code-gauge 3.1.0 → 4.1.0

package/dist/duplication.js.map

@@ -1 +1 @@
-{"version":3,"file":"duplication.js","names":[],"sources":["../src/duplication.ts"],"sourcesContent":["import type Parser from 'tree-sitter';\nimport { dedupeByRegion, selectMaximalGroups } from './duplicateSelection.js';\nimport type { DuplicationMetrics, DuplicationOptions } from './types.js';\n\n/**\n * Block-like nodes considered as whole-subtree duplicate candidates. Detection itself is\n * token-based, so this set only decides which subtrees are compared; Ruby's keyword-like node\n * types (`if`, `case`, ...) are safe here because only named nodes become candidates.\n */\nconst duplicateBlockTypes = new Set([\n  'statement_block',\n  'block',\n  'compound_statement',\n  'body_statement',\n  'constructor_body',\n  'do_block',\n  'if_statement',\n  'for_statement',\n  'for_in_statement',\n  'enhanced_for_statement',\n  'for_range_loop',\n  'while_statement',\n  'do_statement',\n  'try_statement',\n  'try_with_resources_statement',\n  'with_statement',\n  'switch_statement',\n  'switch_expression',\n  'switch_case',\n  'switch_block_statement_group',\n  'switch_rule',\n  'case_clause',\n  'case_statement',\n  'match_statement',\n  'match_arm',\n  'except_clause',\n  'catch_clause',\n  'finally_clause',\n  'elif_clause',\n  'ensure',\n  'expression_statement',\n  'return_statement',\n  'return_expression',\n  'if_expression',\n  'for_expression',\n  'while_expression',\n  'loop_expression',\n  'match_expression',\n  'jsx_element',\n  'jsx_self_closing_element',\n  // Ruby\n  'if',\n  'unless',\n  'case',\n  'case_match',\n  'while',\n  'until',\n  'for',\n  'begin',\n  'when',\n]);\n\n/** Nodes whose direct named children form statement sequences scanned for copy-pasted runs. */\nconst statementContainerTypes = new Set([\n  'program',\n  'source_file',\n  'translation_unit',\n  'module',\n  'statement_block',\n  'block',\n  'compound_statement',\n  'body_statement',\n  'constructor_body',\n  'class_body',\n  'block_body',\n  'do_block',\n  // Ruby loop bodies are a named `do` node, and `ensure` holds statements directly.\n  'do',\n  'ensure',\n  'then',\n  'else',\n  // Case-like nodes hold their statements directly, without an inner block.\n  'case_statement',\n  'switch_block_statement_group',\n  'switch_rule',\n  'expression_case',\n  'type_case',\n  'communication_case',\n  'default_case',\n]);\n\n/**\n * Identifier leaves anonymized by occurrence order so consistently renamed copies still match.\n * Member/type names (`property_identifier`, `field_identifier`, `type_identifier`, ...) are kept\n * verbatim instead: calling a different API is a semantic difference, not a rename.\n */\nconst anonymizedIdentifierTypes = new Set([\n  'identifier',\n  'constant',\n  'instance_variable',\n  'class_variable',\n  'global_variable',\n]);\n\n/**\n * JS shorthand properties (`{ alpha }`) both emit the property name (semantic output shape) and\n * reference the binding, so they tokenize as the desugared `name: binding` — one verbatim text\n * token plus one anonymized id token — matching how the explicit form is tokenized.\n */\nconst shorthandPropertyTypes = new Set(['shorthand_property_identifier', 'shorthand_property_identifier_pattern']);\n\n/** Literal leaves normalized to a kind tag so copies differing only in literal values still match. */\nconst literalKindByType = new Map([\n  ['number', '#num'],\n  ['number_literal', '#num'],\n  ['integer', '#num'],\n  ['float', '#num'],\n  ['integer_literal', '#num'],\n  ['float_literal', '#num'],\n  ['int_literal', '#num'],\n  ['rune_literal', '#char'],\n  ['imaginary_literal', '#num'],\n  ['decimal_integer_literal', '#num'],\n  ['hex_integer_literal', '#num'],\n  ['octal_integer_literal', '#num'],\n  ['binary_integer_literal', '#num'],\n  ['decimal_floating_point_literal', '#num'],\n  ['hex_floating_point_literal', '#num'],\n  ['string_fragment', '#str'],\n  ['multiline_string_fragment', '#str'],\n  ['string_content', '#str'],\n  ['raw_string_content', '#str'],\n  ['heredoc_content', '#str'],\n  // Heredoc marker names (`<<~SQL` vs `<<~QUERY`) have no string-value significance.\n  ['heredoc_beginning', '#heredoc'],\n  ['heredoc_end', '#heredoc'],\n  // Strings are leaves in some grammars (Go/Rust) and fragment containers in others.\n  ['string', '#str'],\n  ['template_string', '#str'],\n  ['string_literal', '#str'],\n  ['interpreted_string_literal', '#str'],\n  ['raw_string_literal', '#str'],\n  ['raw_string', '#str'],\n  ['escape_sequence', '#str'],\n  ['char_literal', '#char'],\n  ['character_literal', '#char'],\n  ['character', '#char'],\n  ['regex_pattern', '#regex'],\n]);\n\n/**\n * Kind tags whose raw source text re-enters the fingerprint in literal-dense (data-like) regions.\n * `#heredoc` is excluded: heredoc marker names are naming choices, not data values.\n */\nconst valueCarryingLiteralKinds = new Set(['#num', '#str', '#char', '#regex']);\n\nconst commentTypes = new Set(['comment', 'line_comment', 'block_comment']);\n\n/**\n * String children that carry actual content, i.e. stringFragmentTypes minus the delimiter nodes\n * (Python's `string_start`/`string_end`), for delimiter-independent literal values.\n */\nconst stringContentFragmentTypes = new Set([\n  'string_fragment',\n  'multiline_string_fragment',\n  'string_content',\n  'raw_string_content',\n  'escape_sequence',\n  'heredoc_content',\n]);\n\n/** Children of a string node that carry only literal content; anything else is interpolation. */\nconst stringFragmentTypes = new Set([\n  'string_fragment',\n  'multiline_string_fragment',\n  'string_content',\n  'raw_string_content',\n  'escape_sequence',\n  'heredoc_content',\n  // Python string delimiters are named children; they never carry interpolation.\n  'string_start',\n  'string_end',\n]);\n\n/**\n * Where a grammar names callees/members with a plain `identifier` (Java `method_invocation.name`,\n * Ruby `call.method`, Python `attribute.attribute`, plain calls elsewhere), the leaf in that field\n * must stay verbatim like `property_identifier` does: calling a different API is a semantic\n * difference, not a rename.\n */\nconst semanticNameFieldByParentType = new Map([\n  ['call_expression', 'function'],\n  ['method_invocation', 'name'],\n  ['call', 'method'],\n  ['attribute', 'attribute'],\n  ['macro_invocation', 'macro'],\n  // Java names accessed fields with a plain identifier in the `field` field.\n  ['field_access', 'field'],\n  // JS/TS `new Foo(...)` names the constructed API in the `constructor` field.\n  ['new_expression', 'constructor'],\n  // Python `f(timeout=...)` and Java `@Anno(key=...)` name parameters of the callee's API.\n  ['keyword_argument', 'name'],\n  ['element_value_pair', 'key'],\n  // Rust turbofish and C++ template callees (`compute::<u32>(...)`); type arguments stay\n  // anonymized via their own node types.\n  ['generic_function', 'function'],\n  ['template_function', 'name'],\n]);\n\nexport const defaultDuplicationOptions: Required<DuplicationOptions> = {\n  minTokens: 40,\n  maxGapTokens: 30,\n  minSimilarityPercent: 70,\n};\n\n/**\n * N-gram size for the near-miss candidate index. 5 is NIL's published default (Nakagawa et al.,\n * ESEC/FSE 2021): short enough that edited clones still share many n-grams, long enough that\n * unrelated blocks rarely collide.\n */\nconst nearMissNgramSize = 5;\n/**\n * Blocks sharing fewer than this percentage of the smaller block's distinct n-grams skip LCS\n * verification entirely (NIL's filtration phase, default 10%): a pair below it cannot reach any\n * useful similarity, and the cheap set-overlap check prunes the quadratic candidate space.\n */\nconst nearMissFiltrationPercent = 10;\n/**\n * Token-level LCS over normalized code is dominated by punctuation and keywords, so a structural\n * 70% match alone is weak evidence of copying: two same-skeleton functions calling entirely\n * different APIs can exceed it. A near-miss pair must therefore also share at least half of the\n * larger block's content-bearing tokens (verbatim-kept names and literal values — the tokens that\n * distinguish WHAT the code does rather than how it is shaped). The bound is exclusive: a\n * value-mapping predicate pair (`x.type === 'a' || ...` with equal branch counts and zero shared\n * values) shares exactly half its content — the repeated member name — and must not pass.\n */\nconst minContentSimilarityPercent = 50;\n\n/** Minimum consecutive statements for a statement-sequence duplicate candidate. */\nconst minSequenceStatementCount = 2;\n/**\n * Caps the window length so statement-sequence enumeration stays linear in the statement count.\n * Heterogeneous clones longer than the cap are reported as capped windows (a deliberate\n * conservative undercount trading completeness for bounded discovery cost).\n */\nconst maxSequenceStatementCount = 100;\n\n/**\n * A region whose normalized tokens are at least 20% literal values is data-like (a lookup table, a\n * constant list, a value-mapping switch), not logic: literal values re-enter its fingerprint so\n * tables that merely share their shape stop counting as copy-paste. Logic-heavy code sits well\n * below the bound (5-10% literals) while object/array tables sit above it (25-50%); punctuation\n * and member names dilute tables, which is why the bound is far below half. Compared in integer\n * math (5 * literals >= total) so the TypeScript and native backends cannot disagree on the\n * boundary.\n */\nfunction isLiteralDense(literalCount: number, tokenCount: number): boolean {\n  return literalCount * 5 >= tokenCount;\n}\n\nexport interface Token {\n  /** Normalization target: identifiers to anonymize, literal kind tags, or the raw token text. */\n  kind: 'id' | 'text';\n  text: string;\n  /**\n   * Two INDEPENDENT hashes of `text` (djb2 and FNV-1a), precomputed so fingerprinting nested\n   * regions never re-hashes a token. Feeding the same per-token hash to both fingerprint\n   * accumulators would collapse the key to 32 effective bits: one djb2 collision between two\n   * token texts would then equate whole regions.\n   */\n  textHash: number;\n  textHash2: number;\n  /** Hash pair of a value-carrying literal's value, folded into data-like region fingerprints. */\n  literalHash?: number;\n  literalHash2?: number;\n  /**\n   * True for verbatim-kept NAMES (member/callee/type names, named grammar leaves): together with\n   * value-carrying literals these are the content-bearing tokens the near-miss content gate\n   * counts. Keywords, operators, and punctuation come from unnamed nodes and stay false.\n   */\n  isName?: boolean;\n  /** 0-based source rows the token occupies, so line coverage counts only matched-token lines. */\n  startRow: number;\n  endRow: number;\n}\n\n/** A token span with its source position; plain data so cross-file matching can retain it. */\nexport interface TokenRange {\n  startTokenIndex: number;\n  endTokenIndex: number;\n  startIndex: number;\n  endIndex: number;\n  startLine: number;\n  endLine: number;\n}\n\n/** A contiguous run of matched tokens; gapped (merged) duplicates carry several per occurrence. */\ninterface TokenSegment {\n  startTokenIndex: number;\n  endTokenIndex: number;\n}\n\ninterface DuplicateCandidate {\n  fingerprint: string;\n  tokenCount: number;\n  startTokenIndex: number;\n  endTokenIndex: number;\n  startIndex: number;\n  endIndex: number;\n  startLine: number;\n  endLine: number;\n}\n\nexport interface CountedOccurrence {\n  /** Matched token runs; more than one once gapped groups are merged. */\n  segments: TokenSegment[];\n  /**\n   * Set on a retained group's occurrences that a partial gapped merge also paired into a merged\n   * group: their spans are counted there, so block counting must not count them again.\n   */\n  sharedWithMergedGroup?: boolean;\n  /** Sum of segment token counts (the gap tokens are not matched content). */\n  tokenCount: number;\n  startTokenIndex: number;\n  endTokenIndex: number;\n  startIndex: number;\n  endIndex: number;\n  startLine: number;\n  endLine: number;\n}\n\n/** A duplicate region found in one file, exported for cross-file matching by fingerprint. */\nexport interface CrossFileDuplicateCandidate {\n  /** Content key: equal fingerprints mean equal normalized token sequences (up to hash collision). */\n  fingerprint: string;\n  tokenCount: number;\n  /** Token positions within the owning file, for cross-file gapped (Type-3) merging. */\n  startTokenIndex: number;\n  endTokenIndex: number;\n  startIndex: number;\n  endIndex: number;\n  startLine: number;\n  endLine: number;\n}\n\n/**\n * One file's contribution to cross-file clone detection: catalogued candidates plus the normalized\n * token stream and statement structure, so the project-level pass can match partial statement runs\n * (windows that a single file cannot know repeat elsewhere) and merge gap-adjacent groups.\n */\nexport interface CrossFileDuplicationFileData {\n  candidates: CrossFileDuplicateCandidate[];\n  tokens: Token[];\n  containerStatements: TokenRange[][];\n  /**\n   * 1-based lines that are neither blank nor comment-only, so cross-file line coverage counts only\n   * code lines (blank rows inside multi-row tokens such as template literals carry no content).\n   * Optional for backward compatibility; without it, every matched-token row counts.\n   */\n  codeLineNumbers?: Set<number>;\n}\n\n/**\n * Detects copy-pasted regions within a file. Regions are compared by their normalized token\n * sequence: identifiers are anonymized consistently by first-occurrence order (`a.f(a, b)` matches\n * `x.f(x, y)` but not `x.f(y, z)`), literals are normalized by kind, and member/type names and all\n * keywords/operators are kept verbatim. Literal-dense (data-like) regions additionally require\n * equal literal values. Candidates are whole block-like subtrees plus runs of consecutive sibling\n * statements, so a copy pasted into the middle of a longer block is still found. Only maximal,\n * non-overlapping regions are counted, and adjacent groups separated by a small token gap merge\n * into one gapped (Type-3) clone group. Blocks the exact pipeline misses are additionally compared\n * by similarity (n-gram filtration then token-level LCS, NiCad/NIL-style), so near-miss (Type-3)\n * clones whose edits fragment them below `minTokens` are still reported.\n */\nexport function measureDuplication(\n  root: Parser.SyntaxNode,\n  codeLineNumbers: Set<number>,\n  options?: DuplicationOptions\n): DuplicationMetrics {\n  const minTokens = options?.minTokens ?? defaultDuplicationOptions.minTokens;\n  const maxGapTokens = options?.maxGapTokens ?? defaultDuplicationOptions.maxGapTokens;\n  const minSimilarityPercent = options?.minSimilarityPercent ?? defaultDuplicationOptions.minSimilarityPercent;\n  const tokens: Token[] = [];\n  const blockRanges: TokenRange[] = [];\n  const containerStatementRanges: TokenRange[][] = [];\n  collectTokens(root, tokens, blockRanges, containerStatementRanges);\n  const literalCountPrefix = buildLiteralCountPrefix(tokens);\n\n  const candidates = [\n    ...collectBlockCandidates(tokens, literalCountPrefix, blockRanges, minTokens),\n    ...collectSequenceCandidates(tokens, literalCountPrefix, containerStatementRanges, minTokens),\n  ];\n  const counted = selectMaximalGroups(candidates, (group) => group.length >= 2);\n  const groups = mergeAdjacentGroups(toCountedGroups(counted), maxGapTokens);\n  const nearMissGroups = collectNearMissGroups(\n    tokens,\n    literalCountPrefix,\n    blockRanges,\n    minTokens,\n    minSimilarityPercent,\n    groups\n  );\n  // Near-miss clustering can merge exact groups away, leaving empty entries behind.\n  const reportedGroups = [...groups.filter((group) => group.length > 0), ...nearMissGroups];\n  return summarizeDuplicates(reportedGroups, codeLineNumbers, tokens);\n}\n\n/**\n * Collects this file's contribution to cross-file clone detection: fingerprinted candidates for\n * whole block-like subtrees plus each statement container's full run (so wholly copied files and\n * class bodies match even when no inner block clears the threshold on its own), together with the\n * normalized token stream and statement structure that let measureCrossFileDuplication match\n * partial statement runs and merge gap-adjacent groups project-wide. Nested and overlapping\n * candidates are all returned; the project-level selection keeps only maximal ones.\n */\nexport function collectCrossFileDuplicateCandidates(\n  root: Parser.SyntaxNode,\n  options?: DuplicationOptions\n): CrossFileDuplicationFileData {\n  const minTokens = options?.minTokens ?? defaultDuplicationOptions.minTokens;\n  const tokens: Token[] = [];\n  const blockRanges: TokenRange[] = [];\n  const containerStatementRanges: TokenRange[][] = [];\n  collectTokens(root, tokens, blockRanges, containerStatementRanges);\n  const literalCountPrefix = buildLiteralCountPrefix(tokens);\n\n  const candidates = collectBlockCandidates(tokens, literalCountPrefix, blockRanges, minTokens);\n  for (const statements of containerStatementRanges) {\n    const first = statements[0];\n    const last = statements.at(-1);\n    if (!first || !last) {\n      continue;\n    }\n    const tokenCount = last.endTokenIndex - first.startTokenIndex;\n    if (tokenCount < minTokens) {\n      continue;\n    }\n    candidates.push(\n      toCandidate(\n        `s:${fingerprintKey(tokens, literalCountPrefix, first.startTokenIndex, last.endTokenIndex)}`,\n        first.startTokenIndex,\n        last.endTokenIndex,\n        first,\n        last\n      )\n    );\n  }\n  return { candidates: dedupeByRegion(candidates), tokens, containerStatements: containerStatementRanges };\n}\n\nfunction collectTokens(\n  root: Parser.SyntaxNode,\n  tokens: Token[],\n  blockRanges: TokenRange[],\n  containerStatementRanges: TokenRange[][]\n): void {\n  function visit(node: Parser.SyntaxNode): TokenRange {\n    const startTokenIndex = tokens.length;\n    const atomicKind = node.childCount === 0 ? undefined : atomicLiteralKind(node);\n    if (node.childCount === 0) {\n      appendLeafToken(node, tokens);\n    } else if (atomicKind !== undefined) {\n      // Interpolation-free strings collapse to their kind tag so copies differing only in quote\n      // style or content still match; delimiter tokens would otherwise break the equivalence.\n      tokens.push(\n        makeTextToken(atomicKind, literalValueText(node, atomicKind), node.startPosition.row, node.endPosition.row)\n      );\n    } else if (!commentTypes.has(node.type)) {\n      const statementRanges: TokenRange[] = [];\n      const isContainer = node.isNamed && statementContainerTypes.has(node.type);\n      for (const child of node.children) {\n        const childRange = visit(child);\n        if (isContainer && child.isNamed && !commentTypes.has(child.type)) {\n          statementRanges.push(childRange);\n        }\n      }\n      // Single-statement containers are recorded too: within-file window enumeration needs two\n      // statements and simply yields nothing for them, but cross-file matching must still see a\n      // file whose only top-level statement is not a catalogued block type (a lone exported table).\n      if (isContainer && statementRanges.length > 0) {\n        containerStatementRanges.push(statementRanges);\n      }\n    }\n\n    const range = {\n      startTokenIndex,\n      endTokenIndex: tokens.length,\n      startIndex: node.startIndex,\n      endIndex: node.endIndex,\n      startLine: node.startPosition.row + 1,\n      endLine: node.endPosition.row + 1,\n    };\n    if (node.isNamed && duplicateBlockTypes.has(node.type)) {\n      blockRanges.push(range);\n    }\n    return range;\n  }\n\n  visit(root);\n}\n\n/** The kind tag of a string-like node with no interpolation, or undefined to descend normally. */\nfunction atomicLiteralKind(node: Parser.SyntaxNode): string | undefined {\n  const kind = node.isNamed ? literalKindByType.get(node.type) : undefined;\n  if (kind === undefined) {\n    return undefined;\n  }\n  return node.namedChildren.every((child) => stringFragmentTypes.has(child.type)) ? kind : undefined;\n}\n\nfunction appendLeafToken(node: Parser.SyntaxNode, tokens: Token[]): void {\n  if (commentTypes.has(node.type)) {\n    return;\n  }\n\n  const startRow = node.startPosition.row;\n  const endRow = node.endPosition.row;\n  if (node.isNamed && shorthandPropertyTypes.has(node.type)) {\n    tokens.push(\n      makeTextToken(node.text, undefined, startRow, endRow, true),\n      makeTextToken(':', undefined, startRow, endRow),\n      { kind: 'id', text: node.text, textHash: 0, textHash2: 0, startRow, endRow }\n    );\n    return;\n  }\n\n  if (node.isNamed && anonymizedIdentifierTypes.has(node.type) && !isSemanticNameLeaf(node)) {\n    tokens.push({ kind: 'id', text: node.text, textHash: 0, textHash2: 0, startRow, endRow });\n    return;\n  }\n\n  // Anything else keeps its text: keywords, operators, punctuation, and semantic names such as\n  // `property_identifier`/`type_identifier`, which must distinguish otherwise-identical structures.