB,kFAGpB,EAA0B,+IAUhC,SAAS,EAAmB,EAAqC,CAEhE,IAAI,EAAQ,EAAQ,QAAQ,iCAAkC,GAAG,CAAC,MAAM,CAQxE,MANA,GAAQ,EAAM,QAAQ,eAAgB,GAAG,CAAC,MAAM,CAEhD,EAAQ,EAAM,QAAQ,2BAA4B,GAAG,CAAC,MAAM,CAC5D,EAAQ,EAAM,QAAY,OAAO,OAAO,EAAc,OAAO,OAAQ,IAAI,CAAE,GAAG,CAAC,MAAM,CAErF,EAAQ,EAAM,QAAQ,QAAS,GAAG,CAAC,MAAM,CAClC,GAAS,WAAW,KAAK,EAAM,CAAG,EAAQ,IAAA,GAkBlD,SAAS,EACR,EACA,EACA,EAAc,IACwC,CACtD,IAAM,EAAc,KAAK,IAAI,EAAG,EAAY,EAAY,CACpD,EAAgB,EAAY,UAAU,EAAa,EAAU,CAC7D,EAAsB,EAKtB,EAAoB,GACpB,EACJ,MAAQ,EAAQ,EAAwB,KAAK,EAAc,IAAM,MAChE,EAAoB,EAAM,MAAQ,EAAM,GAAG,OAGxC,IAAsB,KACzB,EAAgB,EAAc,UAAU,EAAkB,CAC1D,EAAsB,EAAc,GAKrC,IAAM,EAAS,EAAkB,KAAK,EAAc,CACpD,GAAI,GAEC,CAAC,EAAO,GAAG,SAAS,IAAI,CAG3B,MAAO,CAAE,SAFQ,GAAG,EAAO,GAAG,MAAM,CAAC,MAAM,EAAO,GAAG,MAAM,GAExC,UADD,EAAsB,EAAO,MACjB,CAKhC,IAAM,EAAY,EAAwB,KAAK,EAAc,CAC7D,GAAI,GAEC,CAAC,EAAU,GAAG,SAAS,IAAI,CAG9B,MAAO,CAAE,SAFQ,GAAG,EAAU,GAAG,GAAG,EAAU,GAAG,MAAM,GAEpC,UADD,EAAsB,EAAU,MACpB,CAsBjC,SAAS,GACR,EACA,EACA,EAAe,IACN,CACT,IAAI,EAAM,EACJ,EAAW,KAAK,IAAI,EAAY,OAAQ,EAAc,EAAa,CACrE,EAAQ,EACR,EAAgB,GAEpB,KAAO,EAAM,GAAU,CACtB,IAAM,EAAO,EAAY,GAEzB,GAAI,IAAS,IACZ,IACA,EAAgB,GAChB,YACU,IAAS,IAKnB,IAJA,IACA,IAGI,IAAU,EAAG,CAEhB,IAAI,EAAU,EACd,KAAO,EAAU,GAAY,KAAK,KAAK,EAAY,GAAS,EAC3D,IAID,GAAI,EAAY,KAAa,IAAK,CACjC,EAAM,EACN,SAID,IAAM,EAAgB,EAAY,UAAU,EAAS,EAAS,CAG9D,GADC,kEACgB,KAAK,EAAc,CAAE,CAErC,EAAM,EACN,SAID,OAAO,QAGR,IAMF,OAAO,EAAgB,EAAM,EAsB9B,SAAS,EAAmB,EAK1B,CACD,IAAM,EAKF,EAAE,CAGA,EAAa,EAAU,EAAQ,CACjC,IACH,EAAO,KAAO,EACd,EAAO,KAAO,EAAW,OAAO,MAIjC,IAAM,EAAc,EAAmB,EAAQ,CAY/C,OAXI,IACH,EAAO,MAAQ,GAIZ,eAAe,KAAK,EAAQ,CAC/B,EAAO,YAAc,UACX,kBAAkB,KAAK,EAAQ,GACzC,EAAO,YAAc,cAGf,EAwBR,SAAS,EAAmB,EAAsB,CACjD,IAAI,EAAa,EAGjB,EAAa,EAAW,QAAQ,iBAAkB,GAAG,CAGrD,EAAa,EAAW,QAAQ,mBAAoB,GAAG,CAGvD,EAAa,EAAW,QAAQ,eAAgB,GAAG,CAInD,IAAI,EAAO,GACX,KAAO,IAAS,GACf,EAAO,EACP,EAAa,EAAW,QAAQ,+CAAgD,GAAG,CAUpF,MANA,GAAa,EAAW,QAAQ,kBAAmB,GAAG,CAGtD,EAAa,EAAW,QAAQ,OAAQ,IAAI,CAGrC,EAAW,MAAM,CAAC,aAAa,CAsBvC,SAAS,EAAkB,EAMzB,CAcD,IAAK,IAAM,IAZgB,CAC1B,QACA,WACA,YACA,gBACA,wBACA,iBACA,cACA,YACA,CAGwC,CAExC,IAAM,