\n  const literalKind = node.isNamed ? literalKindByType.get(node.type) : undefined;\n  if (literalKind === undefined) {\n    tokens.push(makeTextToken(node.text, undefined, startRow, endRow, node.isNamed));\n  } else {\n    tokens.push(makeTextToken(literalKind, literalValueText(node, literalKind), startRow, endRow));\n  }\n}\n\nfunction makeTextToken(\n  text: string,\n  literalValueText: string | undefined,\n  startRow: number,\n  endRow: number,\n  isName = false\n): Token {\n  const token: Token = { kind: 'text', text, textHash: hashText(text), textHash2: hashText2(text), startRow, endRow };\n  if (literalValueText !== undefined && valueCarryingLiteralKinds.has(text)) {\n    token.literalHash = hashText(literalValueText);\n    token.literalHash2 = hashText2(literalValueText);\n  }\n  if (isName) {\n    token.isName = true;\n  }\n  return token;\n}\n\n/**\n * The value of a literal as folded into literal-dense fingerprints. Strings hash their CONTENT,\n * not their source spelling: formatters rewrite quote style on paste (`'one'` vs `\"one\"`), so\n * delimiters must not make two copied tables differ. Content comes from the fragment children when\n * the grammar provides them (which also drops Python's `string_start`/`string_end` delimiter\n * nodes), else from the text with one matching pair of surrounding quotes stripped. Numbers keep\n * their raw text: formatters preserve numeric spelling, and canonicalizing values (`0x10` vs `16`)\n * identically in JavaScript and Rust would be far riskier than the rare mismatch it would unify.\n */\nfunction literalValueText(node: Parser.SyntaxNode, kind: string): string {\n  if (kind !== '#str' && kind !== '#char') {\n    return node.text;\n  }\n  // Fragment leaves (string_fragment, escape_sequence, heredoc_content, ...) already carry bare\n  // content; a quote appearing there is content, not a delimiter.\n  if (stringContentFragmentTypes.has(node.type)) {\n    return node.text;\n  }\n  const fragments = node.namedChildren.filter((child) => stringContentFragmentTypes.has(child.type));\n  if (fragments.length > 0) {\n    return fragments.map((child) => child.text).join('');\n  }\n  return stripMatchingQuotes(node.text);\n}\n\nconst quoteCharacters = new Set(['\"', \"'\", '`']);\n\nfunction stripMatchingQuotes(text: string): string {\n  const first = text[0];\n  return text.length >= 2 && first !== undefined && quoteCharacters.has(first) && text.endsWith(first)\n    ? text.slice(1, -1)\n    : text;\n}\n\n/** literalCountPrefix[i] = value-carrying literal tokens in tokens[0..i), for O(1) density checks. */\nexport function buildLiteralCountPrefix(tokens: Token[]): Int32Array {\n  const prefix = new Int32Array(tokens.length + 1);\n  for (const [index, token] of tokens.entries()) {\n    prefix[index + 1] = (prefix[index] ?? 0) + (token.literalHash === undefined ? 0 : 1);\n  }\n  return prefix;\n}\n\nfunction isSemanticNameLeaf(node: Parser.SyntaxNode): boolean {\n  const parent = node.parent;\n  if (!parent) {\n    return false;\n  }\n\n  // Java method references (`Foo::bar`) name their identifiers without grammar fields; both the\n  // type/object and the referenced method are semantic.\n  if (parent.type === 'method_reference') {\n    return true;\n  }\n\n  // `call` names its callee `method` in Ruby but `function` in Python; accept both fields.\n  if (parent.type === 'call' && parent.childForFieldName('function')?.id === node.id) {\n    return true;\n  }\n\n  // A Ruby constant receiving a call (`Alpha.new(...)`) names the invoked API; constants used as\n  // plain values stay anonymized so renamed clones referencing constants still match.\n  if (node.type === 'constant' && parent.type === 'call' && parent.childForFieldName('receiver')?.id === node.id) {\n    return true;\n  }\n\n  // Java static receivers (`Alpha.run(...)`) name the invoked type. The tokenizer has no symbol\n  // table, so PascalCase — Java's universal type-naming convention — is the discriminator;\n  // camelCase instance receivers stay anonymized for rename tolerance.\n  if (\n    parent.type === 'method_invocation' &&\n    parent.childForFieldName('object')?.id === node.id &&\n    /^\\p{Lu}/u.test(node.text)\n  ) {\n    return true;\n  }\n\n  // Qualified callees (Rust `crate::alpha::make(...)`, C++ `detail::make(...)`) and generic\n  // callees (`compute::<u32>(...)`, `compute<int>(...)`) wrap their identifiers arbitrarily deep;\n  // every path/name segment is semantic there — but only in call position, so renamed clones that\n  // merely reference scoped constants or `use` paths still match.\n  if (\n    (parent.type === 'scoped_identifier' || parent.type === 'qualified_identifier') &&\n    (parent.childForFieldName('name')?.id === node.id || parent.childForFieldName('path')?.id === node.id)\n  ) {\n    let outer = parent;\n    while (\n      outer.parent &&\n      (outer.parent.type === 'scoped_identifier' ||\n        outer.parent.type === 'qualified_identifier' ||\n        outer.parent.type === 'generic_function' ||\n        outer.parent.type === 'template_function')\n    ) {\n      outer = outer.parent;\n    }\n    if (outer.parent?.type === 'call_expression' && outer.parent.childForFieldName('function')?.id === outer.id) {\n      return true;\n    }\n  }\n\n  // Go struct-literal keys (`Config{Timeout: ...}`) have no `key` field in the grammar: the key\n  // is the keyed_element's first named child, a literal_element wrapping the identifier.\n  if (\n    parent.type === 'literal_element' &&\n    parent.parent?.type === 'keyed_element' &&\n    parent.parent.namedChild(0)?.id === parent.id\n  ) {\n    return true;\n  }\n\n  const field = semanticNameFieldByParentType.get(parent.type);\n  return field !== undefined && parent.childForFieldName(field)?.id === node.id;\n}\n\nfunction collectBlockCandidates(\n  tokens: Token[],\n  literalCountPrefix: Int32Array,\n  blockRanges: TokenRange[],\n  minTokens: number\n): DuplicateCandidate[] {\n  const candidates: DuplicateCandidate[] = [];\n  for (const range of blockRanges) {\n    const tokenCount = range.endTokenIndex - range.startTokenIndex;\n    if (tokenCount < minTokens) {\n      continue;\n    }\n    candidates.push(\n      toCandidate(\n        `b:${fingerprintKey(tokens, literalCountPrefix, range.startTokenIndex, range.endTokenIndex)}`,\n        range.startTokenIndex,\n        range.endTokenIndex,\n        range,\n        range\n      )\n    );\n  }\n  return candidates;\n}\n\ninterface WindowOccurrences {\n  count: number;\n  /** -1 once occurrences span more than one container. */\n  containerIndex: number;\n  /** -1 once occurrences span more than one context (file). */\n  contextIndex: number;\n  minStart: number;\n  maxStart: number;\n}\n\ninterface SequenceWindow {\n  containerIndex: number;\n  start: number;\n  length: number;\n}\n\nfunction collectSequenceCandidates(\n  tokens: Token[],\n  literalCountPrefix: Int32Array,\n  containers: TokenRange[][],\n  minTokens: number\n): DuplicateCandidate[] {\n  return collectSequenceWindowCandidates([{ tokens, literalCountPrefix, containers }], minTokens, false).map(\n    ({ candidate }) => candidate\n  );\n}\n\n/** One file's token stream and statement containers, as a window-matching context. */\nexport interface SequenceWindowContext {\n  tokens: Token[];\n  literalCountPrefix: Int32Array;\n  containers: TokenRange[][];\n}\n\nexport interface ContextualSequenceCandidate {\n  candidate: DuplicateCandidate;\n  contextIndex: number;\n}\n\n/**\n * Enumerates runs of consecutive sibling statements over one or more contexts (files). Every\n * container statement participates; only the window length is capped, so enumeration stays linear\n * in the statement count. Windows are grouped by a cheap rolling hash of per-statement\n * fingerprints, and only locally maximal repeated windows — those whose one-statement extensions\n * stop repeating — become candidates with an exact (window-consistent) fingerprint. Without the\n * maximality filter a degenerate file of near-identical statements would fingerprint every\n * sub-window of every repeated region. With `requireMultipleContexts` a window only counts as\n * repeated when its occurrences span at least two contexts (CPD-style cross-file matching): a\n * repeat confined to one file is that file's own concern, and emitting it here would flood the\n * project-level selection with unusable single-file groups.\n */\nexport function collectSequenceWindowCandidates(\n  contexts: SequenceWindowContext[],\n  minTokens: number,\n  requireMultipleContexts: boolean\n): ContextualSequenceCandidate[] {\n  const candidates: ContextualSequenceCandidate[] = [];\n  const contextIndexByContainer: number[] = [];\n  const containers: TokenRange[][] = [];\n  for (const [contextIndex, context] of contexts.entries()) {\n    for (const statements of context.containers) {\n      contextIndexByContainer.push(contextIndex);\n      containers.push(statements);\n    }\n  }\n  const contextAt = (containerIndex: number): SequenceWindowContext | undefined =>\n    contexts[contextIndexByContainer[containerIndex] ?? 0];\n  const occurrencesByWindowKey = new Map<number, WindowOccurrences>();\n  const containerWindows = containers.map((statements, containerIndex) =>\n    enumerateContainerWindows(contextAt(containerIndex)?.tokens ?? [], statements, minTokens)\n  );\n  for (const [containerIndex, windows] of containerWindows.entries()) {\n    const contextIndex = contextIndexByContainer[containerIndex] ?? 0;\n    for (const [start, row] of windows.windowKeysByStart.entries()) {\n      for (const windowKey of row) {\n        if (windowKey === undefined) {\n          continue;\n        }\n        const occurrences = occurrencesByWindowKey.get(windowKey);\n        if (occurrences) {\n          occurrences.count += 1;\n          if (occurrences.containerIndex !== containerIndex) {\n            occurrences.containerIndex = -1;\n          }\n          if (occurrences.contextIndex !== contextIndex) {\n            occurrences.contextIndex = -1;\n          }\n          occurrences.minStart = Math.min(occurrences.minStart, start);\n          occurrences.maxStart = Math.max(occurrences.maxStart, start);\n        } else {\n          occurrencesByWindowKey.set(windowKey, {\n            count: 1,\n            containerIndex,\n            contextIndex,\n            minStart: start,\n            maxStart: start,\n          });\n        }\n      }\n    }\n  }\n\n  // A window only \"repeats\" when two of its occurrences can coexist without overlapping: sliding\n  // matches inside a homogeneous run (start spread smaller than the window length) can never both\n  // be counted and must neither qualify a window nor dominate its sub-windows. Cross-context\n  // matching instead requires occurrences in two contexts, which coexist by construction.\n  const repeats = (windowKey: number | undefined, length: number): boolean => {\n    if (windowKey === undefined) {\n      return false;\n    }\n    const occurrences = occurrencesByWindowKey.get(windowKey);\n    if (occurrences === undefined || occurrences.count < 2) {\n      return false;\n    }\n    if (requireMultipleContexts) {\n      return occurrences.contextIndex === -1;\n    }\n    return occurrences.containerIndex === -1 || occurrences.maxStart - occurrences.minStart >= length;\n  };\n\n  // A window whose statements all share one normalized shape (sixteen `let x = 0;` declarations,\n  // a constant table) is a homogeneous preamble, not a copy-paste: requiring two distinct\n  // per-statement shapes keeps such runs out of duplicate groups and the duplication ratio.\n  const hasDistinctStatements = (window: SequenceWindow): boolean => {\n    const hashes = containerWindows[window.containerIndex]?.statementHashes ?? [];\n    const firstHash = hashes[window.start];\n    for (let index = window.start + 1; index < window.start + window.length; index += 1) {\n      if (hashes[index] !== firstHash) {\n        return true;\n      }\n    }\n    return false;\n  };\n\n  const maximalWindows: SequenceWindow[] = [];\n  for (const [containerIndex, windows] of containerWindows.entries()) {\n    for (const [start, row] of windows.windowKeysByStart.entries()) {\n      for (const [length, windowKey] of row.entries()) {\n        if (!repeats(windowKey, length) || !hasDistinctStatements({ containerIndex, start, length })) {\n          continue;\n        }\n        // Dominated windows are skipped: the one-statement extension also repeats, so a larger\n        // candidate covering this window exists.\n        const extendedRight = windows.windowKeysByStart[start]?.[length + 1];\n        const extendedLeft = windows.windowKeysByStart[start - 1]?.[length + 1];\n        if (repeats(extendedRight, length + 1) || repeats(extendedLeft, length + 1)) {\n          continue;\n        }\n        maximalWindows.push({ containerIndex, start, length });\n      }\n    }\n  }\n\n  // The rolling hash anonymizes identifiers per statement, so a window can look repeated coarsely\n  // while its exact (window-consistent) fingerprints differ, and a longer window's match can\n  // dominate sub-windows that other copies still need (three copies where only two extend one\n  // statement further). Every emitted window therefore exposes its repeating, unvisited\n  // sub-windows; `visited` bounds the worklist and lengths strictly decrease, so it terminates.\n  const visited = new Set(maximalWindows.map(windowId));\n  let frontier = maximalWindows;\n  while (frontier.length > 0) {\n    const emitted: SequenceWindow[] = [];\n    for (const window of frontier) {\n      const statements = containers[window.containerIndex];\n      const first = statements?.[window.start];\n      const last = statements?.[window.start + window.length - 1];\n      const context = contextAt(window.containerIndex);\n      if (!first || !last || !context) {\n        continue;\n      }\n      const fingerprint = `s:${fingerprintKey(context.tokens, context.literalCountPrefix, first.startTokenIndex, last.endTokenIndex)}`;\n      candidates.push({\n        candidate: toCandidate(fingerprint, first.startTokenIndex, last.endTokenIndex, first, last),\n        contextIndex: contextIndexByContainer[window.containerIndex] ?? 0,\n      });\n      emitted.push(window);\n    }\n    frontier = [];\n    for (const window of emitted) {\n      for (const start of [window.start, window.start + 1]) {\n        const subWindow = { containerIndex: window.containerIndex, start, length: window.length - 1 };\n        const subWindowKey = containerWindows[window.containerIndex]?.windowKeysByStart[start]?.[subWindow.length];\n        if (\n          visited.has(windowId(subWindow)) ||\n          !repeats(subWindowKey, subWindow.length) ||\n          !hasDistinctStatements(subWindow)\n        ) {\n          continue;\n        }\n        visited.add(windowId(subWindow));\n        frontier.push(subWindow);\n      }\n    }\n  }\n  return candidates;\n}\n\nfunction windowId(window: SequenceWindow): string {\n  return `${window.containerIndex}:${window.start}:${window.length}`;\n}\n\ninterface ContainerWindows {\n  /** windowKeysByStart[start][length] is the rolling-hash key of the window, or undefined if it is below the size thresholds. */\n  windowKeysByStart: (number | undefined)[][];\n  /** Per-statement fingerprint hashes, for the distinct-shape requirement on windows. */\n  statementHashes: number[];\n}\n\nfunction enumerateContainerWindows(tokens: Token[], statements: TokenRange[], minTokens: number): ContainerWindows {\n  const statementHashes = statements.map((statement) =>\n    fingerprintHash(tokens, statement.startTokenIndex, statement.endTokenIndex)\n  );\n  const windowKeysByStart: (number | undefined)[][] = [];\n  for (let start = 0; start < statements.length; start += 1) {\n    const row: (number | undefined)[] = [];\n    let hash = 5381;\n    let tokenCount = 0;\n    const maxEnd = Math.min(statements.length, start + maxSequenceStatementCount);\n    for (let end = start; end < maxEnd; end += 1) {\n      const statement = statements[end];\n      const statementHash = statementHashes[end];\n      if (!statement || statementHash === undefined) {\n        break;\n      }\n      hash = combineHashes(hash, statementHash);\n      tokenCount += statement.endTokenIndex - statement.startTokenIndex;\n      const statementCount = end - start + 1;\n      row[statementCount] =\n        statementCount >= minSequenceStatementCount && tokenCount >= minTokens\n          ? combineHashes(hash, statementCount)\n          : undefined;\n    }\n    windowKeysByStart.push(row);\n  }\n  return { windowKeysByStart, statementHashes };\n}\n\n/**\n * Detects near-miss (Type-3) clone groups among block candidates the exact pipeline left\n * unreported, following the locate-filter-verify design of NIL (ESEC/FSE 2021) with NiCad's\n * per-fragment similarity semantics: an inverted index of 5-grams over normalized tokens proposes\n * candidate pairs, a cheap shared-n-gram ratio prunes them, and token-level LCS verifies that the\n * pair is at least `minSimilarityPercent` similar relative to the LARGER block (so a small block\n * embedded in a big one does not count), with a content gate requiring shared names/literal values\n * so same-skeleton code calling different APIs does not pair on punctuation alone. Verified pairs\n * are clustered transitively; each cluster becomes one clone group whose occurrences span whole\n * blocks. Only blocks that are not literal-dense (the data-table guard) participate, and wrappers\n * containing several comparable sub-blocks (a `describe(...)` call, a class body) are descended\n * through. Blocks already covered by an exactly-reported region still take part as ANCHORS — two\n * identical copies plus one edited copy is the commonest copy-paste-then-edit shape, and without\n * anchors the exact pair would swallow both comparison partners of the edited copy — but they are\n * never re-reported: an edited copy verified against an anchor is appended to the anchor's\n * fully-clustered exact group, so no reported region ever overlaps another. Reported occurrences\n * cover their whole block: unlike exact matches, a near-miss occurrence includes its edited\n * tokens, which is how NiCad-style detectors report Type-3 fragments.\n */\nfunction collectNearMissGroups(\n  tokens: Token[],\n  literalCountPrefix: Int32Array,\n  blockRanges: TokenRange[],\n  minTokens: number,\n  minSimilarityPercent: number,\n  reportedGroups: CountedOccurrence[][]\n): CountedOccurrence[][] {\n  if (minSimilarityPercent >= 100) {\n    return [];\n  }\n  const eligible = blockRanges\n    .filter((range) => {\n      const tokenCount = range.endTokenIndex - range.startTokenIndex;\n      const literalCount =\n        (literalCountPrefix[range.endTokenIndex] ?? 0) - (literalCountPrefix[range.startTokenIndex] ?? 0);\n      return tokenCount >= minTokens && !isLiteralDense(literalCount, tokenCount);\n    })\n    .toSorted(\n      (left, right) => left.startTokenIndex - right.startTokenIndex || right.endTokenIndex - left.endTokenIndex\n    );\n  const comparable = selectComparableBlocks(eligible);\n  if (comparable.length < 2) {\n    return [];\n  }\n\n  // Reported-group indices whose occurrences overlap each comparable block: such blocks anchor\n  // near-miss comparisons but are never re-reported.\n  const touchedGroupsByBlock = comparable.map((range) => {\n    const touched: number[] = [];\n    for (const [groupIndex, group] of reportedGroups.entries()) {\n      if (\n        group.some(\n          (occurrence) =>\n            occurrence.startTokenIndex < range.endTokenIndex && range.startTokenIndex < occurrence.endTokenIndex\n        )\n      ) {\n        touched.push(groupIndex);\n      }\n    }\n    return touched;\n  });\n\n  // Interned per call so a file's symbol ids (and thus its n-gram hashes) never depend on which\n  // other files the process measured before it.\n  const symbolIdByTokenHashes = new Map<string, number>();\n  const sequences = comparable.map((range) => normalizeBlockSequence(tokens, range, symbolIdByTokenHashes));\n  const ngramSets = sequences.map(({ sequence }) => collectNgramSet(sequence));\n  const sharedNgramCounts = countSharedNgrams(ngramSets);\n\n  const parent = comparable.map((_, index) => index);\n  const find = (index: number): number => {\n    let root = index;\n    while (parent[root] !== root) {\n      root = parent[root] ?? root;\n    }\n    while (parent[index] !== root) {\n      const next = parent[index] ?? root;\n      parent[index] = root;\n      index = next;\n    }\n    return root;\n  };\n  for (const [pairKey, shared] of sharedNgramCounts) {\n    // Decoded with the same modulus countSharedNgrams encodes with.\n    const leftIndex = Math.floor(pairKey / ngramSets.length);\n    const rightIndex = pairKey % ngramSets.length;\n    const left = sequences[leftIndex];\n    const right = sequences[rightIndex];\n    const leftNgrams = ngramSets[leftIndex];\n    const rightNgrams = ngramSets[rightIndex];\n    if (!left || !right || !leftNgrams || !rightNgrams) {\n      continue;\n    }\n    // Two already-reported blocks have nothing new to contribute to each other.\n    if ((touchedGroupsByBlock[leftIndex]?.length ?? 0) > 0 && (touchedGroupsByBlock[rightIndex]?.length ?? 0) > 0) {\n      continue;\n    }\n    if (shared * 100 < nearMissFiltrationPercent * Math.min(leftNgrams.size, rightNgrams.size)) {\n      continue;\n    }\n    // A structural match must be backed by shared content (names and literal values), or two\n    // same-skeleton blocks calling entirely different APIs would pair on punctuation alone.\n    if (\n      contentOverlap(left, right) * 100 <=\n      minContentSimilarityPercent * Math.max(left.contentTotal, right.contentTotal)\n    ) {\n      continue;\n    }\n    // Per-fragment similarity against the larger block (NiCad semantics): both blocks must be\n    // mostly covered by the common subsequence, which is stricter than NIL's min-denominator and\n    // keeps a generic small block from \"matching\" inside every big one.\n    if (\n      lcsLength(left.sequence, right.sequence) * 100 >=\n      minSimilarityPercent * Math.max(left.sequence.length, right.sequence.length)\n    ) {\n      const leftRoot = find(leftIndex);\n      const rightRoot = find(rightIndex);\n      parent[Math.max(leftRoot, rightRoot)] = Math.min(leftRoot, rightRoot);\n    }\n  }\n\n  const membersByRoot = new Map<number, number[]>();\n  for (const index of comparable.keys()) {\n    const root = find(index);\n    const members = membersByRoot.get(root) ?? [];\n    members.push(index);\n    membersByRoot.set(root, members);\n  }\n  const groups: CountedOccurrence[][] = [];\n  for (const members of membersByRoot.values()) {\n    if (members.length < 2) {\n      continue;\n    }\n    const uncovered = members.filter((index) => (touchedGroupsByBlock[index]?.length ?? 0) === 0);\n    const covered = members.filter((index) => (touchedGroupsByBlock[index]?.length ?? 0) > 0);\n    if (covered.length === 0) {\n      groups.push(members.flatMap((index) => (comparable[index] ? [toNearMissOccurrence(comparable[index])] : [])));\n      continue;\n    }\n    if (uncovered.length === 0) {\n      continue;\n    }\n    // An anchored cluster extends a reported group only when that group lies entirely inside the\n    // cluster (every occurrence overlaps a member); appending the edited copies there keeps one\n    // group per clone family and keeps reported regions disjoint. A group that also has\n    // occurrences elsewhere is left untouched, and the uncovered copies stand alone if they can.\n    const overlapsMember = (occurrence: CountedOccurrence): boolean =>\n      members.some((index) => {\n        const range = comparable[index];\n        return (\n          range !== undefined &&\n          occurrence.startTokenIndex < range.endTokenIndex &&\n          range.startTokenIndex < occurrence.endTokenIndex\n        );\n      });\n    const fullyClustered = [...new Set(covered.flatMap((index) => touchedGroupsByBlock[index] ?? []))]\n      .filter((groupIndex) => {\n        const group = reportedGroups[groupIndex];\n        return group !== undefined && group.length > 0 && group.every(overlapsMember);\n      })\n      .toSorted((leftIndex, rightIndex) => leftIndex - rightIndex);\n    const [targetIndex, ...sourceIndexes] = fullyClustered;\n    if (targetIndex !== undefined) {\n      // Every fully-clustered group belongs to this verified component: rebuild them as ONE group\n      // with one occurrence per member block. Occurrences landing in the same member (a copy's\n      // exact prefix and suffix fragments, split by an over-large edited middle) coalesce into a\n      // single multi-segment occurrence — they are one copy, not two — and each uncovered member\n      // contributes its whole-block occurrence. Merged-away entries are left empty for the caller\n      // to drop.\n      const consumed = new Set<CountedOccurrence>();\n      const merged: CountedOccurrence[] = [];\n      for (const memberIndex of members) {\n        const range = comparable[memberIndex];\n        if (!range) {\n          continue;\n        }\n        // Occurrences of ONE group are distinct copies (a repeated run inside the block); only\n        // fragments from DIFFERENT groups (a copy's exact prefix in one group and its suffix in\n        // another) belong to the same copy. Walking the member's fragments in position order and\n        // starting a new copy whenever a source group repeats keeps same-group copies separate,\n        // and — because the fragments are disjoint and each copy takes a consecutive slice —\n        // guarantees the coalesced spans never overlap, even when groups contribute unequal\n        // fragment counts to this member.\n        const fragments: { occurrence: CountedOccurrence; groupIndex: number }[] = [];\n        for (const groupIndex of fullyClustered) {\n          for (const occurrence of reportedGroups[groupIndex] ?? []) {\n            if (\n              !consumed.has(occurrence) &&\n              occurrence.startTokenIndex < range.endTokenIndex &&\n              range.startTokenIndex < occurrence.endTokenIndex\n            ) {\n              consumed.add(occurrence);\n              fragments.push({ occurrence, groupIndex });\n            }\n          }\n        }\n        fragments.sort(\n          (left, right) =>\n            left.occurrence.startTokenIndex - right.occurrence.startTokenIndex ||\n            left.occurrence.endTokenIndex - right.occurrence.endTokenIndex\n        );\n        let copyParts: CountedOccurrence[] = [];\n        const copyGroups = new Set<number>();\n        for (const { occurrence, groupIndex } of fragments) {\n          if (copyGroups.has(groupIndex)) {\n            merged.push(coalesceOccurrences(copyParts));\n            copyParts = [];\n            copyGroups.clear();\n          }\n          copyParts.push(occurrence);\n          copyGroups.add(groupIndex);\n        }\n        if (copyParts.length > 0) {\n          merged.push(coalesceOccurrences(copyParts));\n        }\n        if (fragments.length === 0 && (touchedGroupsByBlock[memberIndex]?.length ?? 0) === 0) {\n          merged.push(toNearMissOccurrence(range));\n        }\n      }\n      merged.sort(\n        (left, right) => left.startTokenIndex - right.startTokenIndex || left.endTokenIndex - right.endTokenIndex\n      );\n      // The rebuild consumed every fully-clustered group, so shared-span marks from earlier\n      // partial merges no longer point at a separate merged group; the rebuilt group counts alone.\n      for (const occurrence of merged) {\n        occurrence.sharedWithMergedGroup = undefined;\n      }\n      reportedGroups[targetIndex] = merged;\n      for (const sourceIndex of sourceIndexes) {\n        reportedGroups[sourceIndex] = [];\n      }\n    } else if (uncovered.length >= 2) {\n      groups.push(uncovered.flatMap((index) => (comparable[index] ? [toNearMissOccurrence(comparable[index])] : [])));\n    }\n  }\n  groups.sort(compareGroups);\n  return groups;\n}\n\n/**\n * Keeps the block ranges the near-miss phase compares. Block ranges nest or are disjoint, so they\n * form a forest; a candidate whose subtree branches into two or more disjoint eligible sub-blocks\n * is a WRAPPER (a `describe(...)` statement, an IIFE, a class body) and is descended through —\n * otherwise a file whose code sits inside one enclosing construct would yield a single candidate\n * and silently disable near-miss detection. A linear chain keeps its top (the most context), and\n * the kept set is an antichain, so reported near-miss occurrences stay disjoint by construction.\n */\nfunction selectComparableBlocks(eligible: TokenRange[]): TokenRange[] {\n  const roots: ForestNode[] = [];\n  const stack: ForestNode[] = [];\n  for (const range of eligible) {\n    while (stack.length > 0 && (stack.at(-1) as ForestNode).range.endTokenIndex <= range.startTokenIndex) {\n      stack.pop();\n    }\n    const top = stack.at(-1);\n    // Equal spans (two node types covering the same tokens) collapse into the first.\n    if (top && top.range.startTokenIndex === range.startTokenIndex && top.range.endTokenIndex === range.endTokenIndex) {\n      continue;\n    }\n    const node: ForestNode = { range, children: [] };\n    if (top) {\n      top.children.push(node);\n    } else {\n      roots.push(node);\n    }\n    stack.push(node);\n  }\n\n  const kept: TokenRange[] = [];\n  const visit = (node: ForestNode): void => {\n    if (branches(node)) {\n      for (const child of node.children) {\n        visit(child);\n      }\n    } else {\n      kept.push(node.range);\n    }\n  };\n  for (const root of roots) {\n    visit(root);\n  }\n  return kept;\n}\n\ninterface ForestNode {\n  range: TokenRange;\n  children: ForestNode[];\n}\n\n/** Whether the node's subtree splits into two or more disjoint eligible sub-blocks. */\nfunction branches(node: ForestNode): boolean {\n  const [firstChild] = node.children;\n  return node.children.length >= 2 || (firstChild !== undefined && branches(firstChild));\n}\n\n/** One copy's fragments (an exact prefix and suffix split by a large edit) as one occurrence. */\nfunction coalesceOccurrences(occurrences: CountedOccurrence[]): CountedOccurrence {\n  const first = occurrences[0];\n  if (!first || occurrences.length === 1) {\n    return first ?? toEmptyOccurrence();\n  }\n  // A partial gapped merge retains the leftover group with ALL its occurrences, so a copy's\n  // fragments can arrive both standalone and embedded in a merged occurrence: union overlapping\n  // segments so no token span is reported or counted twice.\n  const segments: TokenSegment[] = [];\n  for (const segment of occurrences\n    .flatMap((occurrence) => occurrence.segments)\n    .toSorted(\n      (left, right) => left.startTokenIndex - right.startTokenIndex || left.endTokenIndex - right.endTokenIndex\n    )) {\n    const last = segments.at(-1);\n    if (last && segment.startTokenIndex < last.endTokenIndex) {\n      last.endTokenIndex = Math.max(last.endTokenIndex, segment.endTokenIndex);\n    } else {\n      segments.push({ ...segment });\n    }\n  }\n  return {\n    segments,\n    tokenCount: segments.reduce((sum, segment) => sum + segment.endTokenIndex - segment.startTokenIndex, 0),\n    startTokenIndex: Math.min(...occurrences.map((occurrence) => occurrence.startTokenIndex)),\n    endTokenIndex: Math.max(...occurrences.map((occurrence) => occurrence.endTokenIndex)),\n    startIndex: Math.min(...occurrences.map((occurrence) => occurrence.startIndex)),\n    endIndex: Math.max(...occurrences.map((occurrence) => occurrence.endIndex)),\n    startLine: Math.min(...occurrences.map((occurrence) => occurrence.startLine)),\n    endLine: Math.max(...occurrences.map((occurrence) => occurrence.endLine)),\n  };\n}\n\n/** Unreachable fallback keeping coalesceOccurrences total without a non-null assertion. */\nfunction toEmptyOccurrence(): CountedOccurrence {\n  return {\n    segments: [],\n    tokenCount: 0,\n    startTokenIndex: 0,\n    endTokenIndex: 0,\n    startIndex: 0,\n    endIndex: 0,\n    startLine: 0,\n    endLine: 0,\n  };\n}\n\n/** A near-miss occurrence spans its whole block as one segment, edited tokens included. */\nfunction toNearMissOccurrence(range: TokenRange): CountedOccurrence {\n  return {\n    segments: [{ startTokenIndex: range.startTokenIndex, endTokenIndex: range.endTokenIndex }],\n    tokenCount: range.endTokenIndex - range.startTokenIndex,\n    startTokenIndex: range.startTokenIndex,\n    endTokenIndex: range.endTokenIndex,\n    startIndex: range.startIndex,\n    endIndex: range.endIndex,\n    startLine: range.startLine,\n    endLine: range.endLine,\n  };\n}\n\ninterface NormalizedBlock {\n  sequence: Int32Array;\n  /** Occurrences per content-bearing symbol (names and literal values), for the content gate. */\n  contentCountBySymbol: Map<number, number>;\n  contentTotal: number;\n}\n\n/**\n * A block's tokens as comparable integers: anonymized identifiers become negative first-occurrence\n * indexes (per block, mirroring the exact fingerprint's rename tolerance) and every other token\n * becomes a non-negative id interned over BOTH 32-bit text hashes — with literal VALUES folded in,\n * unlike the exact fingerprint's kind tags — so one hash collision cannot equate two different\n * tokens and a fuzzy 70% match cannot mistake same-shape tables or dispatch predicates with\n * entirely different values for copies.\n */\nfunction normalizeBlockSequence(\n  tokens: Token[],\n  range: TokenRange,\n  symbolIdByTokenHashes: Map<string, number>\n): NormalizedBlock {\n  const sequence = new Int32Array(range.endTokenIndex - range.startTokenIndex);\n  const indexByIdentifier = new Map<string, number>();\n  const contentCountBySymbol = new Map<number, number>();\n  let contentTotal = 0;\n  for (let index = range.startTokenIndex; index < range.endTokenIndex; index += 1) {\n    const token = tokens[index];\n    if (!token) {\n      continue;\n    }\n    let value: number;\n    if (token.kind === 'id') {\n      let identifierIndex = indexByIdentifier.get(token.text);\n      if (identifierIndex === undefined) {\n        identifierIndex = indexByIdentifier.size;\n        indexByIdentifier.set(token.text, identifierIndex);\n      }\n      value = -(identifierIndex + 1);\n    } else {\n      const key = `${token.textHash}:${token.textHash2}:${token.literalHash ?? 0}:${token.literalHash2 ?? 0}`;\n      let id = symbolIdByTokenHashes.get(key);\n      if (id === undefined) {\n        id = symbolIdByTokenHashes.size;\n        symbolIdByTokenHashes.set(key, id);\n      }\n      value = id;\n      if (token.isName === true || token.literalHash !== undefined) {\n        contentCountBySymbol.set(value, (contentCountBySymbol.get(value) ?? 0) + 1);\n        contentTotal += 1;\n      }\n    }\n    sequence[index - range.startTokenIndex] = value;\n  }\n  return { sequence, contentCountBySymbol, contentTotal };\n}\n\n/** Multiset overlap of two blocks' content-bearing symbols, for the content gate. */\nfunction contentOverlap(left: NormalizedBlock, right: NormalizedBlock): number {\n  const [smaller, larger] =\n    left.contentCountBySymbol.size <= right.contentCountBySymbol.size ? [left, right] : [right, left];\n  let overlap = 0;\n  for (const [symbol, count] of smaller.contentCountBySymbol) {\n    overlap += Math.min(count, larger.contentCountBySymbol.get(symbol) ?? 0);\n  }\n  return overlap;\n}\n\n/** The distinct 5-gram hashes of a normalized block sequence, for the filtration set overlap. */\nfunction collectNgramSet(sequence: Int32Array): Set<number> {\n  const ngrams = new Set<number>();\n  for (let start = 0; start + nearMissNgramSize <= sequence.length; start += 1) {\n    let hash = 5381;\n    for (let offset = 0; offset < nearMissNgramSize; offset += 1) {\n      // oxlint-disable-next-line unicorn/prefer-math-trunc -- `| 0` wraps the sum to int32 to match the native backend's wrapping i32 arithmetic.\n      hash = (Math.imul(hash, 31) + (sequence[start + offset] ?? 0)) | 0;\n    }\n    ngrams.add(hash);\n  }\n  return ngrams;\n}\n\n/** Shared distinct-n-gram counts per block pair, keyed `leftIndex * blockCount + rightIndex`. */\nfunction countSharedNgrams(ngramSets: Set<number>[]): Map<number, number> {\n  const blocksByNgram = new Map<number, number[]>();\n  for (const [blockIndex, ngrams] of ngramSets.entries()) {\n    // Get-or-create: this loop runs once per distinct n-gram of every block, so the redundant\n    // `set` on already-present keys is worth avoiding here.\n    for (const ngram of ngrams) {\n      let blocks = blocksByNgram.get(ngram);\n      if (!blocks) {\n        blocks = [];\n        blocksByNgram.set(ngram, blocks);\n      }\n      blocks.push(blockIndex);\n    }\n  }\n  const sharedCounts = new Map<number, number>();\n  for (const blocks of blocksByNgram.values()) {\n    // Index-based loops: slicing here would allocate per pair in what can be a hot loop.\n    for (let leftPosition = 0; leftPosition < blocks.length; leftPosition += 1) {\n      const leftIndex = blocks[leftPosition] ?? 0;\n      for (let rightPosition = leftPosition + 1; rightPosition < blocks.length; rightPosition += 1) {\n        const pairKey = leftIndex * ngramSets.length + (blocks[rightPosition] ?? 0);\n        sharedCounts.set(pairKey, (sharedCounts.get(pairKey) ?? 0) + 1);\n      }\n    }\n  }\n  return sharedCounts;\n}\n\n/**\n * Longest-common-subsequence LENGTH of two symbol sequences via the Allison–Dix bit-parallel\n * recurrence (O(|a|/32 · |b|) words): per symbol of `b`, `x = match | v` and\n * `v = x & ~(x - ((v << 1) | 1))` over multi-word bit vectors; the set bits of `v` count the LCS.\n * Only the length is needed (similarity is a ratio), and LCS length is algorithm-independent, so\n * the native backend may use a different word size and still agree bit-for-bit.\n */\nexport function lcsLength(a: Int32Array, b: Int32Array): number {\n  const wordCount = (a.length + 31) >>> 5;\n  const positionMasks = new Map<number, Uint32Array>();\n  for (const [index, symbol] of a.entries()) {\n    let mask = positionMasks.get(symbol);\n    if (!mask) {\n      mask = new Uint32Array(wordCount);\n      positionMasks.set(symbol, mask);\n    }\n    const word = index >>> 5;\n    // The Uint32Array store wraps the signed int32 bit pattern to unsigned. The `?? 0` guards in\n    // this function are required by noUncheckedIndexedAccess (typed-array reads type as\n    // `number | undefined`), not redundancy: every index is in bounds.\n    mask[word] = (mask[word] ?? 0) | (1 << (index & 31));\n  }\n\n  const v = new Uint32Array(wordCount);\n  for (const symbol of b) {\n    const matchMask = positionMasks.get(symbol);\n    // `(v << 1) | 1` shifts a carry bit across words; subtraction borrows across words.\n    let shiftCarry = 1;\n    let borrow = 0;\n    for (let word = 0; word < wordCount; word += 1) {\n      const previous = v[word] ?? 0;\n      // oxlint-disable-next-line unicorn/prefer-math-trunc -- `>>> 0` reinterprets the signed int32 bit pattern as unsigned so the borrow subtraction below compares magnitudes; Math.trunc would keep it negative.\n      const x = ((matchMask?.[word] ?? 0) | previous) >>> 0;\n      // oxlint-disable-next-line unicorn/prefer-math-trunc -- same unsigned reinterpretation as `x`.\n      const shifted = ((previous << 1) | shiftCarry) >>> 0;\n      shiftCarry = previous >>> 31;\n      const difference = x - shifted - borrow;\n      borrow = difference < 0 ? 1 : 0;\n      // The Uint32Array store wraps the signed int32 bit pattern to unsigned.\n      v[word] = x & ~difference;\n    }\n  }\n\n  let length = 0;\n  for (const word of v) {\n    length += popCount(word);\n  }\n  return length;\n}\n\nfunction popCount(value: number): number {\n  let count = value - ((value >>> 1) & 0x55_55_55_55);\n  count = (count & 0x33_33_33_33) + ((count >>> 2) & 0x33_33_33_33);\n  return (Math.imul((count + (count >>> 4)) & 0x0F_0F_0F_0F, 0x01_01_01_01) >>> 24) & 0xFF;\n}\n\nfunction toCandidate(\n  fingerprint: string,\n  startTokenIndex: number,\n  endTokenIndex: number,\n  first: TokenRange,\n  last: TokenRange\n): DuplicateCandidate {\n  return {\n    fingerprint,\n    tokenCount: endTokenIndex - startTokenIndex,\n    startTokenIndex,\n    endTokenIndex,\n    startIndex: first.startIndex,\n    endIndex: last.endIndex,\n    startLine: first.startLine,\n    endLine: last.endLine,\n  };\n}\n\n/** Caches of hashText/hashText2 over '$0', '$1', ... so anonymized identifiers hash without allocating. */\nconst anonymizedIndexHashes: number[] = [];\nconst anonymizedIndexHashes2: number[] = [];\n\nfunction anonymizedIndexHash(index: number): number {\n  let hash = anonymizedIndexHashes[index];\n  if (hash === undefined) {\n    hash = hashText(`$${index}`);\n    anonymizedIndexHashes[index] = hash;\n  }\n  return hash;\n}\n\nfunction anonymizedIndexHash2(index: number): number {\n  let hash = anonymizedIndexHashes2[index];\n  if (hash === undefined) {\n    hash = hashText2(`$${index}`);\n    anonymizedIndexHashes2[index] = hash;\n  }\n  return hash;\n}\n\n/**\n * Content key of a token range: two independent 32-bit hashes over the normalized token sequence\n * (identifiers anonymized consistently by first-occurrence order) plus the token count. Regions\n * with equal keys are treated as equal content; a collision would need both 32-bit hashes and the\n * length to coincide, which is negligible for a metrics report. Hashing per-token instead of\n * serializing the whole range to a string keeps fingerprinting allocation-free for nested regions.