EADkB,OAAO,KAAK,EAAO,YAAa,IAAI,CAClC,KAAK,EAAS,CACxC,GAAI,EAAO,CACV,IAAM,EAAgB,EAAM,GACtB,EAAU,EAAM,GAGtB,GAAI,cAAc,KAAK,EAAQ,CAAE,CAGhC,IAAM,EAAS,yBAAyB,KAAK,EAAS,CACtD,GAAI,EAAQ,CACX,IAAM,EAAY,EAAO,GAAG,MAAM,CAC5B,EAAY,EAAO,GAAG,MAAM,CAClC,MAAO,CACN,YACA,oBAAqB,EAAmB,EAAU,CAClD,YACA,oBAAqB,EAAmB,EAAU,CAClD,OAIF,MAAO,CACN,UAAW,EACX,oBAAqB,EAAmB,EAAQ,CAChD,iBAAkB,EAClB,EAOJ,IAAM,EADS,yBACO,KAAK,EAAS,CACpC,GAAI,EAAQ,CACX,IAAI,EAAY,EAAO,GAAG,MAAM,CAC1B,EAAY,EAAO,GAAG,MAAM,CAMlC,MAFA,GAAY,EAAU,QAAQ,iFAAkF,GAAG,CAAC,MAAM,CAEnH,CACN,UAAW,GAAa,EAAO,GAAG,MAAM,CACxC,oBAAqB,EAAmB,GAAa,EAAO,GAAG,MAAM,CAAC,CACtE,YACA,oBAAqB,EAAmB,EAAU,CAClD,CAIF,MAAO,EAAE,CAsDV,SAAgB,EACf,EACA,EACA,EACmB,CACnB,GAAM,CAAE,OAAM,QAAS,EAKjB,EAAQ,EAA2B,KAAK,EAAK,CAEnD,GAAI,CAAC,EAEJ,MAAU,MAAM,kCAAkC,IAAO,CAG1D,IAAM,EAAS,EAAY,EAAM,GAAG,CAC9B,EAAW,EAAM,GAAG,MAAM,CAG1B,EAAU,EAAM,GAChB,EAAc,EAAiB,KAAK,EAAQ,CAC5C,EAAO,EAAc,IAAA,GAAY,OAAO,SAAS,EAAS,GAAG,CAC7D,EAAe,EAAc,GAAO,IAAA,GAIpC,EAAe,EAAc,KAAK,EAAK,CACzC,EAAU,EAAe,OAAO,SAAS,EAAa,GAAI,GAAG,CAAG,IAAA,GAGhE,EACA,EACA,EACA,EACA,EACA,EAGA,EAEE,EAAa,GAAY,KAAK,EAAK,CACzC,GAAI,EAAY,CACf,EAAuB,EAAW,GAElC,IAAM,EAAc,EAAmB,EAAqB,CAC5D,EAAO,EAAY,KACnB,EAAQ,EAAY,MACpB,EAAO,EAAY,KACnB,EAAc,EAAY,YAM3B,GAAI,GAAe,CAAC,EAAsB,CACzC,IAAM,EAAa,EAAY,UAAU,EAAK,SAAS,CACjD,EAAiB,GAAsB,KAAK,EAAW,CAC7D,GAAI,EAAgB,CACnB,EAAuB,EAAe,GAEtC,IAAM,EAAc,EAAmB,EAAqB,CAO5D,GANA,EAAO,EAAY,KACnB,EAAQ,EAAY,MACpB,EAAO,EAAY,KACnB,EAAc,EAAY,YAGtB,IAAY,IAAA,GAAW,CAC1B,IAAM,EAAiB,EAAwB,KAAK,EAAW,CAC3D,IACH,EAAU,OAAO,SAAS,EAAe,GAAI,GAAG,IAOpD,GAAI,GAAe,CAAC,EAAa,CAChC,IAAM,EAAa,EAAY,UAAU,EAAK,SAAS,CAEjD,EAAe,EAA0B,KAAK,EAAW,CAC3D,IACH,EAAc,EAAa,GAAG,aAAa,EAU7C,GALI,CAAC,GAAS,EAAsB,KAAK,EAAS,GACjD,EAAQ,UAIL,EAAa,CAChB,IAAM,EAAiB,EAAgB,EAAa,EAAK,WAAW,CACpE,GAAI,EAAgB,CACnB,EAAW,EAAe,SAG1B,IAAM,EAAW,GAAqB,EAAa,EAAK,SAAS,CAC3D,EAAiB,EAAe,UAChC,EAAe,EAAW,EAAK,SAAW,EAAW,EAAK,SAQhE,EAAW,CACV,WAAY,EACZ,SAAU,EACV,cAPA,EAAkB,gBAAgB,IAAI,EAAe,EAAI,EAQzD,YANA,EAAkB,gBAAgB,IAAI,EAAa,EAAI,EAOvD,EAKH,IAAI,EACA,EACA,EACA,EACA,EAEJ,GAAI,EAAU,CACb,IAAM,EAAc,EAAkB,EAAS,CAC/C,EAAY,EAAY,UACxB,EAAsB,EAAY,oBAClC,EAAY,EAAY,UACxB,EAAsB,EAAY,oBAClC,EAAmB,EAAY,iBAIhC,IAAM,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAG1D,EAAa,GAqCjB,GAlCwB,4JA2BvB,CAEmB,KAAM,GAAM,EAAS,SAAS,EAAE,CAAC,GACpD,GAAc,IAIX,IAAS,IAAA,GAAW,CACvB,IAAM,EAAc,IAAI,MAAM,CAAC,aAAa,CACxC,GAAQ,IACX,GAAc,IAYhB,MAPA,GAAa,KAAK,IAAI,EAAY,EAAI,CAGlC,IACH,EAAa,IAGP,CACN,KAAM,OACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,aACA,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,SACA,WACA,OACA,UACA,QACA,OACA,eACA,OACA,WACA,WACA,cACA,YACA,sBACA,YACA,sBACA,mBACA,CC3oBF,SAAgB,EACf,EACA,EACkB,CAClB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADe,yDACM,KAAK,EAAK,CAErC,GAAI,CAAC,EACJ,MAAU,MAAM,qCAAqC,IAAO,CAG7D,IAAM,EAAQ,EAAM,GAAK,OAAO,SAAS,EAAM,GAAI,GAAG,CAAG,IAAA,GACnD,EAAO,EAAM,GAAG,MAAM,CACtB,EAAU,EAAM,GAGhB,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAG1D,EAAa,GAkBjB,MAfmB,CAClB,SACA,SACA,iBACA,kBACA,0BACA,8BACA,CAEc,KAAM,GAAM,EAAK,SAAS,EAAE,CAAC,GAC3C,GAAc,IAGf,EAAa,KAAK,IAAI,EAAY,EAAI,CAE/B,CACN,KAAM,UACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,aACA,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,QACA,OACA,UACA,CC1DF,SAAgB,EACf,EACA,EACkB,CAClB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADe,4CACM,KAAK,EAAK,CAErC,GAAI,CAAC,EACJ,MAAU,MAAM,qCAAqC,IAAO,CAG7D,IAAM,EAAY,EAAM,GAClB,EAAS,QAAQ,KAAK,EAAU,CAAG,OAAO,SAAS,EAAW,GAAG,CAAG,EACpE,EAAU,EAAM,GAAG,MAAM,CACzB,EAAO,OAAO,SAAS,EAAM,GAAI,GAAG,CAIpC,EADe,YACa,KAAK,EAAK,CACtC,EAAU,EAAe,OAAO,SAAS,EAAa,GAAI,GAAG,CAAG,IAAA,GAGhE,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,UACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,SACA,UACA,aAAc,EACd,OACA,UACA,CCxDF,SAAgB,EACf,EACA,EACkB,CAClB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADe,+CACM,KAAK,EAAK,CAErC,GAAI,CAAC,EACJ,MAAU,MAAM,qCAAqC,IAAO,CAG7D,IAAM,EAAO,OAAO,SAAS,EAAM,GAAI,GAAG,CACpC,EAAQ,EAAM,GACd,EAAiB,EAAM,GAGvB,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,UACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,EAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,OACA,QACA,iBACA,CC3CF,SAAgB,EACf,EACA,EACoB,CACpB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADiB,wCACM,KAAK,EAAK,CAEvC,GAAI,CAAC,EACJ,MAAU,MAAM,wCAAwC,IAAO,CAGhE,IAAM,EAAW,OAAO,SAAS,EAAM,GAAI,GAAG,CACxC,EAAY,OAAO,SAAS,EAAM,GAAI,GAAG,CAGzC,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,YACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,WACA,YACA,CC5CF,SAAgB,EACf,EACA,EAC0B,CAC1B,GAAM,CAAE,OAAM,QAAS,EAKjB,EADuB,0CACM,KAAK,EAAK,CAE7C,GAAI,CAAC,EACJ,MAAU,MAAM,8CAA8C,IAAO,CAGtE,IAAM,EAAY,EAAM,GAClB,EAAS,QAAQ,KAAK,EAAU,CAAG,OAAO,SAAS,EAAW,GAAG,CAAG,EACpE,EAAO,OAAO,SAAS,EAAM,GAAI,GAAG,CAKpC,EADY,wBACU,KAAK,EAAK,CAChC,EAAO,EAAY,OAAO,SAAS,EAAU,GAAI,GAAG,CAAG,IAAA,GAGvD,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,kBACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,SACA,OACA,OACA,CChFF,SAAgB,EACf,EACA,EAC0B,CAC1B,GAAM,CAAE,OAAM,QAAS,EAIjB,EADY,mCACM,KAAK,EAAK,CAElC,GAAI,CAAC,EACJ,MAAU,MAAM,+CAA+C,IAAO,CAGvE,IAAM,EAAY,EAAM,GAClB,EAAS,QAAQ,KAAK,EAAU,CAAG,OAAO,SAAS,EAAW,GAAG,CAAG,EACpE,EAAO,OAAO,SAAS,EAAM,GAAI,GAAG,CAIpC,EADY,wBACU,KAAK,EAAK,CAChC,EAAO,EAAY,OAAO,SAAS,EAAU,GAAI,GAAG,CAAG,IAAA,GAGvD,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,kBACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,SACA,OACA,OACA,CCrBF,SAAgB,EACf,EACA,EACa,CACb,GAAM,CAAE,OAAM,QAAS,EAKjB,EADU,oCACM,KAAK,EAAK,CAEhC,GAAI,CAAC,EACJ,MAAU,MAAM,iCAAiC,IAAO,CAIzD,IAAM,EAAU,EAAM,GAAK,OAAO,SAAS,EAAM,GAAI,GAAG,CAAG,IAAA,GAGrD,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,KACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,EAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,UACA,CAmCF,SAAgB,EACf,EACA,EACgB,CAChB,GAAM,CAAE,OAAM,QAAS,EAMjB,EADa,2FACM,KAAK,EAAK,CAEnC,GAAI,CAAC,EACJ,MAAU,MAAM,mCAAmC,IAAO,CAG3D,IAAM,EAAY,EAAM,GAClB,EAAU,EAAM,GAAK,OAAO,SAAS,EAAM,GAAI,GAAG,CAAG,IAAA,GAGrD,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,QACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,YACA,UACA,CAqCF,SAAgB,EACf,EACA