\n */\nfunction fingerprintKey(\n  tokens: Token[],\n  literalCountPrefix: Int32Array,\n  startTokenIndex: number,\n  endTokenIndex: number\n): string {\n  const literalCount = (literalCountPrefix[endTokenIndex] ?? 0) - (literalCountPrefix[startTokenIndex] ?? 0);\n  const literalDense = isLiteralDense(literalCount, endTokenIndex - startTokenIndex);\n  const [primary, secondary] = fingerprintHashPair(tokens, startTokenIndex, endTokenIndex, literalDense);\n  return `${primary}:${secondary}:${endTokenIndex - startTokenIndex}`;\n}\n\n/**\n * A single 32-bit summary of a range for the coarse rolling-hash phase. Deliberately\n * density-agnostic: density is a property of the final candidate REGION, and folding literal\n * values into per-statement hashes would make a dense statement inside a logic-heavy window\n * (`const weights = [1, 2, 3];`) block the window from ever being enumerated. The coarse phase\n * over-approximates on shape alone; the exact region fingerprint still applies the density rule.\n */\nfunction fingerprintHash(tokens: Token[], startTokenIndex: number, endTokenIndex: number): number {\n  const [primary, secondary] = fingerprintHashPair(tokens, startTokenIndex, endTokenIndex, false);\n  // XOR already coerces to int32, matching the native backend's i32 arithmetic.\n  return primary ^ Math.imul(secondary, 31);\n}\n\nfunction fingerprintHashPair(\n  tokens: Token[],\n  startTokenIndex: number,\n  endTokenIndex: number,\n  foldLiteralValues: boolean\n): [number, number] {\n  const indexByIdentifier = new Map<string, number>();\n  let primary = 5381;\n  let secondary = 52_711;\n  for (let index = startTokenIndex; index < endTokenIndex; index += 1) {\n    const token = tokens[index];\n    if (!token) {\n      continue;\n    }\n    // Each accumulator consumes its own independent per-token hash: sharing one would collapse\n    // the key to 32 effective bits (a single djb2 collision would equate whole regions).\n    let part: number;\n    let part2: number;\n    if (token.kind === 'id') {\n      let identifierIndex = indexByIdentifier.get(token.text);\n      if (identifierIndex === undefined) {\n        identifierIndex = indexByIdentifier.size;\n        indexByIdentifier.set(token.text, identifierIndex);\n      }\n      part = anonymizedIndexHash(identifierIndex);\n      part2 = anonymizedIndexHash2(identifierIndex);\n    } else {\n      part = token.textHash;\n      part2 = token.textHash2;\n    }\n    // oxlint-disable-next-line unicorn/prefer-math-trunc -- `| 0` wraps the sum to int32 (Math.trunc does not), which must match the native backend's wrapping i32 arithmetic.\n    primary = (Math.imul(primary, 31) + part) | 0;\n    secondary = Math.imul(secondary, 37) ^ part2;\n    if (foldLiteralValues && token.literalHash !== undefined && token.literalHash2 !== undefined) {\n      // oxlint-disable-next-line unicorn/prefer-math-trunc -- `| 0` wraps the sum to int32 (Math.trunc does not), which must match the native backend's wrapping i32 arithmetic.\n      primary = (Math.imul(primary, 31) + token.literalHash) | 0;\n      secondary = Math.imul(secondary, 37) ^ token.literalHash2;\n    }\n  }\n  return [primary, secondary];\n}\n\n/** djb2-style hash; XOR keeps the value in signed 32-bit range, which is fine for a grouping key. */\nexport function hashText(text: string): number {\n  let hash = 5381;\n  for (let index = 0; index < text.length; index += 1) {\n    // oxlint-disable-next-line unicorn/prefer-code-point -- djb2 hashes UTF-16 code units; codePointAt would hash surrogate pairs twice (full code point, then the lone low surrogate).\n    hash = Math.imul(hash, 33) ^ text.charCodeAt(index);\n  }\n  return hash;\n}\n\n/** FNV-1a over UTF-16 code units: independent of hashText so the two accumulators never share input. */\nfunction hashText2(text: string): number {\n  let hash = -2_128_831_035; // 2166136261 as int32 (the FNV-1a offset basis)\n  for (let index = 0; index < text.length; index += 1) {\n    // oxlint-disable-next-line unicorn/prefer-code-point -- hashes UTF-16 code units like hashText.\n    hash = Math.imul(hash ^ text.charCodeAt(index), 16_777_619);\n  }\n  return hash;\n}\n\nfunction combineHashes(hash: number, value: number): number {\n  return Math.imul(hash, 31) + value;\n}\n\nfunction toCountedGroups(counted: Map<string, DuplicateCandidate[]>): CountedOccurrence[][] {\n  const groups: CountedOccurrence[][] = [];\n  for (const group of counted.values()) {\n    const occurrences = group.map((candidate) => ({\n      segments: [{ startTokenIndex: candidate.startTokenIndex, endTokenIndex: candidate.endTokenIndex }],\n      tokenCount: candidate.tokenCount,\n      startTokenIndex: candidate.startTokenIndex,\n      endTokenIndex: candidate.endTokenIndex,\n      startIndex: candidate.startIndex,\n      endIndex: candidate.endIndex,\n      startLine: candidate.startLine,\n      endLine: candidate.endLine,\n    }));\n    occurrences.sort(\n      (left, right) => left.startTokenIndex - right.startTokenIndex || left.endTokenIndex - right.endTokenIndex\n    );\n    groups.push(occurrences);\n  }\n  return groups;\n}\n\n/**\n * Merges duplicate groups separated by a small token gap into one gapped (Type-3) clone group: a\n * copy edited in one spot splits into two exact groups whose occurrences sit side by side in the\n * same order. Occurrences are paired greedily in source order; a merge happens when the pairing\n * fully consumes at least one group with at least two pairs. Equal-cardinality groups whose\n * occurrences all pair merge into one group as before. When cardinalities differ (a fragment also\n * occurs standalone: prefix ×3, suffix ×2), the fully-paired group is subsumed into the merged\n * gapped group while the other group is RETAINED with ALL its occurrences: dropping the leftover\n * would lose duplicated-line coverage, and reporting it alone would make a single-occurrence group\n * (contradicting duplicateBlockGroupCount's \"appears more than once\" meaning). Line coverage\n * unions ranges, so the overlap between the retained exact group and the merged group is harmless.\n * Merging repeats to a fixpoint so a clone edited in several spots still reassembles; it\n * terminates because a full merge shrinks the group count and a partial merge keeps it while\n * strictly growing the bounded total span of merged occurrences. Gap tokens are not matched\n * content: line coverage and sizes count only the matched segments. Generic so cross-file merging\n * can thread file identity through occurrences.\n */\nexport function mergeAdjacentGroups<T extends CountedOccurrence>(groups: T[][], maxGapTokens: number): T[][] {\n  if (maxGapTokens <= 0 || groups.length < 2) {\n    return groups;\n  }\n  // Deterministic processing order (mirrored by the native backend): by first occurrence position.\n  groups.sort(compareGroups);\n  for (let restart = true; restart;) {\n    restart = false;\n    for (let leftIndex = 0; leftIndex < groups.length && !restart; leftIndex += 1) {\n      for (let rightIndex = leftIndex + 1; rightIndex < groups.length; rightIndex += 1) {\n        const left = groups[leftIndex];\n        const right = groups[rightIndex];\n        if (!left || !right) {\n          continue;\n        }\n        const forward = mergeGroups(left, right, maxGapTokens);\n        const result = forward ?? mergeGroups(right, left, maxGapTokens);\n        if (!result) {\n          continue;\n        }\n        const leftConsumed = forward ? result.firstConsumed : result.secondConsumed;\n        const rightConsumed = forward ? result.secondConsumed : result.firstConsumed;\n        if (leftConsumed && rightConsumed) {\n          groups[leftIndex] = result.merged;\n          groups.splice(rightIndex, 1);\n        } else if (rightConsumed) {\n          groups[rightIndex] = result.merged;\n        } else {\n          groups[leftIndex] = result.merged;\n        }\n        // A partial merge retains the not-fully-consumed group with ALL its occurrences (line\n        // coverage must not shrink, and a reported group must keep >= 2 occurrences), so its\n        // paired occurrences now also live inside the merged group's occurrences: mark them so\n        // duplicateBlockCount counts each token span once.\n        for (const occurrence of result.pairedRetained) {\n          occurrence.sharedWithMergedGroup = true;\n        }\n        groups.sort(compareGroups);\n        restart = true;\n        break;\n      }\n    }\n  }\n  return groups;\n}\n\nfunction compareGroups(left: CountedOccurrence[], right: CountedOccurrence[]): number {\n  const leftFirst = left[0];\n  const rightFirst = right[0];\n  return (\n    (leftFirst?.startTokenIndex ?? 0) - (rightFirst?.startTokenIndex ?? 0) ||\n    (leftFirst?.endTokenIndex ?? 0) - (rightFirst?.endTokenIndex ?? 0)\n  );\n}\n\ninterface MergeResult<T> {\n  merged: T[];\n  /** Whether every occurrence of the respective input group was paired into the merge. */\n  firstConsumed: boolean;\n  secondConsumed: boolean;\n  /** The retained (not fully consumed) group's occurrences that were paired into the merge. */\n  pairedRetained: T[];\n}\n\n/**\n * Pairs `second` occurrences with gap-preceding `first` occurrences, greedily in source order:\n * each trailing occurrence takes the earliest unused leading occurrence within the gap, and a\n * pair's leading must start at or after the previous pair's trailing end so merged spans never\n * overlap. A merge needs at least two pairs (a merged group must still mean \"appears more than\n * once\") and must fully consume at least one group; for equal cardinalities this reduces to the\n * strict all-pairs merge, so pre-partial-merge behavior is unchanged there.\n */\nfunction mergeGroups<T extends CountedOccurrence>(\n  first: T[],\n  second: T[],\n  maxGapTokens: number\n): MergeResult<T> | undefined {\n  // Occurrences a previous partial merge already paired into a merged group must not pair again:\n  // their spans already live inside that merged group, so re-pairing them would assemble a second,\n  // competing merged group instead of letting the existing merged group extend (and would count\n  // the same span twice). Consumption is still judged against the FULL group, so a group holding\n  // shared occurrences is never subsumed away.\n  const leadings = first.filter((occurrence) => !occurrence.sharedWithMergedGroup);\n  const trailings = second.filter((occurrence) => !occurrence.sharedWithMergedGroup);\n  const pairs: [T, T][] = [];\n  let leadingIndex = 0;\n  let previousTrailingEnd = -1;\n  for (const trailing of trailings) {\n    // Leadings ending too far before this trailing can never pair a later (even farther) one.\n    while (leadingIndex < leadings.length) {\n      const leading = leadings[leadingIndex];\n      if (leading && leading.endTokenIndex + maxGapTokens < trailing.startTokenIndex) {\n        leadingIndex += 1;\n      } else {\n        break;\n      }\n    }\n    const leading = leadings[leadingIndex];\n    if (\n      leading &&\n      leading.endTokenIndex <= trailing.startTokenIndex &&\n      leading.startTokenIndex >= previousTrailingEnd\n    ) {\n      pairs.push([leading, trailing]);\n      previousTrailingEnd = trailing.endTokenIndex;\n      leadingIndex += 1;\n    }\n  }\n  const firstConsumed = pairs.length === first.length;\n  const secondConsumed = pairs.length === second.length;\n  if (pairs.length < 2 || (!firstConsumed && !secondConsumed)) {\n    return undefined;\n  }\n  const merged = pairs.map(([leading, trailing]) => ({\n    ...leading,\n    // A merged occurrence is a fresh span combination; it never inherits shared-span marks.\n    sharedWithMergedGroup: undefined,\n    segments: [...leading.segments, ...trailing.segments],\n    tokenCount: leading.tokenCount + trailing.tokenCount,\n    endTokenIndex: trailing.endTokenIndex,\n    endIndex: trailing.endIndex,\n    endLine: trailing.endLine,\n  }));\n  const pairedRetained =\n    firstConsumed === secondConsumed ? [] : pairs.map(([leading, trailing]) => (firstConsumed ? trailing : leading));\n  return { merged, firstConsumed, secondConsumed, pairedRetained };\n}\n\n/**\n * Redundant copies one group adds to duplicateBlockCount. Each redundant occurrence contributes\n * one count per matched fragment, so merging a gapped clone's fragments into one group does not\n * halve duplicateBlockCount: an edited two-fragment pair still counts 2, exactly as its unmerged\n * fragments did. Occurrence shapes can differ within one group (a\n * gap-merged exact pair plus an appended whole-block near-miss copy), so every occurrence's\n * fragments are summed and one representative — the largest — is deducted, keeping the count\n * independent of source order. Occurrences a partial gapped merge also paired into a merged group\n * are skipped: their spans are already counted there, and such a retained group deducts no\n * representative of its own — the merged group's representative already stands for the shared\n * content — so no token span contributes to the count twice.\n */\nexport function countRedundantFragments(group: CountedOccurrence[]): number {\n  let fragmentCount = 0;\n  let maxFragmentCount = 0;\n  let hasSharedOccurrence = false;\n  for (const occurrence of group) {\n    if (occurrence.sharedWithMergedGroup) {\n      hasSharedOccurrence = true;\n      continue;\n    }\n    fragmentCount += occurrence.segments.length;\n    maxFragmentCount = Math.max(maxFragmentCount, occurrence.segments.length);\n  }\n  return hasSharedOccurrence ? fragmentCount : fragmentCount - maxFragmentCount;\n}\n\nfunction summarizeDuplicates(\n  groups: CountedOccurrence[][],\n  codeLineNumbers: Set<number>,\n  tokens: Token[]\n): DuplicationMetrics {\n  let duplicateBlockCount = 0;\n  let maxDuplicateBlockSize = 0;\n  const duplicateBlockGroups: { startLine: number; endLine: number }[][] = [];\n  const duplicatedLines = new Set<number>();\n  for (const group of groups) {\n    duplicateBlockCount += countRedundantFragments(group);\n    for (const occurrence of group) {\n      maxDuplicateBlockSize = Math.max(maxDuplicateBlockSize, occurrence.tokenCount);\n      // Only CODE lines carrying segment tokens count: comments and blank gaps inside an\n      // occurrence's bounding range — the unmatched gap of a merged clone, and blank rows inside a\n      // multi-row token (heredocs, template literals) — are not duplicated content and would push\n      // the ratio past 1. A near-miss occurrence's single segment deliberately spans its WHOLE\n      // block, edited tokens included: an LCS has no canonical per-line attribution, and\n      // NiCad-style detectors treat the whole near-miss fragment as the clone.\n      for (const segment of occurrence.segments) {\n        collectSegmentLines(segment, tokens, codeLineNumbers, duplicatedLines);\n      }\n    }\n    duplicateBlockGroups.push(\n      group\n        .map(({ startLine, endLine }) => ({ startLine, endLine }))\n        .toSorted((left, right) => left.startLine - right.startLine)\n    );\n  }\n  duplicateBlockGroups.sort((left, right) => (left[0]?.startLine ?? 0) - (right[0]?.startLine ?? 0));\n\n  return {\n    duplicateBlockCount,\n    duplicateBlockGroupCount: groups.length,\n    duplicateBlockGroups,\n    duplicateLineCount: duplicatedLines.size,\n    duplicateLineNumbers: [...duplicatedLines].toSorted((left, right) => left - right),\n    duplicationRatio: codeLineNumbers.size === 0 ? 0 : duplicatedLines.size / codeLineNumbers.size,\n    maxDuplicateBlockSize,\n  };\n}\n\n/** Adds the 1-based code lines the segment's matched tokens cover; shared with cross-file coverage. */\nexport function collectSegmentLines(\n  segment: { startTokenIndex: number; endTokenIndex: number },\n  tokens: Token[],\n  codeLineNumbers: Set<number> | undefined,\n  duplicatedLines: Set<number>\n): void {\n  for (let index = segment.startTokenIndex; index < segment.endTokenIndex; index += 1) {\n    const token = tokens[index];\n    for (let row = token?.startRow ?? 0; row <= (token?.endRow ?? -1); row += 1) {\n      if (!codeLineNumbers || codeLineNumbers.has(row + 1)) {\n        duplicatedLines.add(row + 1);\n      }\n    }\n  }\n}\n"],"mappings":"kFASA,MAAM,EAAsB,IAAI,IAAI,irBAmDpC,CAAC,EAGK,EAA0B,IAAI,IAAI,CACtC,UACA,cACA,mBACA,SACA,kBACA,QACA,qBACA,iBACA,mBACA,aACA,aACA,WAEA,KACA,SACA,OACA,OAEA,iBACA,+BACA,cACA,kBACA,YACA,qBACA,cACF,CAAC,EAOK,EAA4B,IAAI,IAAI,CACxC,aACA,WACA,oBACA,iBACA,iBACF,CAAC,EAOK,EAAyB,IAAI,IAAI,CAAC,gCAAiC,uCAAuC,CAAC,EAG3G,EAAoB,IAAI,IAAI,CAChC,CAAC,SAAU,MAAM,EACjB,CAAC,iBAAkB,MAAM,EACzB,CAAC,UAAW,MAAM,EAClB,CAAC,QAAS,MAAM,EAChB,CAAC,kBAAmB,MAAM,EAC1B,CAAC,gBAAiB,MAAM,EACxB,CAAC,cAAe,MAAM,EACtB,CAAC,eAAgB,OAAO,EACxB,CAAC,oBAAqB,MAAM,EAC5B,CAAC,0BAA2B,MAAM,EAClC,CAAC,sBAAuB,MAAM,EAC9B,CAAC,wBAAyB,MAAM,EAChC,CAAC,yBAA0B,MAAM,EACjC,CAAC,iCAAkC,MAAM,EACzC,CAAC,6BAA8B,MAAM,EACrC,CAAC,kBAAmB,MAAM,EAC1B,CAAC,4BAA6B,MAAM,EACpC,CAAC,iBAAkB,MAAM,EACzB,CAAC,qBAAsB,MAAM,EAC7B,CAAC,kBAAmB,MAAM,EAE1B,CAAC,oBAAqB,UAAU,EAChC,CAAC,cAAe,UAAU,EAE1B,CAAC,SAAU,MAAM,EACjB,CAAC,kBAAmB,MAAM,EAC1B,CAAC,iBAAkB,MAAM,EACzB,CAAC,6BAA8B,MAAM,EACrC,CAAC,qBAAsB,MAAM,EAC7B,CAAC,aAAc,MAAM,EACrB,CAAC,kBAAmB,MAAM,EAC1B,CAAC,eAAgB,OAAO,EACxB,CAAC,oBAAqB,OAAO,EAC7B,CAAC,YAAa,OAAO,EACrB,CAAC,gBAAiB,QAAQ,CAC5B,CAAC,EAMK,EAA4B,IAAI,IAAI,CAAC,OAAQ,OAAQ,QAAS,QAAQ,CAAC,EAEvE,EAAe,IAAI,IAAI,CAAC,UAAW,eAAgB,eAAe,CAAC,EAMnE,EAA6B,IAAI,IAAI,CACzC,kBACA,4BACA,iBACA,qBACA,kBACA,iBACF,CAAC,EAGK,EAAsB,IAAI,IAAI,CAClC,kBACA,4BACA,iBACA,qBACA,kBACA,kBAEA,eACA,YACF,CAAC,EAQK,EAAgC,IAAI,IAAI,CAC5C,CAAC,kBAAmB,UAAU,EAC9B,CAAC,oBAAqB,MAAM,EAC5B,CAAC,OAAQ,QAAQ,EACjB,CAAC,YAAa,WAAW,EACzB,CAAC,mBAAoB,OAAO,EAE5B,CAAC,eAAgB,OAAO,EAExB,CAAC,iBAAkB,aAAa,EAEhC,CAAC,mBAAoB,MAAM,EAC3B,CAAC,qBAAsB,KAAK,EAG5B,CAAC,mBAAoB,UAAU,EAC/B,CAAC,oBAAqB,MAAM,CAC9B,CAAC,EAEY,EAA0D,CACrE,UAAW,GACX,aAAc,GACd,qBAAsB,EACxB,EA2CA,SAAS,EAAe,EAAsB,EAA6B,CACzE,OAAO,EAAe,GAAK,CAC7B,CAoHA,SAAgB,EACd,EACA,EACA,EACoB,CACpB,IAAM,EAAY,GAAS,WAAa,EAA0B,UAC5D,EAAe,GAAS,cAAgB,EAA0B,aAClE,EAAuB,GAAS,sBAAwB,EAA0B,qBAClF,EAAkB,CAAC,EACnB,EAA4B,CAAC,EAC7B,EAA2C,CAAC,EAClD,EAAc,EAAM,EAAQ,EAAa,CAAwB,EACjE,IAAM,EAAqB,EAAwB,CAAM,EAEnD,EAAa,CACjB,GAAG,EAAuB,EAAQ,EAAoB,EAAa,CAAS,EAC5E,GAAG,EAA0B,EAAQ,EAAoB,EAA0B,CAAS,CAC9F,EAEM,EAAS,EAAoB,EADnB,EAAoB,EAAa,GAAU,EAAM,QAAU,CAClB,CAAC,EAAG,CAAY,EACnE,EAAiB,EACrB,EACA,EACA,EACA,EACA,EACA,CACF,EAGA,OAAO,GAAoB,CADH,GAAG,EAAO,OAAQ,GAAU,EAAM,OAAS,CAAC,EAAG,GAAG,CAClC,EAAG,EAAiB,CAAM,CACpE,CAUA,SAAgB,EACd,EACA,EAC8B,CAC9B,IAAM,EAAY,GAAS,WAAa,EAA0B,UAC5D,EAAkB,CAAC,EACnB,EAA4B,CAAC,EAC7B,EAA2C,CAAC,EAClD,EAAc,EAAM,EAAQ,EAAa,CAAwB,EACjE,IAAM,EAAqB,EAAwB,CAAM,EAEnD,EAAa,EAAuB,EAAQ,EAAoB,EAAa,CAAS,EAC5F,IAAK,IAAM,KAAc,EAA0B,CACjD,IAAM,EAAQ,EAAW,GACnB,EAAO,EAAW,GAAG,EAAE,EACzB,CAAC,GAAS,CAAC,GAGI,EAAK,cAAgB,EAAM,gBAC7B,GAGjB,EAAW,KACT,EACE,KAAK,EAAe,EAAQ,EAAoB,EAAM,gBAAiB,EAAK,aAAa,IACzF,EAAM,gBACN,EAAK,cACL,EACA,CACF,CACF,CACF,CACA,MAAO,CAAE,WAAY,EAAe,CAAU,EAAG,SAAQ,oBAAqB,CAAyB,CACzG,CAEA,SAAS,EACP,EACA,EACA,EACA,EACM,CACN,SAAS,EAAM,EAAqC,CAClD,IAAM,EAAkB,EAAO,OACzB,EAAa,EAAK,aAAe,EAAI,IAAA,GAAY,EAAkB,CAAI,EAC7E,GAAI,EAAK,aAAe,EACtB,EAAgB,EAAM,CAAM,OACvB,GAAI,IAAe,IAAA,GAGxB,EAAO,KACL,EAAc,EAAY,EAAiB,EAAM,CAAU,EAAG,EAAK,cAAc,IAAK,EAAK,YAAY,GAAG,CAC5G,OACK,GAAI,CAAC,EAAa,IAAI,EAAK,IAAI,EAAG,CACvC,IAAM,EAAgC,CAAC,EACjC,EAAc,EAAK,SAAW,EAAwB,IAAI,EAAK,IAAI,EACzE,IAAK,IAAM,KAAS,EAAK,SAAU,CACjC,IAAM,EAAa,EAAM,CAAK,EAC1B,GAAe,EAAM,SAAW,CAAC,EAAa,IAAI,EAAM,IAAI,GAC9D,EAAgB,KAAK,CAAU,CAEnC,CAII,GAAe,EAAgB,OAAS,GAC1C,EAAyB,KAAK,CAAe,CAEjD,CAEA,IAAM,EAAQ,CACZ,kBACA,cAAe,EAAO,OACtB,WAAY,EAAK,WACjB,SAAU,EAAK,SACf,UAAW,EAAK,cAAc,IAAM,EACpC,QAAS,EAAK,YAAY,IAAM,CAClC,EAIA,OAHI,EAAK,SAAW,EAAoB,IAAI,EAAK,IAAI,GACnD,EAAY,KAAK,CAAK,EAEjB,CACT,CAEA,EAAM,CAAI,CACZ,CAGA,SAAS,EAAkB,EAA6C,CACtE,IAAM,EAAO,EAAK,QAAU,EAAkB,IAAI,EAAK,IAAI,EAAI,IAAA,GAC3D,OAAS,IAAA,GAGb,OAAO,EAAK,cAAc,MAAO,GAAU,EAAoB,IAAI,EAAM,IAAI,CAAC,EAAI,EAAO,IAAA,EAC3F,CAEA,SAAS,EAAgB,EAAyB,EAAuB,CACvE,GAAI,EAAa,IAAI,EAAK,IAAI,EAC5B,OAGF,IAAM,EAAW,EAAK,cAAc,IAC9B,EAAS,EAAK,YAAY,IAChC,GAAI,EAAK,SAAW,EAAuB,IAAI,EAAK,IAAI,EAAG,CACzD,EAAO,KACL,EAAc,EAAK,KAAM,IAAA,GAAW,EAAU,EAAQ,EAAI,EAC1D,EAAc,IAAK,IAAA,GAAW,EAAU,CAAM,EAC9C,CAAE,KAAM,KAAM,KAAM,EAAK,KAAM,SAAU,EAAG,UAAW,EAAG,WAAU,QAAO,CAC7E,EACA,MACF,CAEA,GAAI,EAAK,SAAW,EAA0B,IAAI,EAAK,IAAI,GAAK,CAAC,EAAmB,CAAI,EAAG,CACzF,EAAO,KAAK,CAAE,KAAM,KAAM,KAAM,EAAK,KAAM,SAAU,EAAG,UAAW,EAAG,WAAU,QAAO,CAAC,EACxF,MACF,CAIA,IAAM,EAAc,EAAK,QAAU,EAAkB,IAAI,EAAK,IAAI,EAAI,IAAA,GAClE,IAAgB,IAAA,GAClB,EAAO,KAAK,EAAc,EAAK,KAAM,IAAA,GAAW,EAAU,EAAQ,EAAK,OAAO,CAAC,EAE/E,EAAO,KAAK,EAAc,EAAa,EAAiB,EAAM,CAAW,EAAG,EAAU,CAAM,CAAC,CAEjG,CAEA,SAAS,EACP,EACA,EACA,EACA,EACA,EAAS,GACF,CACP,IAAM,EAAe,CAAE,KAAM,OAAQ,OAAM,SAAU,EAAS,CAAI,EAAG,UAAW,EAAU,CAAI,EAAG,WAAU,QAAO,EAQlH,OAPI,IAAqB,IAAA,IAAa,EAA0B,IAAI,CAAI,IACtE,EAAM,YAAc,EAAS,CAAgB,EAC7C,EAAM,aAAe,EAAU,CAAgB,GAE7C,IACF,EAAM,OAAS,IAEV,CACT,CAWA,SAAS,EAAiB,EAAyB,EAAsB,CAMvE,GALI,IAAS,QAAU,IAAS,SAK5B,EAA2B,IAAI,EAAK,IAAI,EAC1C,OAAO,EAAK,KAEd,IAAM,EAAY,EAAK,cAAc,OAAQ,GAAU,EAA2B,IAAI,EAAM,IAAI,CAAC,EAIjG,OAHI,EAAU,OAAS,EACd,EAAU,IAAK,GAAU,EAAM,IAAI,CAAC,CAAC,KAAK,EAAE,EAE9C,EAAoB,EAAK,IAAI,CACtC,CAEA,MAAM,EAAkB,IAAI,IAAI,CAAC,IAAK,IAAK,GAAG,CAAC,EAE/C,SAAS,EAAoB,EAAsB,CACjD,IAAM,EAAQ,EAAK,GACnB,OAAO,EAAK,QAAU,GAAK,IAAU,IAAA,IAAa,EAAgB,IAAI,CAAK,GAAK,EAAK,SAAS,CAAK,EAC/F,EAAK,MAAM,EAAG,EAAE,EAChB,CACN,CAGA,SAAgB,EAAwB,EAA6B,CACnE,IAAM,EAAS,IAAI,WAAW,EAAO,OAAS,CAAC,EAC/C,IAAK,GAAM,CAAC,EAAO,KAAU,EAAO,QAAQ,EAC1C,EAAO,EAAQ,IAAM,EAAO,IAAU,IAAM,EAAM,cAAgB,IAAA,GAAY,EAAI,GAEpF,OAAO,CACT,CAEA,SAAS,EAAmB,EAAkC,CAC5D,IAAM,EAAS,EAAK,OACpB,GAAI,CAAC,EACH,MAAO,GAuBT,GAlBI,EAAO,OAAS,oBAKhB,EAAO,OAAS,QAAU,EAAO,kBAAkB,UAAU,CAAC,EAAE,KAAO,EAAK,IAM5E,EAAK,OAAS,YAAc,EAAO,OAAS,QAAU,EAAO,kBAAkB,UAAU,CAAC,EAAE,KAAO,EAAK,IAQ1G,EAAO,OAAS,qBAChB,EAAO,kBAAkB,QAAQ,CAAC,EAAE,KAAO,EAAK,IAChD,WAAW,KAAK,EAAK,IAAI,EAEzB,MAAO,GAOT,IACG,EAAO,OAAS,qBAAuB,EAAO,OAAS,0BACvD,EAAO,kBAAkB,MAAM,CAAC,EAAE,KAAO,EAAK,IAAM,EAAO,kBAAkB,MAAM,CAAC,EAAE,KAAO,EAAK,IACnG,CACA,IAAI,EAAQ,EACZ,KACE,EAAM,SACL,EAAM,OAAO,OAAS,qBACrB,EAAM,OAAO,OAAS,wBACtB,EAAM,OAAO,OAAS,oBACtB,EAAM,OAAO,OAAS,sBAExB,EAAQ,EAAM,OAEhB,GAAI,EAAM,QAAQ,OAAS,mBAAqB,EAAM,OAAO,kBAAkB,UAAU,CAAC,EAAE,KAAO,EAAM,GACvG,MAAO,EAEX,CAIA,GACE,EAAO,OAAS,mBAChB,EAAO,QAAQ,OAAS,iBACxB,EAAO,OAAO,WAAW,CAAC,CAAC,EAAE,KAAO,EAAO,GAE3C,MAAO,GAGT,IAAM,EAAQ,EAA8B,IAAI,EAAO,IAAI,EAC3D,OAAO,IAAU,IAAA,IAAa,EAAO,kBAAkB,CAAK,CAAC,EAAE,KAAO,EAAK,EAC7E,CAEA,SAAS,EACP,EACA,EACA,EACA,EACsB,CACtB,IAAM,EAAmC,CAAC,EAC1C,IAAK,IAAM,KAAS,EACC,EAAM,cAAgB,EAAM,gBAC9B,GAGjB,EAAW,KACT,EACE,KAAK,EAAe,EAAQ,EAAoB,EAAM,gBAAiB,EAAM,aAAa,IAC1F,EAAM,gBACN,EAAM,cACN,EACA,CACF,CACF,EAEF,OAAO,CACT,CAkBA,SAAS,EACP,EACA,EACA,EACA,EACsB,CACtB,OAAO,EAAgC,CAAC,CAAE,SAAQ,qBAAoB,YAAW,CAAC,EAAG,EAAW,EAAK,CAAC,CAAC,KACpG,CAAE,eAAgB,CACrB,CACF,CA0BA,SAAgB,EACd,EACA,EACA,EAC+B,CAC/B,IAAM,EAA4C,CAAC,EAC7C,EAAoC,CAAC,EACrC,EAA6B,CAAC,EACpC,IAAK,GAAM,CAAC,EAAc,KAAY,EAAS,QAAQ,EACrD,IAAK,IAAM,KAAc,EAAQ,WAC/B,EAAwB,KAAK,CAAY,EACzC,EAAW,KAAK,CAAU,EAG9B,IAAM,EAAa,GACjB,EAAS,EAAwB,IAAmB,GAChD,EAAyB,IAAI,IAC7B,EAAmB,EAAW,KAAK,EAAY,IACnD,EAA0B,EAAU,CAAc,CAAC,EAAE,QAAU,CAAC,EAAG,EAAY,CAAS,CAC1F,EACA,IAAK,GAAM,CAAC,EAAgB,KAAY,EAAiB,QAAQ,EAAG,CAClE,IAAM,EAAe,EAAwB,IAAmB,EAChE,IAAK,GAAM,CAAC,EAAO,KAAQ,EAAQ,kBAAkB,QAAQ,EAC3D,IAAK,IAAM,KAAa,EAAK,CAC3B,GAAI,IAAc,IAAA,GAChB,SAEF,IAAM,EAAc,EAAuB,IAAI,CAAS,EACpD,GACF,EAAY,OAAS,EACjB,EAAY,iBAAmB,IACjC,EAAY,eAAiB,IAE3B,EAAY,eAAiB,IAC/B,EAAY,aAAe,IAE7B,EAAY,SAAW,KAAK,IAAI,EAAY,SAAU,CAAK,EAC3D,EAAY,SAAW,KAAK,IAAI,EAAY,SAAU,CAAK,GAE3D,EAAuB,IAAI,EAAW,CACpC,MAAO,EACP,iBACA,eACA,SAAU,EACV,SAAU,CACZ,CAAC,CAEL,CAEJ,CAMA,IAAM,GAAW,EAA+B,IAA4B,CAC1E,GAAI,IAAc,IAAA,GAChB,MAAO,GAET,IAAM,EAAc,EAAuB,IAAI,CAAS,EAOxD,OANI,IAAgB,IAAA,IAAa,EAAY,MAAQ,EAC5C,GAEL,EACK,EAAY,eAAiB,GAE/B,EAAY,iBAAmB,IAAM,EAAY,SAAW,EAAY,UAAY,CAC7F,EAKM,EAAyB,GAAoC,CACjE,IAAM,EAAS,EAAiB,EAAO,eAAe,EAAE,iBAAmB,CAAC,EACtE,EAAY,EAAO,EAAO,OAChC,IAAK,IAAI,EAAQ,EAAO,MAAQ,EAAG,EAAQ,EAAO,MAAQ,EAAO,OAAQ,GAAS,EAChF,GAAI,EAAO,KAAW,EACpB,MAAO,GAGX,MAAO,EACT,EAEM,EAAmC,CAAC,EAC1C,IAAK,GAAM,CAAC,EAAgB,KAAY,EAAiB,QAAQ,EAC/D,IAAK,GAAM,CAAC,EAAO,KAAQ,EAAQ,kBAAkB,QAAQ,EAC3D,IAAK,GAAM,CAAC,EAAQ,KAAc,EAAI,QAAQ,EAAG,CAC/C,GAAI,CAAC,EAAQ,EAAW,CAAM,GAAK,CAAC,EAAsB,CAAE,iBAAgB,QAAO,QAAO,CAAC,EACzF,SAIF,IAAM,EAAgB,EAAQ,kBAAkB,EAAM,GAAG,EAAS,GAC5D,EAAe,EAAQ,kBAAkB,EAAQ,EAAE,GAAG,EAAS,GACjE,EAAQ,EAAe,EAAS,CAAC,GAAK,EAAQ,EAAc,EAAS,CAAC,GAG1E,EAAe,KAAK,CAAE,iBAAgB,QAAO,QAAO,CAAC,CACvD,CASJ,IAAM,EAAU,IAAI,IAAI,EAAe,IAAI,CAAQ,CAAC,EAChD,EAAW,EACf,KAAO,EAAS,OAAS,GAAG,CAC1B,IAAM,EAA4B,CAAC,EACnC,IAAK,IAAM,KAAU,EAAU,CAC7B,IAAM,EAAa,EAAW,EAAO,gBAC/B,EAAQ,IAAa,EAAO,OAC5B,EAAO,IAAa,EAAO,MAAQ,EAAO,OAAS,GACnD,EAAU,EAAU,EAAO,cAAc,EAC/C,GAAI,CAAC,GAAS,CAAC,GAAQ,CAAC,EACtB,SAEF,IAAM,EAAc,KAAK,EAAe,EAAQ,OAAQ,EAAQ,mBAAoB,EAAM,gBAAiB,EAAK,aAAa,IAC7H,EAAW,KAAK,CACd,UAAW,EAAY,EAAa,EAAM,gBAAiB,EAAK,cAAe,EAAO,CAAI,EAC1F,aAAc,EAAwB,EAAO,iBAAmB,CAClE,CAAC,EACD,EAAQ,KAAK,CAAM,CACrB,CACA,EAAW,CAAC,EACZ,IAAK,IAAM,KAAU,EACnB,IAAK,IAAM,IAAS,CAAC,EAAO,MAAO,EAAO,MAAQ,CAAC,EAAG,CACpD,IAAM,EAAY,CAAE,eAAgB,EAAO,eAAgB,QAAO,OAAQ,EAAO,OAAS,CAAE,EACtF,EAAe,EAAiB,EAAO,eAAe,EAAE,kBAAkB,EAAM,GAAG,EAAU,QAEjG,EAAQ,IAAI,EAAS,CAAS,CAAC,GAC/B,CAAC,EAAQ,EAAc,EAAU,MAAM,GACvC,CAAC,EAAsB,CAAS,IAIlC,EAAQ,IAAI,EAAS,CAAS,CAAC,EAC/B,EAAS,KAAK,CAAS,EACzB,CAEJ,CACA,OAAO,CACT,CAEA,SAAS,EAAS,EAAgC,CAChD,MAAO,GAAG,EAAO,eAAe,GAAG,EAAO,MAAM,GAAG,EAAO,QAC5D,CASA,SAAS,EAA0B,EAAiB,EAA0B,EAAqC,CACjH,IAAM,EAAkB,EAAW,IAAK,GACtC,EAAgB,EAAQ,EAAU,gBAAiB,EAAU,aAAa,CAC5E,EACM,EAA8C,CAAC,EACrD,IAAK,IAAI,EAAQ,EAAG,EAAQ,EAAW,OAAQ,GAAS,EAAG,CACzD,IAAM,EAA8B,CAAC,EACjC,EAAO,KACP,EAAa,EACX,EAAS,KAAK,IAAI,EAAW,OAAQ,EAAQ,GAAyB,EAC5E,IAAK,IAAI,EAAM,EAAO,EAAM,EAAQ,GAAO,EAAG,CAC5C,IAAM,EAAY,EAAW,GACvB,EAAgB,EAAgB,GACtC,GAAI,CAAC,GAAa,IAAkB,IAAA,GAClC,MAEF,EAAO,EAAc,EAAM,CAAa,EACxC,GAAc,EAAU,cAAgB,EAAU,gBAClD,IAAM,EAAiB,EAAM,EAAQ,EACrC,EAAI,GACF,GAAkB,GAA6B,GAAc,EACzD,EAAc,EAAM,CAAc,EAClC,IAAA,EACR,CACA,EAAkB,KAAK,CAAG,CAC5B,CACA,MAAO,CAAE,oBAAmB,iBAAgB,CAC9C,CAqBA,SAAS,EACP,EACA,EACA,EACA,EACA,EACA,EACuB,CACvB,GAAI,GAAwB,IAC1B,MAAO,CAAC,EAYV,IAAM,EAAa,GAVF,EACd,OAAQ,GAAU,CACjB,IAAM,EAAa,EAAM,cAAgB,EAAM,gBACzC,GACH,EAAmB,EAAM,gBAAkB,IAAM,EAAmB,EAAM,kBAAoB,GACjG,OAAO,GAAc,GAAa,CAAC,EAAe,EAAc,CAAU,CAC5E,CAAC,CAAC,CACD,UACE,EAAM,IAAU,EAAK,gBAAkB,EAAM,iBAAmB,EAAM,cAAgB,EAAK,aAE/C,CAAC,EAClD,GAAI,EAAW,OAAS,EACtB,MAAO,CAAC,EAKV,IAAM,EAAuB,EAAW,IAAK,GAAU,CACrD,IAAM,EAAoB,CAAC,EAC3B,IAAK,GAAM,CAAC,EAAY,KAAU,EAAe,QAAQ,EAErD,EAAM,KACH,GACC,EAAW,gBAAkB,EAAM,eAAiB,EAAM,gBAAkB,EAAW,aAC3F,GAEA,EAAQ,KAAK,CAAU,EAG3B,OAAO,CACT,CAAC,EAIK,EAAwB,IAAI,IAC5B,EAAY,EAAW,IAAK,GAAU,GAAuB,EAAQ,EAAO,CAAqB,CAAC,EAClG,EAAY,EAAU,KAAK,CAAE,cAAe,GAAgB,CAAQ,CAAC,EACrE,EAAoB,GAAkB,CAAS,EAE/C,EAAS,EAAW,KAAK,EAAG,IAAU,CAAK,EAC3C,EAAQ,GAA0B,CACtC,IAAI,EAAO,EACX,KAAO,EAAO,KAAU,GACtB,EAAO,EAAO,IAAS,EAEzB,KAAO,EAAO,KAAW,GAAM,CAC7B,IAAM,EAAO,EAAO,IAAU,EAC9B,EAAO,GAAS,EAChB,EAAQ,CACV,CACA,OAAO,CACT,EACA,IAAK,GAAM,CAAC,EAAS,KAAW,EAAmB,CAEjD,IAAM,EAAY,KAAK,MAAM,EAAU,EAAU,MAAM,EACjD,EAAa,EAAU,EAAU,OACjC,EAAO,EAAU,GACjB,EAAQ,EAAU,GAClB,EAAa,EAAU,GACvB,EAAc,EAAU,GAC1B,MAAC,GAAQ,CAAC,GAAS,CAAC,GAAc,CAAC,IAIvC,GAAK,EAAqB,EAAU,EAAE,QAAU,GAAK,IAAM,EAAqB,EAAW,EAAE,QAAU,GAAK,IAGxG,IAAS,IAAM,GAA4B,KAAK,IAAI,EAAW,KAAM,EAAY,IAAI,IAMvF,KAAe,EAAM,CAAK,EAAI,KAC9B,GAA8B,KAAK,IAAI,EAAK,aAAc,EAAM,YAAY,IAQ5E,EAAU,EAAK,SAAU,EAAM,QAAQ,EAAI,KAC3C,EAAuB,KAAK,IAAI,EAAK,SAAS,OAAQ,EAAM,SAAS,MAAM,EAC3E,CACA,IAAM,EAAW,EAAK,CAAS,EACzB,EAAY,EAAK,CAAU,EACjC,EAAO,KAAK,IAAI,EAAU,CAAS,GAAK,KAAK,IAAI,EAAU,CAAS,CACtE,CACF,CAEA,IAAM,EAAgB,IAAI,IAC1B,IAAK,IAAM,KAAS,EAAW,KAAK,EAAG,CACrC,IAAM,EAAO,EAAK,CAAK,EACjB,EAAU,EAAc,IAAI,CAAI,GAAK,CAAC,EAC5C,EAAQ,KAAK,CAAK,EAClB,EAAc,IAAI,EAAM,CAAO,CACjC,CACA,IAAM,EAAgC,CAAC,EACvC,IAAK,IAAM,KAAW,EAAc,OAAO,EAAG,CAC5C,GAAI,EAAQ,OAAS,EACnB,SAEF,IAAM,EAAY,EAAQ,OAAQ,IAAW,EAAqB,EAAM,EAAE,QAAU,KAAO,CAAC,EACtF,EAAU,EAAQ,OAAQ,IAAW,EAAqB,EAAM,EAAE,QAAU,GAAK,CAAC,EACxF,GAAI,EAAQ,SAAW,EAAG,CACxB,EAAO,KAAK,EAAQ,QAAS,GAAW,EAAW,GAAS,CAAC,EAAqB,EAAW,EAAM,CAAC,EAAI,CAAC,CAAE,CAAC,EAC5G,QACF,CACA,GAAI,EAAU,SAAW,EACvB,SAMF,IAAM,EAAkB,GACtB,EAAQ,KAAM,GAAU,CACtB,IAAM,EAAQ,EAAW,GACzB,OACE,IAAU,IAAA,IACV,EAAW,gBAAkB,EAAM,eACnC,EAAM,gBAAkB,EAAW,aAEvC,CAAC,EACG,EAAiB,CAAC,GAAG,IAAI,IAAI,EAAQ,QAAS,GAAU,EAAqB,IAAU,CAAC,CAAC,CAAC,CAAC,CAAC,CAC/F,OAAQ,GAAe,CACtB,IAAM,EAAQ,EAAe,GAC7B,OAAO,IAAU,IAAA,IAAa,EAAM,OAAS,GAAK,EAAM,MAAM,CAAc,CAC9E,CAAC,CAAC,CACD,UAAU,EAAW,IAAe,EAAY,CAAU,EACvD,CAAC,EAAa,GAAG,GAAiB,EACxC,GAAI,IAAgB,IAAA,GAAW,CAO7B,IAAM,EAAW,IAAI,IACf,EAA8B,CAAC,EACrC,IAAK,IAAM,KAAe,EAAS,CACjC,IAAM,EAAQ,EAAW,GACzB,GAAI,CAAC,EACH,SASF,IAAM,EAAqE,CAAC,EAC5E,IAAK,IAAM,KAAc,EACvB,IAAK,IAAM,KAAc,EAAe,IAAe,CAAC,EAEpD,CAAC,EAAS,IAAI,CAAU,GACxB,EAAW,gBAAkB,EAAM,eACnC,EAAM,gBAAkB,EAAW,gBAEnC,EAAS,IAAI,CAAU,EACvB,EAAU,KAAK,CAAE,aAAY,YAAW,CAAC,GAI/C,EAAU,MACP,EAAM,IACL,EAAK,WAAW,gBAAkB,EAAM,WAAW,iBACnD,EAAK,WAAW,cAAgB,EAAM,WAAW,aACrD,EACA,IAAI,EAAiC,CAAC,EAChC,EAAa,IAAI,IACvB,IAAK,GAAM,CAAE,aAAY,gBAAgB,EACnC,EAAW,IAAI,CAAU,IAC3B,EAAO,KAAK,EAAoB,CAAS,CAAC,EAC1C,EAAY,CAAC,EACb,EAAW,MAAM,GAEnB,EAAU,KAAK,CAAU,EACzB,EAAW,IAAI,CAAU,EAEvB,EAAU,OAAS,GACrB,EAAO,KAAK,EAAoB,CAAS,CAAC,EAExC,EAAU,SAAW,IAAM,EAAqB,EAAY,EAAE,QAAU,KAAO,GACjF,EAAO,KAAK,EAAqB,CAAK,CAAC,CAE3C,CACA,EAAO,MACJ,EAAM,IAAU,EAAK,gBAAkB,EAAM,iBAAmB,EAAK,cAAgB,EAAM,aAC9F,EAGA,IAAK,IAAM,KAAc,EACvB,EAAW,sBAAwB,IAAA,GAErC,EAAe,GAAe,EAC9B,IAAK,IAAM,KAAe,EACxB,EAAe,GAAe,CAAC,CAEnC,MAAW,EAAU,QAAU,GAC7B,EAAO,KAAK,EAAU,QAAS,GAAW,EAAW,GAAS,CAAC,EAAqB,EAAW,EAAM,CAAC,EAAI,CAAC,CAAE,CAAC,CAElH,CAEA,OADA,EAAO,KAAK,CAAa,EAClB,CACT,CAUA,SAAS,GAAuB,EAAsC,CACpE,IAAM,EAAsB,CAAC,EACvB,EAAsB,CAAC,EAC7B,IAAK,IAAM,KAAS,EAAU,CAC5B,KAAO,EAAM,OAAS,GAAM,EAAM,GAAG,EAAE,CAAC,CAAgB,MAAM,eAAiB,EAAM,iBACnF,EAAM,IAAI,EAEZ,IAAM,EAAM,EAAM,GAAG,EAAE,EAEvB,GAAI,GAAO,EAAI,MAAM,kBAAoB,EAAM,iBAAmB,EAAI,MAAM,gBAAkB,EAAM,cAClG,SAEF,IAAM,EAAmB,CAAE,QAAO,SAAU,CAAC,CAAE,EAC3C,EACF,EAAI,SAAS,KAAK,CAAI,EAEtB,EAAM,KAAK,CAAI,EAEjB,EAAM,KAAK,CAAI,CACjB,CAEA,IAAM,EAAqB,CAAC,EACtB,EAAS,GAA2B,CACxC,GAAI,EAAS,CAAI,EACf,IAAK,IAAM,KAAS,EAAK,SACvB,EAAM,CAAK,OAGb,EAAK,KAAK,EAAK,KAAK,CAExB,EACA,IAAK,IAAM,KAAQ,EACjB,EAAM,CAAI,EAEZ,OAAO,CACT,CAQA,SAAS,EAAS,EAA2B,CAC3C,GAAM,CAAC,GAAc,EAAK,SAC1B,OAAO,EAAK,SAAS,QAAU,GAAM,IAAe,IAAA,IAAa,EAAS,CAAU,CACtF,CAGA,SAAS,EAAoB,EAAqD,CAChF,IAAM,EAAQ,EAAY,GAC1B,GAAI,CAAC,GAAS,EAAY,SAAW,EACnC,OAAO,GAAS,EAAkB,EAKpC,IAAM,EAA2B,CAAC,EAClC,IAAK,IAAM,KAAW,EACnB,QAAS,GAAe,EAAW,QAAQ,CAAC,CAC5C,UACE,EAAM,IAAU,EAAK,gBAAkB,EAAM,iBAAmB,EAAK,cAAgB,EAAM,aAC9F,EAAG,CACH,IAAM,EAAO,EAAS,GAAG,EAAE,EACvB,GAAQ,EAAQ,gBAAkB,EAAK,cACzC,EAAK,cAAgB,KAAK,IAAI,EAAK,cAAe,EAAQ,aAAa,EAEvE,EAAS,KAAK,CAAE,GAAG,CAAQ,CAAC,CAEhC,CACA,MAAO,CACL,WACA,WAAY,EAAS,QAAQ,EAAK,IAAY,EAAM,EAAQ,cAAgB,EAAQ,gBAAiB,CAAC,EACtG,gBAAiB,KAAK,IAAI,GAAG,EAAY,IAAK,GAAe,EAAW,eAAe,CAAC,EACxF,cAAe,KAAK,IAAI,GAAG,EAAY,IAAK,GAAe,EAAW,aAAa,CAAC,EACpF,WAAY,KAAK,IAAI,GAAG,EAAY,IAAK,GAAe,EAAW,UAAU,CAAC,EAC9E,SAAU,KAAK,IAAI,GAAG,EAAY,IAAK,GAAe,EAAW,QAAQ,CAAC,EAC1E,UAAW,KAAK,IAAI,GAAG,EAAY,IAAK,GAAe,EAAW,SAAS,CAAC,EAC5E,QAAS,KAAK,IAAI,GAAG,EAAY,IAAK,GAAe,EAAW,OAAO,CAAC,CAC1E,CACF,CAGA,SAAS,GAAuC,CAC9C,MAAO,CACL,SAAU,CAAC,EACX,WAAY,EACZ,gBAAiB,EACjB,cAAe,EACf,WAAY,EACZ,SAAU,EACV,UAAW,EACX,QAAS,CACX,CACF,CAGA,SAAS,EAAqB,EAAsC,CAClE,MAAO,CACL,SAAU,CAAC,CAAE,gBAAiB,EAAM,gBAAiB,cAAe,EAAM,aAAc,CAAC,EACzF,WAAY,EAAM,cAAgB,EAAM,gBACxC,gBAAiB,EAAM,gBACvB,cAAe,EAAM,cACrB,WAAY,EAAM,WAClB,SAAU,EAAM,SAChB,UAAW,EAAM,UACjB,QAAS,EAAM,OACjB,CACF,CAiBA,SAAS,GACP,EACA,EACA,EACiB,CACjB,IAAM,EAAW,IAAI,WAAW,EAAM,cAAgB,EAAM,eAAe,EACrE,EAAoB,IAAI,IACxB,EAAuB,IAAI,IAC7B,EAAe,EACnB,IAAK,IAAI,EAAQ,EAAM,gBAAiB,EAAQ,EAAM,cAAe,GAAS,EAAG,CAC/E,IAAM,EAAQ,EAAO,GACrB,GAAI,CAAC,EACH,SAEF,IAAI,EACJ,GAAI,EAAM,OAAS,KAAM,CACvB,IAAI,EAAkB,EAAkB,IAAI,EAAM,IAAI,EAClD,IAAoB,IAAA,KACtB,EAAkB,EAAkB,KACpC,EAAkB,IAAI,EAAM,KAAM,CAAe,GAEnD,EAAQ,EAAE,EAAkB,EAC9B,KAAO,CACL,IAAM,EAAM,GAAG,EAAM,SAAS,GAAG,EAAM,UAAU,GAAG,EAAM,aAAe,EAAE,GAAG,EAAM,cAAgB,IAChG,EAAK,EAAsB,IAAI,CAAG,EAClC,IAAO,IAAA,KACT,EAAK,EAAsB,KAC3B,EAAsB,IAAI,EAAK,CAAE,GAEnC,EAAQ,GACJ,EAAM,SAAW,IAAQ,EAAM,cAAgB,IAAA,MACjD,EAAqB,IAAI,GAAQ,EAAqB,IAAI,CAAK,GAAK,GAAK,CAAC,EAC1E,GAAgB,EAEpB,CACA,EAAS,EAAQ,EAAM,iBAAmB,CAC5C,CACA,MAAO,CAAE,WAAU,uBAAsB,cAAa,CACxD,CAGA,SAAS,GAAe,EAAuB,EAAgC,CAC7E,GAAM,CAAC,EAAS,GACd,EAAK,qBAAqB,MAAQ,EAAM,qBAAqB,KAAO,CAAC,EAAM,CAAK,EAAI,CAAC,EAAO,CAAI,EAC9F,EAAU,EACd,IAAK,GAAM,CAAC,EAAQ,KAAU,EAAQ,qBACpC,GAAW,KAAK,IAAI,EAAO,EAAO,qBAAqB,IAAI,CAAM,GAAK,CAAC,EAEzE,OAAO,CACT,CAGA,SAAS,GAAgB,EAAmC,CAC1D,IAAM,EAAS,IAAI,IACnB,IAAK,IAAI,EAAQ,EAAG,EAAQ,GAAqB,EAAS,OAAQ,GAAS,EAAG,CAC5E,IAAI,EAAO,KACX,IAAK,IAAI,EAAS,EAAG,EAAS,EAAmB,GAAU,EAEzD,EAAQ,KAAK,KAAK,EAAM,EAAE,GAAK,EAAS,EAAQ,IAAW,GAAM,EAEnE,EAAO,IAAI,CAAI,CACjB,CACA,OAAO,CACT,CAGA,SAAS,GAAkB,EAA+C,CACxE,IAAM,EAAgB,IAAI,IAC1B,IAAK,GAAM,CAAC,EAAY,KAAW,EAAU,QAAQ,EAGnD,IAAK,IAAM,KAAS,EAAQ,CAC1B,IAAI,EAAS,EAAc,IAAI,CAAK,EAC/B,IACH,EAAS,CAAC,EACV,EAAc,IAAI,EAAO,CAAM,GAEjC,EAAO,KAAK,CAAU,CACxB,CAEF,IAAM,EAAe,IAAI,IACzB,IAAK,IAAM,KAAU,EAAc,OAAO,EAExC,IAAK,IAAI,EAAe,EAAG,EAAe,EAAO,OAAQ,GAAgB,EAAG,CAC1E,IAAM,EAAY,EAAO,IAAiB,EAC1C,IAAK,IAAI,EAAgB,EAAe,EAAG,EAAgB,EAAO,OAAQ,GAAiB,EAAG,CAC5F,IAAM,EAAU,EAAY,EAAU,QAAU,EAAO,IAAkB,GACzE,EAAa,IAAI,GAAU,EAAa,IAAI,CAAO,GAAK,GAAK,CAAC,CAChE,CACF,CAEF,OAAO,CACT,CASA,SAAgB,EAAU,EAAe,EAAuB,CAC9D,IAAM,EAAa,EAAE,OAAS,KAAQ,EAChC,EAAgB,IAAI,IAC1B,IAAK,GAAM,CAAC,EAAO,KAAW,EAAE,QAAQ,EAAG,CACzC,IAAI,EAAO,EAAc,IAAI,CAAM,EAC9B,IACH,EAAO,IAAI,YAAY,CAAS,EAChC,EAAc,IAAI,EAAQ,CAAI,GAEhC,IAAM,EAAO,IAAU,EAIvB,EAAK,IAAS,EAAK,IAAS,GAAM,IAAM,EAAQ,GAClD,CAEA,IAAM,EAAI,IAAI,YAAY,CAAS,EACnC,IAAK,IAAM,KAAU,EAAG,CACtB,IAAM,EAAY,EAAc,IAAI,CAAM,EAEtC,EAAa,EACb,EAAS,EACb,IAAK,IAAI,EAAO,EAAG,EAAO,EAAW,GAAQ,EAAG,CAC9C,IAAM,EAAW,EAAE,IAAS,EAEtB,IAAM,IAAY,IAAS,GAAK,KAAc,EAE9C,GAAY,GAAY,EAAK,KAAgB,EACnD,EAAa,IAAa,GAC1B,IAAM,EAAa,EAAI,EAAU,EACjC,EAAS,IAAa,GAEtB,EAAE,GAAQ,EAAI,CAAC,CACjB,CACF,CAEA,IAAI,EAAS,EACb,IAAK,IAAM,KAAQ,EACjB,GAAU,EAAS,CAAI,EAEzB,OAAO,CACT,CAEA,SAAS,EAAS,EAAuB,CACvC,IAAI,EAAQ,GAAU,IAAU,EAAK,YAErC,MADA,IAAS,EAAQ,YAAmB,IAAU,EAAK,WAC3C,KAAK,KAAM,GAAS,IAAU,GAAM,UAAe,QAAa,IAAM,GAAM,GACtF,CAEA,SAAS,EACP,EACA,EACA,EACA,EACA,EACoB,CACpB,MAAO,CACL,cACA,WAAY,EAAgB,EAC5B,kBACA,gBACA,WAAY,EAAM,WAClB,SAAU,EAAK,SACf,UAAW,EAAM,UACjB,QAAS,EAAK,OAChB,CACF,CAGA,MAAM,EAAkC,CAAC,EACnC,EAAmC,CAAC,EAE1C,SAAS,EAAoB,EAAuB,CAClD,IAAI,EAAO,EAAsB,GAKjC,OAJI,IAAS,IAAA,KACX,EAAO,EAAS,IAAI,GAAO,EAC3B,EAAsB,GAAS,GAE1B,CACT,CAEA,SAAS,EAAqB,EAAuB,CACnD,IAAI,EAAO,EAAuB,GAKlC,OAJI,IAAS,IAAA,KACX,EAAO,EAAU,IAAI,GAAO,EAC5B,EAAuB,GAAS,GAE3B,CACT,CASA,SAAS,EACP,EACA,EACA,EACA,EACQ,CAGR,GAAM,CAAC,EAAS,GAAa,EAAoB,EAAQ,EAAiB,EADrD,GADC,EAAmB,IAAkB,IAAM,EAAmB,IAAoB,GACtD,EAAgB,CACkC,CAAC,EACrG,MAAO,GAAG,EAAQ,GAAG,EAAU,GAAG,EAAgB,GACpD,CASA,SAAS,EAAgB,EAAiB,EAAyB,EAA+B,CAChG,GAAM,CAAC,EAAS,GAAa,EAAoB,EAAQ,EAAiB,EAAe,EAAK,EAE9F,OAAO,EAAU,KAAK,KAAK,EAAW,EAAE,CAC1C,CAEA,SAAS,EACP,EACA,EACA,EACA,EACkB,CAClB,IAAM,EAAoB,IAAI,IAC1B,EAAU,KACV,EAAY,MAChB,IAAK,IAAI,EAAQ,EAAiB,EAAQ,EAAe,GAAS,EAAG,CACnE,IAAM,EAAQ,EAAO,GACrB,GAAI,CAAC,EACH,SAIF,IAAI,EACA,EACJ,GAAI,EAAM,OAAS,KAAM,CACvB,IAAI,EAAkB,EAAkB,IAAI,EAAM,IAAI,EAClD,IAAoB,IAAA,KACtB,EAAkB,EAAkB,KACpC,EAAkB,IAAI,EAAM,KAAM,CAAe,GAEnD,EAAO,EAAoB,CAAe,EAC1C,EAAQ,EAAqB,CAAe,CAC9C,KACE,GAAO,EAAM,SACb,EAAQ,EAAM,UAGhB,EAAW,KAAK,KAAK,EAAS,EAAE,EAAI,EAAQ,EAC5C,EAAY,KAAK,KAAK,EAAW,EAAE,EAAI,EACnC,GAAqB,EAAM,cAAgB,IAAA,IAAa,EAAM,eAAiB,IAAA,KAEjF,EAAW,KAAK,KAAK,EAAS,EAAE,EAAI,EAAM,YAAe,EACzD,EAAY,KAAK,KAAK,EAAW,EAAE,EAAI,EAAM,aAEjD,CACA,MAAO,CAAC,EAAS,CAAS,CAC5B,CAGA,SAAgB,EAAS,EAAsB,CAC7C,IAAI,EAAO,KACX,IAAK,IAAI,EAAQ,EAAG,EAAQ,EAAK,OAAQ,GAAS,EAEhD,EAAO,KAAK,KAAK,EAAM,EAAE,EAAI,EAAK,WAAW,CAAK,EAEpD,OAAO,CACT,CAGA,SAAS,EAAU,EAAsB,CACvC,IAAI,EAAO,YACX,IAAK,IAAI,EAAQ,EAAG,EAAQ,EAAK,OAAQ,GAAS,EAEhD,EAAO,KAAK,KAAK,EAAO,EAAK,WAAW,CAAK,EAAG,QAAU,EAE5D,OAAO,CACT,CAEA,SAAS,EAAc,EAAc,EAAuB,CAC1D,OAAO,KAAK,KAAK,EAAM,EAAE,EAAI,CAC/B,CAEA,SAAS,EAAgB,EAAmE,CAC1F,IAAM,EAAgC,CAAC,EACvC,IAAK,IAAM,KAAS,EAAQ,OAAO,EAAG,CACpC,IAAM,EAAc,EAAM,IAAK,IAAe,CAC5C,SAAU,CAAC,CAAE,gBAAiB,EAAU,gBAAiB,cAAe,EAAU,aAAc,CAAC,EACjG,WAAY,EAAU,WACtB,gBAAiB,EAAU,gBAC3B,cAAe,EAAU,cACzB,WAAY,EAAU,WACtB,SAAU,EAAU,SACpB,UAAW,EAAU,UACrB,QAAS,EAAU,OACrB,EAAE,EACF,EAAY,MACT,EAAM,IAAU,EAAK,gBAAkB,EAAM,iBAAmB,EAAK,cAAgB,EAAM,aAC9F,EACA,EAAO,KAAK,CAAW,CACzB,CACA,OAAO,CACT,CAmBA,SAAgB,EAAiD,EAAe,EAA6B,CAC3G,GAAI,GAAgB,GAAK,EAAO,OAAS,EACvC,OAAO,EAGT,EAAO,KAAK,CAAa,EACzB,IAAK,IAAI,EAAU,GAAM,GAAU,CACjC,EAAU,GACV,IAAK,IAAI,EAAY,EAAG,EAAY,EAAO,QAAU,CAAC,EAAS,GAAa,EAC1E,IAAK,IAAI,EAAa,EAAY,EAAG,EAAa,EAAO,OAAQ,GAAc,EAAG,CAChF,IAAM,EAAO,EAAO,GACd,EAAQ,EAAO,GACrB,GAAI,CAAC,GAAQ,CAAC,EACZ,SAEF,IAAM,EAAU,EAAY,EAAM,EAAO,CAAY,EAC/C,EAAS,GAAW,EAAY,EAAO,EAAM,CAAY,EAC/D,GAAI,CAAC,EACH,SAEF,IAAM,EAAe,EAAU,EAAO,cAAgB,EAAO,eACvD,EAAgB,EAAU,EAAO,eAAiB,EAAO,cAC3D,GAAgB,GAClB,EAAO,GAAa,EAAO,OAC3B,EAAO,OAAO,EAAY,CAAC,GAClB,EACT,EAAO,GAAc,EAAO,OAE5B,EAAO,GAAa,EAAO,OAM7B,IAAK,IAAM,KAAc,EAAO,eAC9B,EAAW,sBAAwB,GAErC,EAAO,KAAK,CAAa,EACzB,EAAU,GACV,KACF,CAEJ,CACA,OAAO,CACT,CAEA,SAAS,EAAc,EAA2B,EAAoC,CACpF,IAAM,EAAY,EAAK,GACjB,EAAa,EAAM,GACzB,OACG,GAAW,iBAAmB,IAAM,GAAY,iBAAmB,KACnE,GAAW,eAAiB,IAAM,GAAY,eAAiB,EAEpE,CAmBA,SAAS,EACP,EACA,EACA,EAC4B,CAM5B,IAAM,EAAW,EAAM,OAAQ,GAAe,CAAC,EAAW,qBAAqB,EACzE,EAAY,EAAO,OAAQ,GAAe,CAAC,EAAW,qBAAqB,EAC3E,EAAkB,CAAC,EACrB,EAAe,EACf,EAAsB,GAC1B,IAAK,IAAM,KAAY,EAAW,CAEhC,KAAO,EAAe,EAAS,QAAQ,CACrC,IAAM,EAAU,EAAS,GACzB,GAAI,GAAW,EAAQ,cAAgB,EAAe,EAAS,gBAC7D,GAAgB,OAEhB,KAEJ,CACA,IAAM,EAAU,EAAS,GAEvB,GACA,EAAQ,eAAiB,EAAS,iBAClC,EAAQ,iBAAmB,IAE3B,EAAM,KAAK,CAAC,EAAS,CAAQ,CAAC,EAC9B,EAAsB,EAAS,cAC/B,GAAgB,EAEpB,CACA,IAAM,EAAgB,EAAM,SAAW,EAAM,OACvC,EAAiB,EAAM,SAAW,EAAO,OAC3C,OAAM,OAAS,GAAM,CAAC,GAAiB,CAAC,GAe5C,MAAO,CAAE,OAZM,EAAM,KAAK,CAAC,EAAS,MAAe,CACjD,GAAG,EAEH,sBAAuB,IAAA,GACvB,SAAU,CAAC,GAAG,EAAQ,SAAU,GAAG,EAAS,QAAQ,EACpD,WAAY,EAAQ,WAAa,EAAS,WAC1C,cAAe,EAAS,cACxB,SAAU,EAAS,SACnB,QAAS,EAAS,OACpB,EAGc,EAAG,gBAAe,iBAAgB,eAD9C,IAAkB,EAAiB,CAAC,EAAI,EAAM,KAAK,CAAC,EAAS,KAAe,EAAgB,EAAW,CAAQ,CAClD,CACjE,CAcA,SAAgB,EAAwB,EAAoC,CAC1E,IAAI,EAAgB,EAChB,EAAmB,EACnB,EAAsB,GAC1B,IAAK,IAAM,KAAc,EAAO,CAC9B,GAAI,EAAW,sBAAuB,CACpC,EAAsB,GACtB,QACF,CACA,GAAiB,EAAW,SAAS,OACrC,EAAmB,KAAK,IAAI,EAAkB,EAAW,SAAS,MAAM,CAC1E,CACA,OAAO,EAAsB,EAAgB,EAAgB,CAC/D,CAEA,SAAS,GACP,EACA,EACA,EACoB,CACpB,IAAI,EAAsB,EACtB,EAAwB,EACtB,EAAmE,CAAC,EACpE,EAAkB,IAAI,IAC5B,IAAK,IAAM,KAAS,EAAQ,CAC1B,GAAuB,EAAwB,CAAK,EACpD,IAAK,IAAM,KAAc,EAAO,CAC9B,EAAwB,KAAK,IAAI,EAAuB,EAAW,UAAU,EAO7E,IAAK,IAAM,KAAW,EAAW,SAC/B,EAAoB,EAAS,EAAQ,EAAiB,CAAe,CAEzE,CACA,EAAqB,KACnB,EACG,KAAK,CAAE,YAAW,cAAe,CAAE,YAAW,SAAQ,EAAE,CAAC,CACzD,UAAU,EAAM,IAAU,EAAK,UAAY,EAAM,SAAS,CAC/D,CACF,CAGA,OAFA,EAAqB,MAAM,EAAM,KAAW,EAAK,EAAE,EAAE,WAAa,IAAM,EAAM,EAAE,EAAE,WAAa,EAAE,EAE1F,CACL,sBACA,yBAA0B,EAAO,OACjC,uBACA,mBAAoB,EAAgB,KACpC,qBAAsB,CAAC,GAAG,CAAe,CAAC,CAAC,UAAU,EAAM,IAAU,EAAO,CAAK,EACjF,iBAAkB,EAAgB,OAAS,EAAI,EAAI,EAAgB,KAAO,EAAgB,KAC1F,uBACF,CACF,CAGA,SAAgB,EACd,EACA,EACA,EACA,EACM,CACN,IAAK,IAAI,EAAQ,EAAQ,gBAAiB,EAAQ,EAAQ,cAAe,GAAS,EAAG,CACnF,IAAM,EAAQ,EAAO,GACrB,IAAK,IAAI,EAAM,GAAO,UAAY,EAAG,IAAQ,GAAO,QAAU,IAAK,GAAO,GACpE,CAAC,GAAmB,EAAgB,IAAI,EAAM,CAAC,IACjD,EAAgB,IAAI,EAAM,CAAC,CAGjC,CACF"}
+{"version":3,"file":"duplication.js","names":[],"sources":["../src/duplication.ts"],"sourcesContent":["import type { DuplicationOptions } from './types.js';\n\n/**\n * Project-level duplication machinery operating on normalized token streams. Tokenization itself\n * (parsing, identifier anonymization, literal normalization) happens in the Rust addon, which\n * serializes each file's Token stream and statement structure; the helpers here match statement\n * windows across files, merge gap-adjacent groups, and count duplicated lines over that data.\n */\n\nexport const defaultDuplicationOptions: Required<DuplicationOptions> = {\n  minTokens: 40,\n  maxGapTokens: 30,\n  minSimilarityPercent: 70,\n};\n\n/**\n * Fills defaults for absent settings, applying the same normalization as the native boundary's\n * clampToU32 — NaN (e.g. `Number(unsetEnvVariable)`) counts as absent, and other values truncate\n * and clamp to [0, u32::MAX] — so the TypeScript half of cross-file matching cannot diverge from\n * the natively collected candidates on such input.\n */\nexport function resolveDuplicationOptions(options?: DuplicationOptions): Required<DuplicationOptions> {\n  return {\n    minTokens: resolveOption(options?.minTokens, defaultDuplicationOptions.minTokens),\n    maxGapTokens: resolveOption(options?.maxGapTokens, defaultDuplicationOptions.maxGapTokens),\n    minSimilarityPercent: resolveOption(options?.minSimilarityPercent, defaultDuplicationOptions.minSimilarityPercent),\n  };\n}\n\nfunction resolveOption(value: number | undefined, fallback: number): number {\n  return value === undefined || Number.isNaN(value)\n    ? fallback\n    : Math.min(Math.max(Math.trunc(value), 0), 0xFF_FF_FF_FF);\n}\n\n/** Minimum consecutive statements for a statement-sequence duplicate candidate. */\nconst minSequenceStatementCount = 2;\n/**\n * Caps the window length so statement-sequence enumeration stays linear in the statement count.\n * Heterogeneous clones longer than the cap are reported as capped windows (a deliberate\n * conservative undercount trading completeness for bounded discovery cost).\n */\nconst maxSequenceStatementCount = 100;\n\n/**\n * A region whose normalized tokens are at least 20% literal values is data-like (a lookup table, a\n * constant list, a value-mapping switch), not logic: literal values re-enter its fingerprint so\n * tables that merely share their shape stop counting as copy-paste. Compared in integer math\n * (5 * literals >= total) so the TypeScript and native sides cannot disagree on the boundary.\n */\nfunction isLiteralDense(literalCount: number, tokenCount: number): boolean {\n  return literalCount * 5 >= tokenCount;\n}\n\nexport interface Token {\n  /** Normalization target: identifiers to anonymize, literal kind tags, or the raw token text. */\n  kind: 'id' | 'text';\n  text: string;\n  /**\n   * Two INDEPENDENT hashes of `text` (djb2 and FNV-1a), precomputed so fingerprinting nested\n   * regions never re-hashes a token. Feeding the same per-token hash to both fingerprint\n   * accumulators would collapse the key to 32 effective bits: one djb2 collision between two\n   * token texts would then equate whole regions.\n   */\n  textHash: number;\n  textHash2: number;\n  /** Hash pair of a value-carrying literal's value, folded into data-like region fingerprints. */\n  literalHash?: number;\n  literalHash2?: number;\n  /**\n   * True for verbatim-kept NAMES (member/callee/type names, named grammar leaves): together with\n   * value-carrying literals these are the content-bearing tokens the near-miss content gate\n   * counts. Keywords, operators, and punctuation come from unnamed nodes and stay false.\n   */\n  isName?: boolean;\n  /** 0-based source rows the token occupies, so line coverage counts only matched-token lines. */\n  startRow: number;\n  endRow: number;\n}\n\n/** A token span with its source position; plain data so cross-file matching can retain it. */\nexport interface TokenRange {\n  startTokenIndex: number;\n  endTokenIndex: number;\n  startIndex: number;\n  endIndex: number;\n  startLine: number;\n  endLine: number;\n}\n\n/** A contiguous run of matched tokens; gapped (merged) duplicates carry several per occurrence. */\ninterface TokenSegment {\n  startTokenIndex: number;\n  endTokenIndex: number;\n}\n\ninterface DuplicateCandidate {\n  fingerprint: string;\n  tokenCount: number;\n  startTokenIndex: number;\n  endTokenIndex: number;\n  startIndex: number;\n  endIndex: number;\n  startLine: number;\n  endLine: number;\n}\n\nexport interface CountedOccurrence {\n  /** Matched token runs; more than one once gapped groups are merged. */\n  segments: TokenSegment[];\n  /**\n   * Set on a retained group's occurrences that a partial gapped merge also paired into a merged\n   * group: their spans are counted there, so block counting must not count them again.\n   */\n  sharedWithMergedGroup?: boolean;\n  /** Sum of segment token counts (the gap tokens are not matched content). */\n  tokenCount: number;\n  startTokenIndex: number;\n  endTokenIndex: number;\n  startIndex: number;\n  endIndex: number;\n  startLine: number;\n  endLine: number;\n}\n\n/** A duplicate region found in one file, exported for cross-file matching by fingerprint. */\nexport interface CrossFileDuplicateCandidate {\n  /** Content key: equal fingerprints mean equal normalized token sequences (up to hash collision). */\n  fingerprint: string;\n  tokenCount: number;\n  /** Token positions within the owning file, for cross-file gapped (Type-3) merging. */\n  startTokenIndex: number;\n  endTokenIndex: number;\n  startIndex: number;\n  endIndex: number;\n  startLine: number;\n  endLine: number;\n}\n\n/**\n * One file's contribution to cross-file clone detection: catalogued candidates plus the normalized\n * token stream and statement structure, so the project-level pass can match partial statement runs\n * (windows that a single file cannot know repeat elsewhere) and merge gap-adjacent groups.\n */\nexport interface CrossFileDuplicationFileData {\n  candidates: CrossFileDuplicateCandidate[];\n  tokens: Token[];\n  containerStatements: TokenRange[][];\n  /**\n   * 1-based lines that are neither blank nor comment-only, so cross-file line coverage counts only\n   * code lines (blank rows inside multi-row tokens such as template literals carry no content).\n   * Optional for backward compatibility; without it, every matched-token row counts.\n   */\n  codeLineNumbers?: Set<number>;\n}\n\n/** literalCountPrefix[i] = value-carrying literal tokens in tokens[0..i), for O(1) density checks. */\nexport function buildLiteralCountPrefix(tokens: Token[]): Int32Array {\n  const prefix = new Int32Array(tokens.length + 1);\n  for (const [index, token] of tokens.entries()) {\n    prefix[index + 1] = (prefix[index] ?? 0) + (token.literalHash === undefined ? 0 : 1);\n  }\n  return prefix;\n}\n\ninterface WindowOccurrences {\n  count: number;\n  /** -1 once occurrences span more than one container. */\n  containerIndex: number;\n  /** -1 once occurrences span more than one context (file). */\n  contextIndex: number;\n  minStart: number;\n  maxStart: number;\n}\n\ninterface SequenceWindow {\n  containerIndex: number;\n  start: number;\n  length: number;\n}\n\n/** One file's token stream and statement containers, as a window-matching context. */\nexport interface SequenceWindowContext {\n  tokens: Token[];\n  literalCountPrefix: Int32Array;\n  containers: TokenRange[][];\n}\n\nexport interface ContextualSequenceCandidate {\n  candidate: DuplicateCandidate;\n  contextIndex: number;\n}\n\n/**\n * Enumerates runs of consecutive sibling statements over one or more contexts (files). Every\n * container statement participates; only the window length is capped, so enumeration stays linear\n * in the statement count. Windows are grouped by a cheap rolling hash of per-statement\n * fingerprints, and only locally maximal repeated windows — those whose one-statement extensions\n * stop repeating — become candidates with an exact (window-consistent) fingerprint. Without the\n * maximality filter a degenerate file of near-identical statements would fingerprint every\n * sub-window of every repeated region. With `requireMultipleContexts` a window only counts as\n * repeated when its occurrences span at least two contexts (CPD-style cross-file matching): a\n * repeat confined to one file is that file's own concern, and emitting it here would flood the\n * project-level selection with unusable single-file groups.\n */\nexport function collectSequenceWindowCandidates(\n  contexts: SequenceWindowContext[],\n  minTokens: number,\n  requireMultipleContexts: boolean\n): ContextualSequenceCandidate[] {\n  const candidates: ContextualSequenceCandidate[] = [];\n  const contextIndexByContainer: number[] = [];\n  const containers: TokenRange[][] = [];\n  for (const [contextIndex, context] of contexts.entries()) {\n    for (const statements of context.containers) {\n      contextIndexByContainer.push(contextIndex);\n      containers.push(statements);\n    }\n  }\n  const contextAt = (containerIndex: number): SequenceWindowContext | undefined =>\n    contexts[contextIndexByContainer[containerIndex] ?? 0];\n  const occurrencesByWindowKey = new Map<number, WindowOccurrences>();\n  const containerWindows = containers.map((statements, containerIndex) =>\n    enumerateContainerWindows(contextAt(containerIndex)?.tokens ?? [], statements, minTokens)\n  );\n  for (const [containerIndex, windows] of containerWindows.entries()) {\n    const contextIndex = contextIndexByContainer[containerIndex] ?? 0;\n    for (const [start, row] of windows.windowKeysByStart.entries()) {\n      for (const windowKey of row) {\n        if (windowKey === undefined) {\n          continue;\n        }\n        const occurrences = occurrencesByWindowKey.get(windowKey);\n        if (occurrences) {\n          occurrences.count += 1;\n          if (occurrences.containerIndex !== containerIndex) {\n            occurrences.containerIndex = -1;\n          }\n          if (occurrences.contextIndex !== contextIndex) {\n            occurrences.contextIndex = -1;\n          }\n          occurrences.minStart = Math.min(occurrences.minStart, start);\n          occurrences.maxStart = Math.max(occurrences.maxStart, start);\n        } else {\n          occurrencesByWindowKey.set(windowKey, {\n            count: 1,\n            containerIndex,\n            contextIndex,\n            minStart: start,\n            maxStart: start,\n          });\n        }\n      }\n    }\n  }\n\n  // A window only \"repeats\" when two of its occurrences can coexist without overlapping: sliding\n  // matches inside a homogeneous run (start spread smaller than the window length) can never both\n  // be counted and must neither qualify a window nor dominate its sub-windows. Cross-context\n  // matching instead requires occurrences in two contexts, which coexist by construction.\n  const repeats = (windowKey: number | undefined, length: number): boolean => {\n    if (windowKey === undefined) {\n      return false;\n    }\n    const occurrences = occurrencesByWindowKey.get(windowKey);\n    if (occurrences === undefined || occurrences.count < 2) {\n      return false;\n    }\n    if (requireMultipleContexts) {\n      return occurrences.contextIndex === -1;\n    }\n    return occurrences.containerIndex === -1 || occurrences.maxStart - occurrences.minStart >= length;\n  };\n\n  // A window whose statements all share one normalized shape (sixteen `let x = 0;` declarations,\n  // a constant table) is a homogeneous preamble, not a copy-paste: requiring two distinct\n  // per-statement shapes keeps such runs out of duplicate groups and the duplication ratio.\n  const hasDistinctStatements = (window: SequenceWindow): boolean => {\n    const hashes = containerWindows[window.containerIndex]?.statementHashes ?? [];\n    const firstHash = hashes[window.start];\n    for (let index = window.start + 1; index < window.start + window.length; index += 1) {\n      if (hashes[index] !== firstHash) {\n        return true;\n      }\n    }\n    return false;\n  };\n\n  const maximalWindows: SequenceWindow[] = [];\n  for (const [containerIndex, windows] of containerWindows.entries()) {\n    for (const [start, row] of windows.windowKeysByStart.entries()) {\n      for (const [length, windowKey] of row.entries()) {\n        if (!repeats(windowKey, length) || !hasDistinctStatements({ containerIndex, start, length })) {\n          continue;\n        }\n        // Dominated windows are skipped: the one-statement extension also repeats, so a larger\n        // candidate covering this window exists.\n        const extendedRight = windows.windowKeysByStart[start]?.[length + 1];\n        const extendedLeft = windows.windowKeysByStart[start - 1]?.[length + 1];\n        if (repeats(extendedRight, length + 1) || repeats(extendedLeft, length + 1)) {\n          continue;\n        }\n        maximalWindows.push({ containerIndex, start, length });\n      }\n    }\n  }\n\n  // The rolling hash anonymizes identifiers per statement, so a window can look repeated coarsely\n  // while its exact (window-consistent) fingerprints differ, and a longer window's match can\n  // dominate sub-windows that other copies still need (three copies where only two extend one\n  // statement further). Every emitted window therefore exposes its repeating, unvisited\n  // sub-windows; `visited` bounds the worklist and lengths strictly decrease, so it terminates.\n  const visited = new Set(maximalWindows.map(windowId));\n  let frontier = maximalWindows;\n  while (frontier.length > 0) {\n    const emitted: SequenceWindow[] = [];\n    for (const window of frontier) {\n      const statements = containers[window.containerIndex];\n      const first = statements?.[window.start];\n      const last = statements?.[window.start + window.length - 1];\n      const context = contextAt(window.containerIndex);\n      if (!first || !last || !context) {\n        continue;\n      }\n      const fingerprint = `s:${fingerprintKey(context.tokens, context.literalCountPrefix, first.startTokenIndex, last.endTokenIndex)}`;\n      candidates.push({\n        candidate: toCandidate(fingerprint, first.startTokenIndex, last.endTokenIndex, first, last),\n        contextIndex: contextIndexByContainer[window.containerIndex] ?? 0,\n      });\n      emitted.push(window);\n    }\n    frontier = [];\n    for (const window of emitted) {\n      for (const start of [window.start, window.start + 1]) {\n        const subWindow = { containerIndex: window.containerIndex, start, length: window.length - 1 };\n        const subWindowKey = containerWindows[window.containerIndex]?.windowKeysByStart[start]?.[subWindow.length];\n        if (\n          visited.has(windowId(subWindow)) ||\n          !repeats(subWindowKey, subWindow.length) ||\n          !hasDistinctStatements(subWindow)\n        ) {\n          continue;\n        }\n        visited.add(windowId(subWindow));\n        frontier.push(subWindow);\n      }\n    }\n  }\n  return candidates;\n}\n\nfunction windowId(window: SequenceWindow): string {\n  return `${window.containerIndex}:${window.start}:${window.length}`;\n}\n\ninterface ContainerWindows {\n  /** windowKeysByStart[start][length] is the rolling-hash key of the window, or undefined if it is below the size thresholds. */\n  windowKeysByStart: (number | undefined)[][];\n  /** Per-statement fingerprint hashes, for the distinct-shape requirement on windows. */\n  statementHashes: number[];\n}\n\nfunction enumerateContainerWindows(tokens: Token[], statements: TokenRange[], minTokens: number): ContainerWindows {\n  const statementHashes = statements.map((statement) =>\n    fingerprintHash(tokens, statement.startTokenIndex, statement.endTokenIndex)\n  );\n  const windowKeysByStart: (number | undefined)[][] = [];\n  for (let start = 0; start < statements.length; start += 1) {\n    const row: (number | undefined)[] = [];\n    let hash = 5381;\n    let tokenCount = 0;\n    const maxEnd = Math.min(statements.length, start + maxSequenceStatementCount);\n    for (let end = start; end < maxEnd; end += 1) {\n      const statement = statements[end];\n      const statementHash = statementHashes[end];\n      if (!statement || statementHash === undefined) {\n        break;\n      }\n      hash = combineHashes(hash, statementHash);\n      tokenCount += statement.endTokenIndex - statement.startTokenIndex;\n      const statementCount = end - start + 1;\n      row[statementCount] =\n        statementCount >= minSequenceStatementCount && tokenCount >= minTokens\n          ? combineHashes(hash, statementCount)\n          : undefined;\n    }\n    windowKeysByStart.push(row);\n  }\n  return { windowKeysByStart, statementHashes };\n}\n\n/**\n * Longest-common-subsequence LENGTH of two symbol sequences via the Allison–Dix bit-parallel\n * recurrence (O(|a|/32 · |b|) words): per symbol of `b`, `x = match | v` and\n * `v = x & ~(x - ((v << 1) | 1))` over multi-word bit vectors; the set bits of `v` count the LCS.\n */\nexport function lcsLength(a: Int32Array, b: Int32Array): number {\n  const wordCount = (a.length + 31) >>> 5;\n  const positionMasks = new Map<number, Uint32Array>();\n  for (const [index, symbol] of a.entries()) {\n    let mask = positionMasks.get(symbol);\n    if (!mask) {\n      mask = new Uint32Array(wordCount);\n      positionMasks.set(symbol, mask);\n    }\n    const word = index >>> 5;\n    // The Uint32Array store wraps the signed int32 bit pattern to unsigned. The `?? 0` guards in\n    // this function are required by noUncheckedIndexedAccess (typed-array reads type as\n    // `number | undefined`), not redundancy: every index is in bounds.\n    mask[word] = (mask[word] ?? 0) | (1 << (index & 31));\n  }\n\n  const v = new Uint32Array(wordCount);\n  for (const symbol of b) {\n    const matchMask = positionMasks.get(symbol);\n    // `(v << 1) | 1` shifts a carry bit across words; subtraction borrows across words.\n    let shiftCarry = 1;\n    let borrow = 0;\n    for (let word = 0; word < wordCount; word += 1) {\n      const previous = v[word] ?? 0;\n      // oxlint-disable-next-line unicorn/prefer-math-trunc -- `>>> 0` reinterprets the signed int32 bit pattern as unsigned so the borrow subtraction below compares magnitudes; Math.trunc would keep it negative.\n      const x = ((matchMask?.[word] ?? 0) | previous) >>> 0;\n      // oxlint-disable-next-line unicorn/prefer-math-trunc -- same unsigned reinterpretation as `x`.\n      const shifted = ((previous << 1) | shiftCarry) >>> 0;\n      shiftCarry = previous >>> 31;\n      const difference = x - shifted - borrow;\n      borrow = difference < 0 ? 1 : 0;\n      // The Uint32Array store wraps the signed int32 bit pattern to unsigned.\n      v[word] = x & ~difference;\n    }\n  }\n\n  let length = 0;\n  for (const word of v) {\n    length += popCount(word);\n  }\n  return length;\n}\n\nfunction popCount(value: number): number {\n  let count = value - ((value >>> 1) & 0x55_55_55_55);\n  count = (count & 0x33_33_33_33) + ((count >>> 2) & 0x33_33_33_33);\n  return (Math.imul((count + (count >>> 4)) & 0x0F_0F_0F_0F, 0x01_01_01_01) >>> 24) & 0xFF;\n}\n\nfunction toCandidate(\n  fingerprint: string,\n  startTokenIndex: number,\n  endTokenIndex: number,\n  first: TokenRange,\n  last: TokenRange\n): DuplicateCandidate {\n  return {\n    fingerprint,\n    tokenCount: endTokenIndex - startTokenIndex,\n    startTokenIndex,\n    endTokenIndex,\n    startIndex: first.startIndex,\n    endIndex: last.endIndex,\n    startLine: first.startLine,\n    endLine: last.endLine,\n  };\n}\n\n/** Caches of hashText/hashText2 over '$0', '$1', ... so anonymized identifiers hash without allocating. */\nconst anonymizedIndexHashes: number[] = [];\nconst anonymizedIndexHashes2: number[] = [];\n\nfunction anonymizedIndexHash(index: number): number {\n  let hash = anonymizedIndexHashes[index];\n  if (hash === undefined) {\n    hash = hashText(`$${index}`);\n    anonymizedIndexHashes[index] = hash;\n  }\n  return hash;\n}\n\nfunction anonymizedIndexHash2(index: number): number {\n  let hash = anonymizedIndexHashes2[index];\n  if (hash === undefined) {\n    hash = hashText2(`$${index}`);\n    anonymizedIndexHashes2[index] = hash;\n  }\n  return hash;\n}\n\n/**\n * Content key of a token range: two independent 32-bit hashes over the normalized token sequence\n * (identifiers anonymized consistently by first-occurrence order) plus the token count. Regions\n * with equal keys are treated as equal content; a collision would need both 32-bit hashes and the\n * length to coincide, which is negligible for a metrics report. The format and arithmetic match\n * fingerprint_key in native/src/duplication.rs exactly, so window candidates fingerprinted here\n * group together with the per-file candidates the addon catalogues.\n */\nfunction fingerprintKey(\n  tokens: Token[],\n  literalCountPrefix: Int32Array,\n  startTokenIndex: number,\n  endTokenIndex: number\n): string {\n  const literalCount = (literalCountPrefix[endTokenIndex] ?? 0) - (literalCountPrefix[startTokenIndex] ?? 0);\n  const literalDense = isLiteralDense(literalCount, endTokenIndex - startTokenIndex);\n  const [primary, secondary] = fingerprintHashPair(tokens, startTokenIndex, endTokenIndex, literalDense);\n  return `${primary}:${secondary}:${endTokenIndex - startTokenIndex}`;\n}\n\n/**\n * A single 32-bit summary of a range for the coarse rolling-hash phase. Deliberately\n * density-agnostic: density is a property of the final candidate REGION, and folding literal\n * values into per-statement hashes would make a dense statement inside a logic-heavy window\n * (`const weights = [1, 2, 3];`) block the window from ever being enumerated. The coarse phase\n * over-approximates on shape alone; the exact region fingerprint still applies the density rule.\n */\nfunction fingerprintHash(tokens: Token[], startTokenIndex: number, endTokenIndex: number): number {\n  const [primary, secondary] = fingerprintHashPair(tokens, startTokenIndex, endTokenIndex, false);\n  // XOR already coerces to int32, matching the native side's i32 arithmetic.\n  return primary ^ Math.imul(secondary, 31);\n}\n\nfunction fingerprintHashPair(\n  tokens: Token[],\n  startTokenIndex: number,\n  endTokenIndex: number,\n  foldLiteralValues: boolean\n): [number, number] {\n  const indexByIdentifier = new Map<string, number>();\n  let primary = 5381;\n  let secondary = 52_711;\n  for (let index = startTokenIndex; index < endTokenIndex; index += 1) {\n    const token = tokens[index];\n    if (!token) {\n      continue;\n    }\n    // Each accumulator consumes its own independent per-token hash: sharing one would collapse\n    // the key to 32 effective bits (a single djb2 collision would equate whole regions).\n    let part: number;\n    let part2: number;\n    if (token.kind === 'id') {\n      let identifierIndex = indexByIdentifier.get(token.text);\n      if (identifierIndex === undefined) {\n        identifierIndex = indexByIdentifier.size;\n        indexByIdentifier.set(token.text, identifierIndex);\n      }\n      part = anonymizedIndexHash(identifierIndex);\n      part2 = anonymizedIndexHash2(identifierIndex);\n    } else {\n      part = token.textHash;\n      part2 = token.textHash2;\n    }\n    // oxlint-disable-next-line unicorn/prefer-math-trunc -- `| 0` wraps the sum to int32 (Math.trunc does not), which must match the native side's wrapping i32 arithmetic.\n    primary = (Math.imul(primary, 31) + part) | 0;\n    secondary = Math.imul(secondary, 37) ^ part2;\n    if (foldLiteralValues && token.literalHash !== undefined && token.literalHash2 !== undefined) {\n      // oxlint-disable-next-line unicorn/prefer-math-trunc -- `| 0` wraps the sum to int32 (Math.trunc does not), which must match the native side's wrapping i32 arithmetic.\n      primary = (Math.imul(primary, 31) + token.literalHash) | 0;\n      secondary = Math.imul(secondary, 37) ^ token.literalHash2;\n    }\n  }\n  return [primary, secondary];\n}\n\n/** djb2-style hash; XOR keeps the value in signed 32-bit range, which is fine for a grouping key. */\nfunction hashText(text: string): number {\n  let hash = 5381;\n  for (let index = 0; index < text.length; index += 1) {\n    // oxlint-disable-next-line unicorn/prefer-code-point -- djb2 hashes UTF-16 code units; codePointAt would hash surrogate pairs twice (full code point, then the lone low surrogate).