,EACwB,CACxB,GAAM,CAAE,OAAM,QAAS,EAKjB,EADiB,iEACM,KAAK,EAAK,CAEvC,GAAI,CAAC,EACJ,MAAU,MAAM,6CAA6C,IAAO,CAGrE,IAAM,EAAY,EAAM,GAClB,EAAS,QAAQ,KAAK,EAAU,CAAG,OAAO,SAAS,EAAW,GAAG,CAAG,EACpE,EAAW,EAAM,GAAG,MAAM,CAC1B,EAAU,OAAO,SAAS,EAAM,GAAI,GAAG,CAGvC,EACL,EAAkB,gBAAgB,IAAI,EAAK,WAAW,EAAI,EAAK,WAC1D,EACL,EAAkB,gBAAgB,IAAI,EAAK,SAAS,EAAI,EAAK,SAK9D,MAAO,CACN,KAAM,gBACN,OACA,KAAM,CACL,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,gBACA,cACA,CACD,WAXkB,GAYlB,YAAa,EACb,cAAe,EACf,gBAAiB,EACjB,SACA,WACA,UACA,CClOF,SAAgB,EACd,EACA,EACA,EAA0B,SACL,CACrB,IAAM,EAAe,IAAI,IAGnB,EAAuB,CAAC,EAAE,CAC5B,EAEJ,MAAQ,EAAQ,EAAgB,KAAK,EAAK,IAAM,MAE9C,EAAW,KAAK,EAAM,MAAQ,EAAM,GAAG,OAAO,CAGhD,EAAW,KAAK,EAAK,OAAO,CAG5B,IAAK,IAAI,EAAI,EAAG,EAAI,EAAU,OAAQ,IAAK,CAEzC,IAAM,EADW,EAAU,GACI,KAAK,cAGhC,EAAe,EACnB,IAAK,IAAI,EAAI,EAAG,EAAI,EAAW,OAAS,EAAG,IACzC,GAAI,GAAiB,EAAW,IAAM,EAAgB,EAAW,EAAI,GAAI,CACvE,EAAe,EACf,MAIJ,EAAa,IAAI,EAAG,EAAa,CAGnC,OAAO,EAYT,SAAgB,EACd,EACA,EACA,EACA,EACS,CACT,GAAI,IAAa,OAEf,MAAO,GAIT,IAAM,EAAsB,EAAa,IAAI,EAAgB,CACvD,EAAmB,EAAa,IAAI,EAAa,CASvD,OANI,IAAwB,IAAA,IAAa,IAAqB,IAAA,GACrD,GAKF,IAAwB,ECpEjC,SAAgB,EAAoB,EAAW,EAAmB,CAEhE,GAAI,EAAE,SAAW,EAAG,OAAO,EAAE,OAC7B,GAAI,EAAE,SAAW,EAAG,OAAO,EAAE,OAI7B,IAAM,EAAiB,MAAM,KAAK,CAAE,OAAQ,EAAE,OAAS,EAAG,KACxD,MAAM,EAAE,OAAS,EAAE,CAAC,KAAK,EAAE,CAC5B,CAGD,IAAK,IAAI,EAAI,EAAG,GAAK,EAAE,OAAQ,IAC7B,EAAG,GAAG,GAAK,EAEb,IAAK,IAAI,EAAI,EAAG,GAAK,EAAE,OAAQ,IAC7B,EAAG,GAAG,GAAK,EAIb,IAAK,IAAI,EAAI,EAAG,GAAK,EAAE,OAAQ,IAC7B,IAAK,IAAI,EAAI,EAAG,GAAK,EAAE,OAAQ,IACzB,EAAE,EAAI,KAAO,EAAE,EAAI,GAErB,EAAG,GAAG,GAAK,EAAG,EAAI,GAAG,EAAI,GAMzB,EAAG,GAAG,GAAK,EAAI,KAAK,IAClB,EAAG,EAAI,GAAG,GACV,EAAG,GAAG,EAAI,GACV,EAAG,EAAI,GAAG,EAAI,GACf,CAKP,OAAO,EAAG,EAAE,QAAQ,EAAE,QAgBxB,SAAgB,EAA8B,EAAW,EAAmB,CAE1E,IAAM,EAAS,EAAE,aAAa,CACxB,EAAS,EAAE,aAAa,CAGxB,EAAW,EAAoB,EAAQ,EAAO,CAG9C,EAAY,KAAK,IAAI,EAAO,OAAQ,EAAO,OAAO,CAIxD,OAHI,IAAc,EAAU,EAGrB,EAAI,EAAW,ECtDxB,IAAa,EAAb,KAA8B,CAa5B,YACE,EACA,EACA,EAA6B,EAAE,CAC/B,CACA,KAAK,UAAY,EACjB,KAAK,KAAO,EAGZ,KAAK,QAAU,CACb,cAAe,EAAQ,eAAiB,YACxC,qBAAsB,EAAQ,sBAAwB,GACtD,yBAA0B,EAAQ,0BAA4B,SAC9D,mBAAoB,EAAQ,oBAAsB,GAClD,oBAAqB,EAAQ,qBAAuB,GACpD,sBAAuB,EAAQ,uBAAyB,GACxD,iBAAkB,EAAQ,kBAAoB,GAC/C,CAGD,KAAK,QAAU,CACb,cAAe,EACf,aAAc,EACd,iBAAkB,IAAA,GAClB,oBAAqB,IAAI,IACzB,aAAc,IAAI,IACnB,CAGG,KAAK,QAAQ,uBACf,KAAK,QAAQ,aAAe,EAC1B,EACA,EACA,KAAK,QAAQ,yBACd,EASL,SAA8B,CAC5B,IAAM,EAA+B,EAAE,CAEvC,IAAK,IAAI,EAAI,EAAG,EAAI,KAAK,UAAU,OAAQ,IAAK,CAC9C,KAAK,QAAQ,cAAgB,EAC7B,IAAM,EAAW,KAAK,UAAU,GAG5B,EAEJ,OAAQ,EAAS,KAAjB,CACE,IAAK,KACH,EAAa,KAAK,UAAU,EAAS,CACrC,MACF,IAAK,QACH,EAAa,KAAK,aAAa,EAAS,CACxC,MACF,IAAK,gBACH,EAAa,KAAK,qBAAqB,EAAS,CAChD,MACF,QAEM,EAAe,EAAS,GAC1B,KAAK,QAAQ,iBAAmB,EAChC,KAAK,kBAAkB,EAAU,EAAE,EAErC,MAKJ,EAAS,KAAK,CACZ,GAAG,EACH,aACD,CAAqB,CAGxB,OAAO,EAMT,UAAkB,EAAqD,CACrE,IAAM,EAAe,KAAK,QAAQ,cAG9B,EACJ,IAAK,IAAI,EAAI,EAAe,EAAG,GAAK,EAAG,IAErC,GADkB,KAAK,UAAU,GACnB,OAAS,OAAQ,CAC7B,EAAkB,EAClB,MAcJ,OATI,IAAoB,IAAA,GACf,KAAK,oBAAoB,wCAAwC,CAIrE,KAAK,cAAc,EAAiB,EAAa,CAI/C,CACL,WAAY,EACZ,WAAY,EACb,CANQ,KAAK,oBAAoB,6CAA6C,CAYjF,aAAqB,EAAuD,CAC1E,IAAM,EAAe,KAAK,QAAQ,cAC5B,EAAkB,KAAK,mBAAmB,EAAS,UAAU,CAG/D,EAEJ,IAAK,GAAM,CAAC,EAAW,KAAkB,KAAK,QAAQ,oBAAqB,CAEzE,GAAI,CAAC,KAAK,cAAc,EAAe,EAAa,CAClD,SAIF,IAAM,EAAa,EAA8B,EAAiB,EAAU,EAGxE,CAAC,GAAa,EAAa,EAAU,cACvC,EAAY,CAAE,MAAO,EAAe,aAAY,EAKpD,GAAI,CAAC,EACH,OAAO,KAAK,oBAAoB,kCAAkC,CAGpE,GAAI,EAAU,WAAa,KAAK,QAAQ,oBACtC,OAAO,KAAK,oBACV,yBAAyB,EAAU,WAAW,QAAQ,EAAE,CAAC,mBAAmB,KAAK,QAAQ,sBAC1F,CAIH,IAAM,EAAqB,EAAE,CAK7B,OAJI,EAAU,WAAa,GACzB,EAAS,KAAK,2BAA2B,EAAU,WAAW,QAAQ,EAAE,GAAG,CAGtE,CACL,WAAY,EAAU,MACtB,WAAY,EAAU,WACtB,SAAU,EAAS,OAAS,EAAI,EAAW,IAAA,GAC5C,CAMH,qBAA6B,EAA+D,CAC1F,IAAM,EAAe,KAAK,QAAQ,cAGlC,IAAK,IAAI,EAAI,EAAe,EAAG,GAAK,EAAG,IAAK,CAC1C,IAAM,EAAY,KAAK,UAAU,GAG7B,KAAU,OAAS,QAMrB,EAAU,SAAW,EAAS,QAC9B,KAAK,kBAAkB,EAAU,SAAS,GAAK,KAAK,kBAAkB,EAAS,SAAS,CAQxF,OALK,KAAK,cAAc,EAAG,EAAa,CAKjC,CACL,WAAY,EACZ,WAAY,IACb,CAPQ,KAAK,oBAAoB,2CAA2C,CAWjF,OAAO,KAAK,oBAAoB,uCAAuC,CAQzE,kBAA0B,EAAoB,EAAqB,CAEjE,GAAI,EAAS,OAAS,SAGhB,EAAS,qBACX,KAAK,QAAQ,oBAAoB,IAAI,EAAS,oBAAqB,EAAM,CAEvE,EAAS,qBACX,KAAK,QAAQ,oBAAoB,IAAI,EAAS,oBAAqB,EAAM,CAIvE,CAAC,EAAS,qBAAuB,CAAC,EAAS,qBAAqB,CAClE,IAAM,EAAY,KAAK,iBAAiB,EAAS,CACjD,GAAI,EAAW,CACb,IAAM,EAAa,KAAK,mBAAmB,EAAU,CACrD,KAAK,QAAQ,oBAAoB,IAAI,EAAY,EAAM,GAU/D,iBAAyB,EAAgD,CAKvE,IAAM,EAAgB,EAAS,KAAK,cAE9B,EAAgB,KAAK,IAAI,EAAG,EAAgB,IAAI,CAChD,EAAa,KAAK,KAAK,UAAU,EAAe,EAAc,CAI9D,EAAS,EAAW,MAAM,4FAA4F,CAC5H,GAAI,EACF,OAAO,KAAK,iBAAiB,EAAO,GAAG,MAAM,CAAC,CAIhD,IAAM,EAAc,EAAW,MAAM,8CAA8C,CACnF,GAAI,EACF,OAAO,KAAK,iBAAiB,EAAY,GAAG,MAAM,CAAC,CAUvD,iBAAyB,EAAsB,CAC7C,IAAM,EAAW,EAAK,QAAQ,iFAAkF,GAAG,CAAC,MAAM,CAE1H,OAAO,EAAS,OAAS,EAAI,EAAW,EAM1C,mBAA2B,EAAsB,CAC/C,OAAO,EACJ,aAAa,CACb,QAAQ,OAAQ,IAAI,CACpB,MAAM,CAMX,kBAA0B,EAA0B,CAClD,OAAO,EACJ,aAAa,CACb,QAAQ,OAAQ,GAAG,CACnB,QAAQ,MAAO,GAAG,CAMvB,cAAsB,EAAyB,EAA+B,CAC5E,OAAO,EACL,EACA,EACA,KAAK,QAAQ,aACb,KAAK,QAAQ,cACd,CAMH,oBAA4B,EAA8C,CACxE,GAAI,KAAK,QAAQ,iBACf,MAAO,CACL,WAAY,IAAA,GACZ,cAAe,EACf,WAAY,EACb,GClUP,SAAgB,EACd,EACA,EACA,EACoB,CAEpB,OADiB,IAAI,EAAiB,EAAW,EAAM,EAAQ,CAC/C,SAAS,CCkB3B,SAAgB,GACf,EACA,EAAc,GACU,CACxB,IAAM,EAAiB,IAAI,IAG3B,GAAI,EAAO,SAAW,GAAK,IAAgB,GAC1C,OAAO,EAIR,IAAM,EAAkB,IAAI,IAE5B,IAAK,IAAI,EAAI,EAAG,EAAI,EAAO,OAAQ,IAAK,CACvC,IAAM,EAAU,EAAO,GAQvB,GALI,EAAQ,OAAS,QAKjB,EAAgB,IAAI,EAAE,CACzB,SAGD,IAAM,EAA6B,EAAE,CAIrC,IAAK,IAAI,EAAI,EAAI,EAAG,EAAI,EAAO,OAAQ,IAAK,CAC3C,IAAM,EAAY,EAAO,GAGzB,GAAI,EAAU,OAAS,OACtB,MAKD,IAAM,GADY,IAAM,EAAI,EAAI,EAAU,EAAO,EAAI,IAC1B,KAAK,SAC1B,EAAS,EAAU,KAAK,WAK9B,GADgB,EAAS,EACX,GACb,MAID,IAAM,EAAU,EAAY,UAAU,EAAU,EAAO,CAGvD,GAAI,CAAC,EAAQ,SAAS,IAAI,CACzB,MAKD,IAAM,EAAa,EAAQ,QAAQ,IAAI,CAiBvC,GAhB2B,EAAQ,OAAS,EAAa,EAEhC,GAQL,EAAY,UAAU,EAAQ,KAAK,SAAU,EAAU,KAAK,SAAS,CACzE,SAAS,IAAI,EAKzB,CAAC,GAAuB,EAAa,EAAU,KAAK,SAAS,CAChE,MAID,EAAiB,KAAK,EAAE,CACxB,EAAgB,IAAI,EAAE,CAInB,EAAiB,OAAS,GAC7B,EAAe,IAAI,EAAG,EAAiB,CAIzC,OAAO,EAaR,SAAS,GAAuB,EAAqB,EAA2B,CAE/E,IAAM,EAAa,EAAY,UAAU,EAAU,EAAW,IAAI,CAG5D,EAAY,EAAW,QAAQ,IAAI,CACzC,GAAI,IAAc,GACjB,MAAO,GAIR,IAAI,EAAQ,EACZ,IAAK,IAAI,EAAI,EAAW,EAAI,EAAW,OAAQ,IAC9C,GAAI,EAAW,KAAO,IACrB,YACU,EAAW,KAAO,MAC5B,IACI,IAAU,GAEb,MAAO,GAKV,MAAO,GCrBR,SAAgB,EACf,EACA,EACkC,CAClC,IAAM,EAAY,YAAY,KAAK,CAG7B,CAAE,UAAS,oBAAmB,YAAa,EAChD,EACA,GAAS,SACT,CAWK,EAAS,EAAS,EAPJ,GAAS,UAAY,CACxC,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACH,GAAG,EACH,CAC4C,CAKvC,EAAoC,EAAE,CACtC,EAAgB,IAAI,IAIpB,EAAgB,EAAQ,OAAS,MAEvC,IAAK,IAAM,KAAS,EAAQ,CAC3B,IAAM,EAAS,EACX,EAAM,KAAK,YAAc,GAAM,EAAM,KAAK,SAC3C,GAAG,EAAM,KAAK,WAAW,GAAG,EAAM,KAAK,WACrC,EAAc,IAAI,EAAO,GAC7B,EAAc,IAAI,EAAO,CACzB,EAAmB,KAAK,EAAM,EAMhC,IAAM,EAAiB,GAAwB,EAAoB,EAAQ,CAGrE,EAAwB,EAAE,CAChC,IAAK,IAAI,EAAI,EAAG,EAAI,EAAmB,OAAQ,IAAK,CACnD,IAAM,EAAQ,EAAmB,GAC7B,EAEJ,OAAQ,EAAM,KAAd,CACC,IAAK,OAEJ,AAOC,EAPG,EAAM,YAAc,MAAQ,EAAM,YAAc,OACxC,EAAU,EAAO,EAAkB,CACpC,EAAM,YAAc,QACnB,EAAa,EAAO,EAAkB,CACvC,EAAM,YAAc,gBACnB,EAAqB,EAAO,EAAkB,CAE9C,EAAY,EAAO,EAAmB,EAAQ,CAE1D,MACD,IAAK,UACJ,EAAW,EAAe,EAAO,EAAkB,CACnD,MACD,IAAK,UACJ,EAAW,EAAe,EAAO,EAAkB,CACnD,MACD,IAAK,UACJ,EAAW,EAAe,EAAO,EAAkB,CACnD,MACD,IAAK,YACJ,EAAW,EAAiB,EAAO,EAAkB,CACrD,MACD,IAAK,kBACJ,EAAW,EAAuB,EAAO,EAAkB,CAC3D,MACD,IAAK,kBACJ,EAAW,EAAuB,EAAO,EAAkB,CAC3D,MACD,QAEC,SAYF,GARI,EAAS,OAAS,IACrB,EAAS,SAAW,CAAC,GAAI,EAAS,UAAY,EAAE,CAAG,GAAG,EAAS,EAIhE,EAAS,cAAgB,YAAY,KAAK,CAAG,EAGzC,EAAS,OAAS,OAAQ,CAE7B,IAAM,EAAY,EAAe,IAAI,EAAE,CACjC,EAAc,MAAM,KAAK,EAAe,QAAQ,CAAC,CAAC,KAAK,GAAe,EAAY,SAAS,EAAE,CAAC,CAEpG,GAAI,GAAa,EAAa,CAE7B,IAAI,EAAe,EACnB,GAAI,OAEE,GAAM,CAAC,EAAS,KAAgB,EAAe,SAAS,CAC5D,GAAI,EAAY,SAAS,EAAE,CAAE,CAC5B,EAAe,EACf,OAMH,IAAM,EAAe,EAAmB,GAGlC,EAAQ,yBAAyB,KAAK,EAAa,KAAK,CAC9D,GAAI,EAAO,CACV,GAAM,EAAG,EAAQ,EAAU,GAAQ,EAInC,GAHA,EAAS,QAAU,GAAG,EAAO,GAAG,EAAS,QAAQ,OAAQ,IAAI,CAAC,GAAG,IAG7D,EAAW,CACd,IAAM,EAAmB,EAAe,IAAI,EAAE,CAC9C,EAAS,kBAAoB,EAAiB,IAAI,GAAU,CAC3D,IAAM,EAAW,EAAmB,GAC9B,EAAW,yBAAyB,KAAK,EAAS,KAAK,CAC7D,GAAI,EAAU,CACb,GAAM,EAAG,EAAQ,EAAQ,GAAW,EACpC,MAAO,CACN,OAAQ,QAAQ,KAAK,EAAO,CAAG,OAAO,SAAS,EAAQ,GAAG,CAAG,EAC7D,SAAU,EACV,KAAM,OAAO,SAAS,EAAS,GAAG,CAClC,CAEF,MAAO,CAAE,OAAQ,EAAG,SAAU,GAAI,KAAM,EAAG,EAC1C,IAMN,EAAU,KAAK,EAAS,CAQzB,OAJI,GAAS,QACL,EAAiB,EAAW,EAAM,EAAQ,kBAAkB,CAG7D,EA2BR,eAAsB,GACrB,EACA,EAC2C,CAG3C,OAAO,EAAiB,EAAM,EAAQ"}