\n    hash = Math.imul(hash, 33) ^ text.charCodeAt(index);\n  }\n  return hash;\n}\n\n/** FNV-1a over UTF-16 code units: independent of hashText so the two accumulators never share input. */\nfunction hashText2(text: string): number {\n  let hash = -2_128_831_035; // 2166136261 as int32 (the FNV-1a offset basis)\n  for (let index = 0; index < text.length; index += 1) {\n    // oxlint-disable-next-line unicorn/prefer-code-point -- hashes UTF-16 code units like hashText.\n    hash = Math.imul(hash ^ text.charCodeAt(index), 16_777_619);\n  }\n  return hash;\n}\n\nfunction combineHashes(hash: number, value: number): number {\n  return Math.imul(hash, 31) + value;\n}\n\n/**\n * Merges duplicate groups separated by a small token gap into one gapped (Type-3) clone group: a\n * copy edited in one spot splits into two exact groups whose occurrences sit side by side in the\n * same order. Occurrences are paired greedily in source order; a merge happens when the pairing\n * fully consumes at least one group with at least two pairs. Equal-cardinality groups whose\n * occurrences all pair merge into one group as before. When cardinalities differ (a fragment also\n * occurs standalone: prefix ×3, suffix ×2), the fully-paired group is subsumed into the merged\n * gapped group while the other group is RETAINED with ALL its occurrences: dropping the leftover\n * would lose duplicated-line coverage, and reporting it alone would make a single-occurrence group\n * (contradicting duplicateBlockGroupCount's \"appears more than once\" meaning). Line coverage\n * unions ranges, so the overlap between the retained exact group and the merged group is harmless.\n * Merging repeats to a fixpoint so a clone edited in several spots still reassembles; it\n * terminates because a full merge shrinks the group count and a partial merge keeps it while\n * strictly growing the bounded total span of merged occurrences. Gap tokens are not matched\n * content: line coverage and sizes count only the matched segments. Generic so cross-file merging\n * can thread file identity through occurrences.\n */\nexport function mergeAdjacentGroups<T extends CountedOccurrence>(groups: T[][], maxGapTokens: number): T[][] {\n  if (maxGapTokens <= 0 || groups.length < 2) {\n    return groups;\n  }\n  // Deterministic processing order (mirrored by the native side): by first occurrence position.\n  groups.sort(compareGroups);\n  for (let restart = true; restart;) {\n    restart = false;\n    for (let leftIndex = 0; leftIndex < groups.length && !restart; leftIndex += 1) {\n      for (let rightIndex = leftIndex + 1; rightIndex < groups.length; rightIndex += 1) {\n        const left = groups[leftIndex];\n        const right = groups[rightIndex];\n        if (!left || !right) {\n          continue;\n        }\n        const forward = mergeGroups(left, right, maxGapTokens);\n        const result = forward ?? mergeGroups(right, left, maxGapTokens);\n        if (!result) {\n          continue;\n        }\n        const leftConsumed = forward ? result.firstConsumed : result.secondConsumed;\n        const rightConsumed = forward ? result.secondConsumed : result.firstConsumed;\n        if (leftConsumed && rightConsumed) {\n          groups[leftIndex] = result.merged;\n          groups.splice(rightIndex, 1);\n        } else if (rightConsumed) {\n          groups[rightIndex] = result.merged;\n        } else {\n          groups[leftIndex] = result.merged;\n        }\n        // A partial merge retains the not-fully-consumed group with ALL its occurrences (line\n        // coverage must not shrink, and a reported group must keep >= 2 occurrences), so its\n        // paired occurrences now also live inside the merged group's occurrences: mark them so\n        // duplicateBlockCount counts each token span once.\n        for (const occurrence of result.pairedRetained) {\n          occurrence.sharedWithMergedGroup = true;\n        }\n        groups.sort(compareGroups);\n        restart = true;\n        break;\n      }\n    }\n  }\n  return groups;\n}\n\nfunction compareGroups(left: CountedOccurrence[], right: CountedOccurrence[]): number {\n  const leftFirst = left[0];\n  const rightFirst = right[0];\n  return (\n    (leftFirst?.startTokenIndex ?? 0) - (rightFirst?.startTokenIndex ?? 0) ||\n    (leftFirst?.endTokenIndex ?? 0) - (rightFirst?.endTokenIndex ?? 0)\n  );\n}\n\ninterface MergeResult<T> {\n  merged: T[];\n  /** Whether every occurrence of the respective input group was paired into the merge. */\n  firstConsumed: boolean;\n  secondConsumed: boolean;\n  /** The retained (not fully consumed) group's occurrences that were paired into the merge. */\n  pairedRetained: T[];\n}\n\n/**\n * Pairs `second` occurrences with gap-preceding `first` occurrences, greedily in source order:\n * each trailing occurrence takes the earliest unused leading occurrence within the gap, and a\n * pair's leading must start at or after the previous pair's trailing end so merged spans never\n * overlap. A merge needs at least two pairs (a merged group must still mean \"appears more than\n * once\") and must fully consume at least one group; for equal cardinalities this reduces to the\n * strict all-pairs merge, so pre-partial-merge behavior is unchanged there.\n */\nfunction mergeGroups<T extends CountedOccurrence>(\n  first: T[],\n  second: T[],\n  maxGapTokens: number\n): MergeResult<T> | undefined {\n  // Occurrences a previous partial merge already paired into a merged group must not pair again:\n  // their spans already live inside that merged group, so re-pairing them would assemble a second,\n  // competing merged group instead of letting the existing merged group extend (and would count\n  // the same span twice). Consumption is still judged against the FULL group, so a group holding\n  // shared occurrences is never subsumed away.\n  const leadings = first.filter((occurrence) => !occurrence.sharedWithMergedGroup);\n  const trailings = second.filter((occurrence) => !occurrence.sharedWithMergedGroup);\n  const pairs: [T, T][] = [];\n  let leadingIndex = 0;\n  let previousTrailingEnd = -1;\n  for (const trailing of trailings) {\n    // Leadings ending too far before this trailing can never pair a later (even farther) one.\n    while (leadingIndex < leadings.length) {\n      const leading = leadings[leadingIndex];\n      if (leading && leading.endTokenIndex + maxGapTokens < trailing.startTokenIndex) {\n        leadingIndex += 1;\n      } else {\n        break;\n      }\n    }\n    const leading = leadings[leadingIndex];\n    if (\n      leading &&\n      leading.endTokenIndex <= trailing.startTokenIndex &&\n      leading.startTokenIndex >= previousTrailingEnd\n    ) {\n      pairs.push([leading, trailing]);\n      previousTrailingEnd = trailing.endTokenIndex;\n      leadingIndex += 1;\n    }\n  }\n  const firstConsumed = pairs.length === first.length;\n  const secondConsumed = pairs.length === second.length;\n  if (pairs.length < 2 || (!firstConsumed && !secondConsumed)) {\n    return undefined;\n  }\n  const merged = pairs.map(([leading, trailing]) => ({\n    ...leading,\n    // A merged occurrence is a fresh span combination; it never inherits shared-span marks.\n    sharedWithMergedGroup: undefined,\n    segments: [...leading.segments, ...trailing.segments],\n    tokenCount: leading.tokenCount + trailing.tokenCount,\n    endTokenIndex: trailing.endTokenIndex,\n    endIndex: trailing.endIndex,\n    endLine: trailing.endLine,\n  }));\n  const pairedRetained =\n    firstConsumed === secondConsumed ? [] : pairs.map(([leading, trailing]) => (firstConsumed ? trailing : leading));\n  return { merged, firstConsumed, secondConsumed, pairedRetained };\n}\n\n/**\n * Redundant copies one group adds to duplicateBlockCount. Each redundant occurrence contributes\n * one count per matched fragment, so merging a gapped clone's fragments into one group does not\n * halve duplicateBlockCount: an edited two-fragment pair still counts 2, exactly as its unmerged\n * fragments did. Occurrence shapes can differ within one group (a\n * gap-merged exact pair plus an appended whole-block near-miss copy), so every occurrence's\n * fragments are summed and one representative — the largest — is deducted, keeping the count\n * independent of source order. Occurrences a partial gapped merge also paired into a merged group\n * are skipped: their spans are already counted there, and such a retained group deducts no\n * representative of its own — the merged group's representative already stands for the shared\n * content — so no token span contributes to the count twice.\n */\nexport function countRedundantFragments(group: CountedOccurrence[]): number {\n  let fragmentCount = 0;\n  let maxFragmentCount = 0;\n  let hasSharedOccurrence = false;\n  for (const occurrence of group) {\n    if (occurrence.sharedWithMergedGroup) {\n      hasSharedOccurrence = true;\n      continue;\n    }\n    fragmentCount += occurrence.segments.length;\n    maxFragmentCount = Math.max(maxFragmentCount, occurrence.segments.length);\n  }\n  return hasSharedOccurrence ? fragmentCount : fragmentCount - maxFragmentCount;\n}\n\n/** Adds the 1-based code lines the segment's matched tokens cover; shared with cross-file coverage. */\nexport function collectSegmentLines(\n  segment: { startTokenIndex: number; endTokenIndex: number },\n  tokens: Token[],\n  codeLineNumbers: Set<number> | undefined,\n  duplicatedLines: Set<number>\n): void {\n  for (let index = segment.startTokenIndex; index < segment.endTokenIndex; index += 1) {\n    const token = tokens[index];\n    for (let row = token?.startRow ?? 0; row <= (token?.endRow ?? -1); row += 1) {\n      if (!codeLineNumbers || codeLineNumbers.has(row + 1)) {\n        duplicatedLines.add(row + 1);\n      }\n    }\n  }\n}\n"],"mappings":"AASA,MAAa,EAA0D,CACrE,UAAW,GACX,aAAc,GACd,qBAAsB,EACxB,EAQA,SAAgB,EAA0B,EAA4D,CACpG,MAAO,CACL,UAAW,EAAc,GAAS,UAAW,EAA0B,SAAS,EAChF,aAAc,EAAc,GAAS,aAAc,EAA0B,YAAY,EACzF,qBAAsB,EAAc,GAAS,qBAAsB,EAA0B,oBAAoB,CACnH,CACF,CAEA,SAAS,EAAc,EAA2B,EAA0B,CAC1E,OAAO,IAAU,IAAA,IAAa,OAAO,MAAM,CAAK,EAC5C,EACA,KAAK,IAAI,KAAK,IAAI,KAAK,MAAM,CAAK,EAAG,CAAC,EAAG,UAAa,CAC5D,CAiBA,SAAS,EAAe,EAAsB,EAA6B,CACzE,OAAO,EAAe,GAAK,CAC7B,CAyGA,SAAgB,EAAwB,EAA6B,CACnE,IAAM,EAAS,IAAI,WAAW,EAAO,OAAS,CAAC,EAC/C,IAAK,GAAM,CAAC,EAAO,KAAU,EAAO,QAAQ,EAC1C,EAAO,EAAQ,IAAM,EAAO,IAAU,IAAM,EAAM,cAAgB,IAAA,GAAY,EAAI,GAEpF,OAAO,CACT,CA0CA,SAAgB,EACd,EACA,EACA,EAC+B,CAC/B,IAAM,EAA4C,CAAC,EAC7C,EAAoC,CAAC,EACrC,EAA6B,CAAC,EACpC,IAAK,GAAM,CAAC,EAAc,KAAY,EAAS,QAAQ,EACrD,IAAK,IAAM,KAAc,EAAQ,WAC/B,EAAwB,KAAK,CAAY,EACzC,EAAW,KAAK,CAAU,EAG9B,IAAM,EAAa,GACjB,EAAS,EAAwB,IAAmB,GAChD,EAAyB,IAAI,IAC7B,EAAmB,EAAW,KAAK,EAAY,IACnD,EAA0B,EAAU,CAAc,CAAC,EAAE,QAAU,CAAC,EAAG,EAAY,CAAS,CAC1F,EACA,IAAK,GAAM,CAAC,EAAgB,KAAY,EAAiB,QAAQ,EAAG,CAClE,IAAM,EAAe,EAAwB,IAAmB,EAChE,IAAK,GAAM,CAAC,EAAO,KAAQ,EAAQ,kBAAkB,QAAQ,EAC3D,IAAK,IAAM,KAAa,EAAK,CAC3B,GAAI,IAAc,IAAA,GAChB,SAEF,IAAM,EAAc,EAAuB,IAAI,CAAS,EACpD,GACF,EAAY,OAAS,EACjB,EAAY,iBAAmB,IACjC,EAAY,eAAiB,IAE3B,EAAY,eAAiB,IAC/B,EAAY,aAAe,IAE7B,EAAY,SAAW,KAAK,IAAI,EAAY,SAAU,CAAK,EAC3D,EAAY,SAAW,KAAK,IAAI,EAAY,SAAU,CAAK,GAE3D,EAAuB,IAAI,EAAW,CACpC,MAAO,EACP,iBACA,eACA,SAAU,EACV,SAAU,CACZ,CAAC,CAEL,CAEJ,CAMA,IAAM,GAAW,EAA+B,IAA4B,CAC1E,GAAI,IAAc,IAAA,GAChB,MAAO,GAET,IAAM,EAAc,EAAuB,IAAI,CAAS,EAOxD,OANI,IAAgB,IAAA,IAAa,EAAY,MAAQ,EAC5C,GAEL,EACK,EAAY,eAAiB,GAE/B,EAAY,iBAAmB,IAAM,EAAY,SAAW,EAAY,UAAY,CAC7F,EAKM,EAAyB,GAAoC,CACjE,IAAM,EAAS,EAAiB,EAAO,eAAe,EAAE,iBAAmB,CAAC,EACtE,EAAY,EAAO,EAAO,OAChC,IAAK,IAAI,EAAQ,EAAO,MAAQ,EAAG,EAAQ,EAAO,MAAQ,EAAO,OAAQ,GAAS,EAChF,GAAI,EAAO,KAAW,EACpB,MAAO,GAGX,MAAO,EACT,EAEM,EAAmC,CAAC,EAC1C,IAAK,GAAM,CAAC,EAAgB,KAAY,EAAiB,QAAQ,EAC/D,IAAK,GAAM,CAAC,EAAO,KAAQ,EAAQ,kBAAkB,QAAQ,EAC3D,IAAK,GAAM,CAAC,EAAQ,KAAc,EAAI,QAAQ,EAAG,CAC/C,GAAI,CAAC,EAAQ,EAAW,CAAM,GAAK,CAAC,EAAsB,CAAE,iBAAgB,QAAO,QAAO,CAAC,EACzF,SAIF,IAAM,EAAgB,EAAQ,kBAAkB,EAAM,GAAG,EAAS,GAC5D,EAAe,EAAQ,kBAAkB,EAAQ,EAAE,GAAG,EAAS,GACjE,EAAQ,EAAe,EAAS,CAAC,GAAK,EAAQ,EAAc,EAAS,CAAC,GAG1E,EAAe,KAAK,CAAE,iBAAgB,QAAO,QAAO,CAAC,CACvD,CASJ,IAAM,EAAU,IAAI,IAAI,EAAe,IAAI,CAAQ,CAAC,EAChD,EAAW,EACf,KAAO,EAAS,OAAS,GAAG,CAC1B,IAAM,EAA4B,CAAC,EACnC,IAAK,IAAM,KAAU,EAAU,CAC7B,IAAM,EAAa,EAAW,EAAO,gBAC/B,EAAQ,IAAa,EAAO,OAC5B,EAAO,IAAa,EAAO,MAAQ,EAAO,OAAS,GACnD,EAAU,EAAU,EAAO,cAAc,EAC/C,GAAI,CAAC,GAAS,CAAC,GAAQ,CAAC,EACtB,SAEF,IAAM,EAAc,KAAK,EAAe,EAAQ,OAAQ,EAAQ,mBAAoB,EAAM,gBAAiB,EAAK,aAAa,IAC7H,EAAW,KAAK,CACd,UAAW,EAAY,EAAa,EAAM,gBAAiB,EAAK,cAAe,EAAO,CAAI,EAC1F,aAAc,EAAwB,EAAO,iBAAmB,CAClE,CAAC,EACD,EAAQ,KAAK,CAAM,CACrB,CACA,EAAW,CAAC,EACZ,IAAK,IAAM,KAAU,EACnB,IAAK,IAAM,IAAS,CAAC,EAAO,MAAO,EAAO,MAAQ,CAAC,EAAG,CACpD,IAAM,EAAY,CAAE,eAAgB,EAAO,eAAgB,QAAO,OAAQ,EAAO,OAAS,CAAE,EACtF,EAAe,EAAiB,EAAO,eAAe,EAAE,kBAAkB,EAAM,GAAG,EAAU,QAEjG,EAAQ,IAAI,EAAS,CAAS,CAAC,GAC/B,CAAC,EAAQ,EAAc,EAAU,MAAM,GACvC,CAAC,EAAsB,CAAS,IAIlC,EAAQ,IAAI,EAAS,CAAS,CAAC,EAC/B,EAAS,KAAK,CAAS,EACzB,CAEJ,CACA,OAAO,CACT,CAEA,SAAS,EAAS,EAAgC,CAChD,MAAO,GAAG,EAAO,eAAe,GAAG,EAAO,MAAM,GAAG,EAAO,QAC5D,CASA,SAAS,EAA0B,EAAiB,EAA0B,EAAqC,CACjH,IAAM,EAAkB,EAAW,IAAK,GACtC,EAAgB,EAAQ,EAAU,gBAAiB,EAAU,aAAa,CAC5E,EACM,EAA8C,CAAC,EACrD,IAAK,IAAI,EAAQ,EAAG,EAAQ,EAAW,OAAQ,GAAS,EAAG,CACzD,IAAM,EAA8B,CAAC,EACjC,EAAO,KACP,EAAa,EACX,EAAS,KAAK,IAAI,EAAW,OAAQ,EAAQ,GAAyB,EAC5E,IAAK,IAAI,EAAM,EAAO,EAAM,EAAQ,GAAO,EAAG,CAC5C,IAAM,EAAY,EAAW,GACvB,EAAgB,EAAgB,GACtC,GAAI,CAAC,GAAa,IAAkB,IAAA,GAClC,MAEF,EAAO,EAAc,EAAM,CAAa,EACxC,GAAc,EAAU,cAAgB,EAAU,gBAClD,IAAM,EAAiB,EAAM,EAAQ,EACrC,EAAI,GACF,GAAkB,GAA6B,GAAc,EACzD,EAAc,EAAM,CAAc,EAClC,IAAA,EACR,CACA,EAAkB,KAAK,CAAG,CAC5B,CACA,MAAO,CAAE,oBAAmB,iBAAgB,CAC9C,CAOA,SAAgB,EAAU,EAAe,EAAuB,CAC9D,IAAM,EAAa,EAAE,OAAS,KAAQ,EAChC,EAAgB,IAAI,IAC1B,IAAK,GAAM,CAAC,EAAO,KAAW,EAAE,QAAQ,EAAG,CACzC,IAAI,EAAO,EAAc,IAAI,CAAM,EAC9B,IACH,EAAO,IAAI,YAAY,CAAS,EAChC,EAAc,IAAI,EAAQ,CAAI,GAEhC,IAAM,EAAO,IAAU,EAIvB,EAAK,IAAS,EAAK,IAAS,GAAM,IAAM,EAAQ,GAClD,CAEA,IAAM,EAAI,IAAI,YAAY,CAAS,EACnC,IAAK,IAAM,KAAU,EAAG,CACtB,IAAM,EAAY,EAAc,IAAI,CAAM,EAEtC,EAAa,EACb,EAAS,EACb,IAAK,IAAI,EAAO,EAAG,EAAO,EAAW,GAAQ,EAAG,CAC9C,IAAM,EAAW,EAAE,IAAS,EAEtB,IAAM,IAAY,IAAS,GAAK,KAAc,EAE9C,GAAY,GAAY,EAAK,KAAgB,EACnD,EAAa,IAAa,GAC1B,IAAM,EAAa,EAAI,EAAU,EACjC,EAAS,IAAa,GAEtB,EAAE,GAAQ,EAAI,CAAC,CACjB,CACF,CAEA,IAAI,EAAS,EACb,IAAK,IAAM,KAAQ,EACjB,GAAU,EAAS,CAAI,EAEzB,OAAO,CACT,CAEA,SAAS,EAAS,EAAuB,CACvC,IAAI,EAAQ,GAAU,IAAU,EAAK,YAErC,MADA,IAAS,EAAQ,YAAmB,IAAU,EAAK,WAC3C,KAAK,KAAM,GAAS,IAAU,GAAM,UAAe,QAAa,IAAM,GAAM,GACtF,CAEA,SAAS,EACP,EACA,EACA,EACA,EACA,EACoB,CACpB,MAAO,CACL,cACA,WAAY,EAAgB,EAC5B,kBACA,gBACA,WAAY,EAAM,WAClB,SAAU,EAAK,SACf,UAAW,EAAM,UACjB,QAAS,EAAK,OAChB,CACF,CAGA,MAAM,EAAkC,CAAC,EACnC,EAAmC,CAAC,EAE1C,SAAS,EAAoB,EAAuB,CAClD,IAAI,EAAO,EAAsB,GAKjC,OAJI,IAAS,IAAA,KACX,EAAO,EAAS,IAAI,GAAO,EAC3B,EAAsB,GAAS,GAE1B,CACT,CAEA,SAAS,EAAqB,EAAuB,CACnD,IAAI,EAAO,EAAuB,GAKlC,OAJI,IAAS,IAAA,KACX,EAAO,EAAU,IAAI,GAAO,EAC5B,EAAuB,GAAS,GAE3B,CACT,CAUA,SAAS,EACP,EACA,EACA,EACA,EACQ,CAGR,GAAM,CAAC,EAAS,GAAa,EAAoB,EAAQ,EAAiB,EADrD,GADC,EAAmB,IAAkB,IAAM,EAAmB,IAAoB,GACtD,EAAgB,CACkC,CAAC,EACrG,MAAO,GAAG,EAAQ,GAAG,EAAU,GAAG,EAAgB,GACpD,CASA,SAAS,EAAgB,EAAiB,EAAyB,EAA+B,CAChG,GAAM,CAAC,EAAS,GAAa,EAAoB,EAAQ,EAAiB,EAAe,EAAK,EAE9F,OAAO,EAAU,KAAK,KAAK,EAAW,EAAE,CAC1C,CAEA,SAAS,EACP,EACA,EACA,EACA,EACkB,CAClB,IAAM,EAAoB,IAAI,IAC1B,EAAU,KACV,EAAY,MAChB,IAAK,IAAI,EAAQ,EAAiB,EAAQ,EAAe,GAAS,EAAG,CACnE,IAAM,EAAQ,EAAO,GACrB,GAAI,CAAC,EACH,SAIF,IAAI,EACA,EACJ,GAAI,EAAM,OAAS,KAAM,CACvB,IAAI,EAAkB,EAAkB,IAAI,EAAM,IAAI,EAClD,IAAoB,IAAA,KACtB,EAAkB,EAAkB,KACpC,EAAkB,IAAI,EAAM,KAAM,CAAe,GAEnD,EAAO,EAAoB,CAAe,EAC1C,EAAQ,EAAqB,CAAe,CAC9C,KACE,GAAO,EAAM,SACb,EAAQ,EAAM,UAGhB,EAAW,KAAK,KAAK,EAAS,EAAE,EAAI,EAAQ,EAC5C,EAAY,KAAK,KAAK,EAAW,EAAE,EAAI,EACnC,GAAqB,EAAM,cAAgB,IAAA,IAAa,EAAM,eAAiB,IAAA,KAEjF,EAAW,KAAK,KAAK,EAAS,EAAE,EAAI,EAAM,YAAe,EACzD,EAAY,KAAK,KAAK,EAAW,EAAE,EAAI,EAAM,aAEjD,CACA,MAAO,CAAC,EAAS,CAAS,CAC5B,CAGA,SAAS,EAAS,EAAsB,CACtC,IAAI,EAAO,KACX,IAAK,IAAI,EAAQ,EAAG,EAAQ,EAAK,OAAQ,GAAS,EAEhD,EAAO,KAAK,KAAK,EAAM,EAAE,EAAI,EAAK,WAAW,CAAK,EAEpD,OAAO,CACT,CAGA,SAAS,EAAU,EAAsB,CACvC,IAAI,EAAO,YACX,IAAK,IAAI,EAAQ,EAAG,EAAQ,EAAK,OAAQ,GAAS,EAEhD,EAAO,KAAK,KAAK,EAAO,EAAK,WAAW,CAAK,EAAG,QAAU,EAE5D,OAAO,CACT,CAEA,SAAS,EAAc,EAAc,EAAuB,CAC1D,OAAO,KAAK,KAAK,EAAM,EAAE,EAAI,CAC/B,CAmBA,SAAgB,EAAiD,EAAe,EAA6B,CAC3G,GAAI,GAAgB,GAAK,EAAO,OAAS,EACvC,OAAO,EAGT,EAAO,KAAK,CAAa,EACzB,IAAK,IAAI,EAAU,GAAM,GAAU,CACjC,EAAU,GACV,IAAK,IAAI,EAAY,EAAG,EAAY,EAAO,QAAU,CAAC,EAAS,GAAa,EAC1E,IAAK,IAAI,EAAa,EAAY,EAAG,EAAa,EAAO,OAAQ,GAAc,EAAG,CAChF,IAAM,EAAO,EAAO,GACd,EAAQ,EAAO,GACrB,GAAI,CAAC,GAAQ,CAAC,EACZ,SAEF,IAAM,EAAU,EAAY,EAAM,EAAO,CAAY,EAC/C,EAAS,GAAW,EAAY,EAAO,EAAM,CAAY,EAC/D,GAAI,CAAC,EACH,SAEF,IAAM,EAAe,EAAU,EAAO,cAAgB,EAAO,eACvD,EAAgB,EAAU,EAAO,eAAiB,EAAO,cAC3D,GAAgB,GAClB,EAAO,GAAa,EAAO,OAC3B,EAAO,OAAO,EAAY,CAAC,GAClB,EACT,EAAO,GAAc,EAAO,OAE5B,EAAO,GAAa,EAAO,OAM7B,IAAK,IAAM,KAAc,EAAO,eAC9B,EAAW,sBAAwB,GAErC,EAAO,KAAK,CAAa,EACzB,EAAU,GACV,KACF,CAEJ,CACA,OAAO,CACT,CAEA,SAAS,EAAc,EAA2B,EAAoC,CACpF,IAAM,EAAY,EAAK,GACjB,EAAa,EAAM,GACzB,OACG,GAAW,iBAAmB,IAAM,GAAY,iBAAmB,KACnE,GAAW,eAAiB,IAAM,GAAY,eAAiB,EAEpE,CAmBA,SAAS,EACP,EACA,EACA,EAC4B,CAM5B,IAAM,EAAW,EAAM,OAAQ,GAAe,CAAC,EAAW,qBAAqB,EACzE,EAAY,EAAO,OAAQ,GAAe,CAAC,EAAW,qBAAqB,EAC3E,EAAkB,CAAC,EACrB,EAAe,EACf,EAAsB,GAC1B,IAAK,IAAM,KAAY,EAAW,CAEhC,KAAO,EAAe,EAAS,QAAQ,CACrC,IAAM,EAAU,EAAS,GACzB,GAAI,GAAW,EAAQ,cAAgB,EAAe,EAAS,gBAC7D,GAAgB,OAEhB,KAEJ,CACA,IAAM,EAAU,EAAS,GAEvB,GACA,EAAQ,eAAiB,EAAS,iBAClC,EAAQ,iBAAmB,IAE3B,EAAM,KAAK,CAAC,EAAS,CAAQ,CAAC,EAC9B,EAAsB,EAAS,cAC/B,GAAgB,EAEpB,CACA,IAAM,EAAgB,EAAM,SAAW,EAAM,OACvC,EAAiB,EAAM,SAAW,EAAO,OAC3C,OAAM,OAAS,GAAM,CAAC,GAAiB,CAAC,GAe5C,MAAO,CAAE,OAZM,EAAM,KAAK,CAAC,EAAS,MAAe,CACjD,GAAG,EAEH,sBAAuB,IAAA,GACvB,SAAU,CAAC,GAAG,EAAQ,SAAU,GAAG,EAAS,QAAQ,EACpD,WAAY,EAAQ,WAAa,EAAS,WAC1C,cAAe,EAAS,cACxB,SAAU,EAAS,SACnB,QAAS,EAAS,OACpB,EAGc,EAAG,gBAAe,iBAAgB,eAD9C,IAAkB,EAAiB,CAAC,EAAI,EAAM,KAAK,CAAC,EAAS,KAAe,EAAgB,EAAW,CAAQ,CAClD,CACjE,CAcA,SAAgB,EAAwB,EAAoC,CAC1E,IAAI,EAAgB,EAChB,EAAmB,EACnB,EAAsB,GAC1B,IAAK,IAAM,KAAc,EAAO,CAC9B,GAAI,EAAW,sBAAuB,CACpC,EAAsB,GACtB,QACF,CACA,GAAiB,EAAW,SAAS,OACrC,EAAmB,KAAK,IAAI,EAAkB,EAAW,SAAS,MAAM,CAC1E,CACA,OAAO,EAAsB,EAAgB,EAAgB,CAC/D,CAGA,SAAgB,EACd,EACA,EACA,EACA,EACM,CACN,IAAK,IAAI,EAAQ,EAAQ,gBAAiB,EAAQ,EAAQ,cAAe,GAAS,EAAG,CACnF,IAAM,EAAQ,EAAO,GACrB,IAAK,IAAI,EAAM,GAAO,UAAY,EAAG,IAAQ,GAAO,QAAU,IAAK,GAAO,GACpE,CAAC,GAAmB,EAAgB,IAAI,EAAM,CAAC,IACjD,EAAgB,IAAI,EAAM,CAAC,CAGjC,CACF"}

package/dist/index.cjs

@@ -1 +1 @@
-"use strict";Object.defineProperty(exports,Symbol.toStringTag,{value:`Module`});const e=require("./crossFileDuplication.cjs"),t=require("./languages.cjs"),n=require("./nativeMetrics.cjs"),r=require("./metrics.cjs"),i=require("./regressionGate.cjs");exports.TreeMeasurer=r.TreeMeasurer,exports.collectCrossFileDuplicationFileData=r.collectCrossFileDuplicationFileData,exports.collectDuplicationCandidates=r.collectDuplicationCandidates,exports.collectFunctionTokenSequences=r.collectFunctionTokenSequences,exports.defaultGateOptions=i.defaultGateOptions,exports.defaultLanguages=t.defaultLanguages,exports.defaultMeasurer=r.defaultMeasurer,exports.evaluateRegressionGate=i.evaluateRegressionGate,exports.isNativeBackendAvailable=n.isNativeBackendAvailable,exports.measureCode=r.measureCode,exports.measureCrossFileDuplication=e.measureCrossFileDuplication,exports.supportedLanguages=t.supportedLanguages;
+"use strict";Object.defineProperty(exports,Symbol.toStringTag,{value:`Module`});const e=require("./crossFileDuplication.cjs"),t=require("./languages.cjs"),n=require("./metrics.cjs"),r=require("./regressionGate.cjs");exports.TreeMeasurer=n.TreeMeasurer,exports.collectCrossFileDuplicationFileData=n.collectCrossFileDuplicationFileData,exports.collectDuplicationCandidates=n.collectDuplicationCandidates,exports.collectFunctionTokenSequences=n.collectFunctionTokenSequences,exports.defaultGateOptions=r.defaultGateOptions,exports.defaultLanguages=t.defaultLanguages,exports.defaultMeasurer=n.defaultMeasurer,exports.evaluateRegressionGate=r.evaluateRegressionGate,exports.measureCode=n.measureCode,exports.measureCrossFileDuplication=e.measureCrossFileDuplication,exports.supportedLanguages=t.supportedLanguages;

package/dist/index.d.ts

@@ -3,7 +3,6 @@ export type { CrossFileDuplicateBlockGroup, CrossFileDuplicateOccurrence, CrossF
 export type { CrossFileDuplicateCandidate, CrossFileDuplicationFileData } from './duplication.js';
 export { defaultLanguages, supportedLanguages } from './languages.js';
 export { TreeMeasurer, collectCrossFileDuplicationFileData, collectDuplicationCandidates, collectFunctionTokenSequences, defaultMeasurer, measureCode, } from './metrics.js';
-export { isNativeBackendAvailable } from './nativeMetrics.js';
 export { defaultGateOptions, evaluateRegressionGate } from './regressionGate.js';
 export type { GateFileInput, GateOptions, GateResult, GateTolerances, GateViolation, NewFunctionThresholds, } from './regressionGate.js';
 export type { CodeMetrics, DuplicationMetrics, DuplicationOptions, FunctionMetrics, HalsteadMetrics, LanguageDefinition, LanguageName, LineMetrics, MeasureOptions, SupportedLanguage, } from './types.js';

package/dist/index.js

@@ -1 +1 @@
-import{measureCrossFileDuplication as e}from"./crossFileDuplication.js";import{defaultLanguages as t,supportedLanguages as n}from"./languages.js";import{isNativeBackendAvailable as r}from"./nativeMetrics.js";import{TreeMeasurer as i,collectCrossFileDuplicationFileData as a,collectDuplicationCandidates as o,collectFunctionTokenSequences as s,defaultMeasurer as c,measureCode as l}from"./metrics.js";import{defaultGateOptions as u,evaluateRegressionGate as d}from"./regressionGate.js";export{i as TreeMeasurer,a as collectCrossFileDuplicationFileData,o as collectDuplicationCandidates,s as collectFunctionTokenSequences,u as defaultGateOptions,t as defaultLanguages,c as defaultMeasurer,d as evaluateRegressionGate,r as isNativeBackendAvailable,l as measureCode,e as measureCrossFileDuplication,n as supportedLanguages};
+import{measureCrossFileDuplication as e}from"./crossFileDuplication.js";import{defaultLanguages as t,supportedLanguages as n}from"./languages.js";import{TreeMeasurer as r,collectCrossFileDuplicationFileData as i,collectDuplicationCandidates as a,collectFunctionTokenSequences as o,defaultMeasurer as s,measureCode as c}from"./metrics.js";import{defaultGateOptions as l,evaluateRegressionGate as u}from"./regressionGate.js";export{r as TreeMeasurer,i as collectCrossFileDuplicationFileData,a as collectDuplicationCandidates,o as collectFunctionTokenSequences,l as defaultGateOptions,t as defaultLanguages,s as defaultMeasurer,u as evaluateRegressionGate,c as measureCode,e as measureCrossFileDuplication,n as supportedLanguages};

package/dist/languages.cjs

@@ -1,2 +1,2 @@
-"use strict";const e=require("./_virtual/_rolldown/runtime.cjs");let t=require("tree-sitter-c");t=e.__toESM(t,1);let n=require("tree-sitter-cpp");n=e.__toESM(n,1);let r=require("tree-sitter-go");r=e.__toESM(r,1);let i=require("tree-sitter-java");i=e.__toESM(i,1);let a=require("tree-sitter-javascript");a=e.__toESM(a,1);let o=require("tree-sitter-python");o=e.__toESM(o,1);let s=require("tree-sitter-ruby");s=e.__toESM(s,1);let c=require("tree-sitter-rust");c=e.__toESM(c,1);let l=require("tree-sitter-typescript");l=e.__toESM(l,1);const u=[`function`,`function_declaration`,`function_definition`,`function_expression`,`function_item`,`function_signature_item`,`function_declarator`,`func_literal`,`method_declaration`,`method_definition`,`method_spec`,`arrow_function`,`generator_function`,`generator_function_declaration`,`lambda`,`lambda_expression`,`closure_expression`],d=[`if_statement`,`elif_clause`,`else_if_clause`,`for_statement`,`for_in_statement`,`while_statement`,`do_statement`,`catch_clause`,`except_clause`,`case_clause`,`switch_case`,`match_arm`,`conditional_expression`,`ternary_expression`,`if_expression`,`while_expression`,`for_expression`,`loop_expression`],f=[...d,`expression_case`,`type_case`,`communication_case`],p=[...d,`switch_default`],m=[...f,`default_case`],h=[...u,`constructor_declaration`,`compact_constructor_declaration`],g=[...d,`enhanced_for_statement`,`switch_block_statement_group`,`switch_rule`],_=[`method`,`singleton_method`,`lambda`,`block`,`do_block`],v=[`if`,`elsif`,`unless`,`while`,`until`,`for`,`when`,`in_clause`,`rescue`,`conditional`,`if_modifier`,`unless_modifier`,`while_modifier`,`until_modifier`,`rescue_modifier`],y=[`function_definition`,`lambda_expression`],b=[...d,`case_statement`],x=[...b,`for_range_loop`],S=`import_statement.export_statement.lexical_declaration.variable_declaration.function_declaration.function_signature.generator_function_declaration.class_declaration.abstract_class_declaration.module.method_definition.abstract_method_signature.class_static_block.field_definition.public_field_definition.type_alias_declaration.interface_declaration.enum_declaration.expression_statement.if_statement.else_clause.switch_statement.switch_case.switch_default.for_statement.for_in_statement.while_statement.do_statement.catch_clause.finally_clause.labeled_statement.return_statement.break_statement.continue_statement.throw_statement.debugger_statement.with_statement`.split(`.`),C=`import_statement.import_from_statement.future_import_statement.print_statement.exec_statement.assert_statement.expression_statement.return_statement.delete_statement.raise_statement.pass_statement.break_statement.continue_statement.global_statement.nonlocal_statement.if_statement.elif_clause.else_clause.for_statement.while_statement.except_clause.except_group_clause.finally_clause.with_statement.match_statement.case_clause.function_definition.class_definition.type_alias_statement`.split(`.`),w=`package_clause.import_spec.type_spec.type_alias.const_spec.var_spec.function_declaration.method_declaration.short_var_declaration.expression_statement.send_statement.inc_statement.dec_statement.assignment_statement.if_statement.for_statement.expression_switch_statement.type_switch_statement.select_statement.expression_case.type_case.communication_case.default_case.return_statement.break_statement.continue_statement.goto_statement.fallthrough_statement.defer_statement.go_statement.labeled_statement`.split(`.`),T=[`use_declaration`,`extern_crate_declaration`,`foreign_mod_item`,`mod_item`,`const_item`,`static_item`,`struct_item`,`enum_item`,`union_item`,`trait_item`,`impl_item`,`function_item`,`function_signature_item`,`type_item`,`associated_type`,`macro_definition`,`field_declaration`,`let_declaration`,`expression_statement`,`else_clause`,`match_arm`],E=`package_declaration.import_declaration.module_declaration.requires_module_directive.exports_module_directive.opens_module_directive.uses_module_directive.provides_module_directive.class_declaration.interface_declaration.enum_declaration.annotation_type_declaration.annotation_type_element_declaration.record_declaration.field_declaration.constant_declaration.method_declaration.constructor_declaration.compact_constructor_declaration.static_initializer.explicit_constructor_invocation.local_variable_declaration.expression_statement.if_statement.while_statement.do_statement.for_statement.enhanced_for_statement.switch_statement.switch_expression.switch_label.break_statement.continue_statement.return_statement.throw_statement.assert_statement.synchronized_statement.labeled_statement.yield_statement.resource.catch_clause.finally_clause`.split(`.`),D=[`elsif`,`else`,`when`,`in_clause`,`rescue`,`ensure`],O=[`program`,`body_statement`,`then`,`else`,`do`,`block_body`,`begin`,`ensure`],k=[`preproc_include`,`preproc_def`,`preproc_function_def`,`declaration`,`type_definition`,`field_declaration`,`function_definition`,`struct_specifier`,`enum_specifier`,`union_specifier`,`expression_statement`,`if_statement`,`else_clause`,`switch_statement`,`case_statement`,`for_statement`,`while_statement`,`do_statement`,`return_statement`,`break_statement`,`continue_statement`,`goto_statement`,`labeled_statement`],A=[...k,`class_specifier`,`namespace_definition`,`using_declaration`,`alias_declaration`,`namespace_alias_definition`,`concept_definition`,`static_assert_declaration`,`for_range_loop`,`catch_clause`,`throw_statement`,`co_return_statement`,`co_yield_statement`];function j(e){return N(e,`default`)?e.default:e}function M(e){return l.default[e]}function N(e,t){return typeof e!=`object`||!e||!(t in e)?!1:!!e[t]}const P=[{name:`javascript`,aliases:[`js`,`mjs`,`cjs`],parserLanguage:j(a.default)},{name:`jsx`,parserLanguage:j(a.default)},{name:`typescript`,aliases:[`ts`],parserLanguage:M(`typescript`)},{name:`tsx`,parserLanguage:M(`tsx`)},{name:`python`,aliases:[`py`],parserLanguage:j(o.default),ncssNodeTypes:C},{name:`go`,parserLanguage:j(r.default),decisionNodeTypes:f,nestingNodeTypes:m,ncssNodeTypes:w},{name:`rust`,aliases:[`rs`],parserLanguage:j(c.default),ncssNodeTypes:T,ncssContainerNodeTypes:[`block`,`else_clause`]},{name:`java`,parserLanguage:j(i.default),functionNodeTypes:h,decisionNodeTypes:g,nestingNodeTypes:g,ncssNodeTypes:E},{name:`ruby`,aliases:[`rb`],parserLanguage:j(s.default),functionNodeTypes:_,decisionNodeTypes:v,nestingNodeTypes:v,ncssNodeTypes:D,ncssContainerNodeTypes:O},{name:`c`,parserLanguage:j(t.default),functionNodeTypes:y,decisionNodeTypes:b,nestingNodeTypes:b,ncssNodeTypes:k},{name:`cpp`,aliases:[`c++`,`cxx`],parserLanguage:j(n.default),functionNodeTypes:y,decisionNodeTypes:x,nestingNodeTypes:x,ncssNodeTypes:A}].map(e=>({functionNodeTypes:u,decisionNodeTypes:d,nestingNodeTypes:p,ncssNodeTypes:S,...e}));function F(e=P){let t=new Map;for(let n of e){t.set(n.name,n);for(let e of n.aliases??[])t.set(e,n)}return t}const I=P.map(e=>e.name);exports.createLanguageRegistry=F,exports.defaultLanguages=P,exports.supportedLanguages=I;
+"use strict";const e=[{name:`javascript`,aliases:[`js`,`mjs`,`cjs`]},{name:`jsx`},{name:`typescript`,aliases:[`ts`]},{name:`tsx`},{name:`python`,aliases:[`py`]},{name:`go`},{name:`rust`,aliases:[`rs`]},{name:`java`},{name:`ruby`,aliases:[`rb`]},{name:`c`},{name:`cpp`,aliases:[`c++`,`cxx`]}];function t(t=e){let n=new Map;for(let e of t){n.set(e.name,e);for(let t of e.aliases??[])n.set(t,e)}return n}const n=e.map(e=>e.name);exports.createLanguageRegistry=t,exports.defaultLanguages=e,exports.supportedLanguages=n;
 //# sourceMappingURL=languages.cjs.map

package/dist/languages.cjs.map

@@ -1 +1 @@
-{"version":3,"file":"languages.cjs","names":["grammars","JavaScript","Python","Go","Rust","Java","Ruby","C","Cpp"],"sources":["../src/languages.ts"],"sourcesContent":["import C from 'tree-sitter-c';\nimport Cpp from 'tree-sitter-cpp';\nimport Go from 'tree-sitter-go';\nimport Java from 'tree-sitter-java';\nimport JavaScript from 'tree-sitter-javascript';\nimport Python from 'tree-sitter-python';\nimport Ruby from 'tree-sitter-ruby';\nimport Rust from 'tree-sitter-rust';\nimport TypeScript from 'tree-sitter-typescript';\nimport type { LanguageDefinition, LanguageName, ParserLanguage } from './types.js';\n\ntype GrammarModule = unknown;\n\nconst commonFunctionNodes = [\n  'function',\n  'function_declaration',\n  'function_definition',\n  'function_expression',\n  'function_item',\n  'function_signature_item',\n  'function_declarator',\n  'func_literal',\n  'method_declaration',\n  'method_definition',\n  'method_spec',\n  'arrow_function',\n  'generator_function',\n  'generator_function_declaration',\n  'lambda',\n  'lambda_expression',\n  'closure_expression',\n] as const;\n\nconst commonDecisionNodes = [\n  'if_statement',\n  'elif_clause',\n  'else_if_clause',\n  'for_statement',\n  'for_in_statement',\n  'while_statement',\n  'do_statement',\n  'catch_clause',\n  'except_clause',\n  'case_clause',\n  'switch_case',\n  'match_arm',\n  'conditional_expression',\n  'ternary_expression',\n  'if_expression',\n  'while_expression',\n  'for_expression',\n  'loop_expression',\n] as const;\n\n// Go `switch`/`select` branches are `*_case` nodes, not the `case_clause`/`switch_case` of other grammars.\nconst goDecisionNodes = [...commonDecisionNodes, 'expression_case', 'type_case', 'communication_case'] as const;\n\n// Default switch branches add no decision, but their contents are nested inside the switch like\n// any other arm, so they appear in the nesting sets only.\nconst commonNestingNodes = [...commonDecisionNodes, 'switch_default'] as const;\nconst goNestingNodes = [...goDecisionNodes, 'default_case'] as const;\n\nconst javaFunctionNodes = [\n  ...commonFunctionNodes,\n  'constructor_declaration',\n  'compact_constructor_declaration',\n] as const;\nconst javaDecisionNodes = [\n  ...commonDecisionNodes,\n  'enhanced_for_statement',\n  'switch_block_statement_group',\n  'switch_rule',\n] as const;\n\n// Ruby node types are keyword-like (`if`, `while`, ...), so they must stay Ruby-specific: the same\n// strings appear as anonymous keyword tokens in other grammars and would be double-counted there.\n// `block`/`do_block` are Ruby's closures (`items.map { ... }`), the analog of JS callbacks.\nconst rubyFunctionNodes = ['method', 'singleton_method', 'lambda', 'block', 'do_block'] as const;\nconst rubyDecisionNodes = [\n  'if',\n  'elsif',\n  'unless',\n  'while',\n  'until',\n  'for',\n  'when',\n  'in_clause',\n  'rescue',\n  'conditional',\n  'if_modifier',\n  'unless_modifier',\n  'while_modifier',\n  'until_modifier',\n  'rescue_modifier',\n] as const;\n\n// `function_declarator` must stay out: it is nested inside every `function_definition` (which\n// would double-count) and also appears in body-less prototypes.\nconst cFunctionNodes = ['function_definition', 'lambda_expression'] as const;\nconst cDecisionNodes = [...commonDecisionNodes, 'case_statement'] as const;\nconst cppDecisionNodes = [...cDecisionNodes, 'for_range_loop'] as const;\n\n// NCSS node sets: every listed type counts as one non-commenting source statement. The Java set is\n// calibrated against PMD's NcssCount rule (`try` counts 0; `else`, `case`/`default` labels,\n// `catch`, `finally`, and try-with-resources resources count 1 each); the other languages follow\n// the same conventions with their grammar's node types.\nconst jsNcssNodes = [\n  'import_statement',\n  'export_statement',\n  'lexical_declaration',\n  'variable_declaration',\n  'function_declaration',\n  'function_signature',\n  'generator_function_declaration',\n  'class_declaration',\n  'abstract_class_declaration',\n  'module',\n  'method_definition',\n  'abstract_method_signature',\n  'class_static_block',\n  'field_definition',\n  'public_field_definition',\n  'type_alias_declaration',\n  'interface_declaration',\n  'enum_declaration',\n  'expression_statement',\n  'if_statement',\n  'else_clause',\n  'switch_statement',\n  'switch_case',\n  'switch_default',\n  'for_statement',\n  'for_in_statement',\n  'while_statement',\n  'do_statement',\n  'catch_clause',\n  'finally_clause',\n  'labeled_statement',\n  'return_statement',\n  'break_statement',\n  'continue_statement',\n  'throw_statement',\n  'debugger_statement',\n  'with_statement',\n] as const;\n\nconst pythonNcssNodes = [\n  'import_statement',\n  'import_from_statement',\n  'future_import_statement',\n  'print_statement',\n  'exec_statement',\n  'assert_statement',\n  'expression_statement',\n  'return_statement',\n  'delete_statement',\n  'raise_statement',\n  'pass_statement',\n  'break_statement',\n  'continue_statement',\n  'global_statement',\n  'nonlocal_statement',\n  'if_statement',\n  'elif_clause',\n  'else_clause',\n  'for_statement',\n  'while_statement',\n  'except_clause',\n  'except_group_clause',\n  'finally_clause',\n  'with_statement',\n  'match_statement',\n  'case_clause',\n  'function_definition',\n  'class_definition',\n  'type_alias_statement',\n] as const;\n\nconst goNcssNodes = [\n  'package_clause',\n  'import_spec',\n  'type_spec',\n  'type_alias',\n  'const_spec',\n  'var_spec',\n  'function_declaration',\n  'method_declaration',\n  // Struct fields and interface members count contextually (see ncss.ts): only inside a named\n  // `type T struct/interface { ... }`, not in inline anonymous types, which are part of one\n  // declaration.\n  'short_var_declaration',\n  'expression_statement',\n  'send_statement',\n  'inc_statement',\n  'dec_statement',\n  'assignment_statement',\n  'if_statement',\n  'for_statement',\n  'expression_switch_statement',\n  'type_switch_statement',\n  'select_statement',\n  'expression_case',\n  'type_case',\n  'communication_case',\n  'default_case',\n  'return_statement',\n  'break_statement',\n  'continue_statement',\n  'goto_statement',\n  'fallthrough_statement',\n  'defer_statement',\n  'go_statement',\n  'labeled_statement',\n] as const;\n\nconst rustNcssNodes = [\n  'use_declaration',\n  'extern_crate_declaration',\n  'foreign_mod_item',\n  'mod_item',\n  'const_item',\n  'static_item',\n  'struct_item',\n  'enum_item',\n  'union_item',\n  'trait_item',\n  'impl_item',\n  'function_item',\n  'function_signature_item',\n  'type_item',\n  'associated_type',\n  'macro_definition',\n  'field_declaration',\n  'let_declaration',\n  'expression_statement',\n  'else_clause',\n  'match_arm',\n] as const;\n\nconst javaNcssNodes = [\n  'package_declaration',\n  'import_declaration',\n  'module_declaration',\n  'requires_module_directive',\n  'exports_module_directive',\n  'opens_module_directive',\n  'uses_module_directive',\n  'provides_module_directive',\n  'class_declaration',\n  'interface_declaration',\n  'enum_declaration',\n  'annotation_type_declaration',\n  'annotation_type_element_declaration',\n  'record_declaration',\n  'field_declaration',\n  'constant_declaration',\n  'method_declaration',\n  'constructor_declaration',\n  'compact_constructor_declaration',\n  'static_initializer',\n  'explicit_constructor_invocation',\n  'local_variable_declaration',\n  'expression_statement',\n  'if_statement',\n  'while_statement',\n  'do_statement',\n  'for_statement',\n  'enhanced_for_statement',\n  'switch_statement',\n  'switch_expression',\n  'switch_label',\n  'break_statement',\n  'continue_statement',\n  'return_statement',\n  'throw_statement',\n  'assert_statement',\n  'synchronized_statement',\n  'labeled_statement',\n  'yield_statement',\n  'resource',\n  'catch_clause',\n  'finally_clause',\n] as const;\n\n// Ruby statements have no wrapper node types, so bodies are counted positionally (see\n// ncssContainerNodeTypes); only clause nodes hanging off non-container parents need listing.\nconst rubyNcssNodes = ['elsif', 'else', 'when', 'in_clause', 'rescue', 'ensure'] as const;\nconst rubyNcssContainers = [\n  'program',\n  'body_statement',\n  'then',\n  'else',\n  'do',\n  'block_body',\n  'begin',\n  'ensure',\n] as const;\n\nconst cNcssNodes = [\n  'preproc_include',\n  'preproc_def',\n  'preproc_function_def',\n  'declaration',\n  'type_definition',\n  'field_declaration',\n  'function_definition',\n  // Counted only when they carry a body (see bodylessNcssSpecifierTypes in ncss.ts).\n  'struct_specifier',\n  'enum_specifier',\n  'union_specifier',\n  'expression_statement',\n  'if_statement',\n  'else_clause',\n  'switch_statement',\n  'case_statement',\n  'for_statement',\n  'while_statement',\n  'do_statement',\n  'return_statement',\n  'break_statement',\n  'continue_statement',\n  'goto_statement',\n  'labeled_statement',\n] as const;\nconst cppNcssNodes = [\n  ...cNcssNodes,\n  'class_specifier',\n  'namespace_definition',\n  'using_declaration',\n  'alias_declaration',\n  'namespace_alias_definition',\n  'concept_definition',\n  'static_assert_declaration',\n  'for_range_loop',\n  'catch_clause',\n  'throw_statement',\n  'co_return_statement',\n  'co_yield_statement',\n] as const;\n\nfunction normalizeGrammar(module: GrammarModule): ParserLanguage {\n  if (isGrammarWrapper(module, 'default')) {\n    return module.default;\n  }\n\n  return module;\n}\n\nfunction getTypeScriptGrammar(name: 'typescript' | 'tsx'): ParserLanguage {\n  const grammars = TypeScript as unknown as Record<string, GrammarModule>;\n  return grammars[name];\n}\n\nfunction isGrammarWrapper(value: GrammarModule, key: 'default'): value is Record<typeof key, ParserLanguage> {\n  if (typeof value !== 'object' || value === null || !(key in value)) {\n    return false;\n  }\n\n  return Boolean((value as Record<string, ParserLanguage>)[key]);\n}\n\nexport const defaultLanguages: readonly LanguageDefinition[] = [\n  {\n    name: 'javascript',\n    aliases: ['js', 'mjs', 'cjs'],\n    parserLanguage: normalizeGrammar(JavaScript as unknown as GrammarModule),\n  },\n  {\n    name: 'jsx',\n    parserLanguage: normalizeGrammar(JavaScript as unknown as GrammarModule),\n  },\n  {\n    name: 'typescript',\n    aliases: ['ts'],\n    parserLanguage: getTypeScriptGrammar('typescript'),\n  },\n  {\n    name: 'tsx',\n    parserLanguage: getTypeScriptGrammar('tsx'),\n  },\n  {\n    name: 'python',\n    aliases: ['py'],\n    parserLanguage: normalizeGrammar(Python as unknown as GrammarModule),\n    ncssNodeTypes: pythonNcssNodes,\n  },\n  {\n    name: 'go',\n    parserLanguage: normalizeGrammar(Go as unknown as GrammarModule),\n    decisionNodeTypes: goDecisionNodes,\n    nestingNodeTypes: goNestingNodes,\n    ncssNodeTypes: goNcssNodes,\n  },\n  {\n    name: 'rust',\n    aliases: ['rs'],\n    parserLanguage: normalizeGrammar(Rust as unknown as GrammarModule),\n    ncssNodeTypes: rustNcssNodes,\n    // `else_clause` is a container so `else if` chains count the nested if_expression; a plain\n    // `else { ... }` is unaffected because its only child is a `block`, itself a container.\n    ncssContainerNodeTypes: ['block', 'else_clause'],\n  },\n  {\n    name: 'java',\n    parserLanguage: normalizeGrammar(Java as unknown as GrammarModule),\n    functionNodeTypes: javaFunctionNodes,\n    decisionNodeTypes: javaDecisionNodes,\n    nestingNodeTypes: javaDecisionNodes,\n    ncssNodeTypes: javaNcssNodes,\n  },\n  {\n    name: 'ruby',\n    aliases: ['rb'],\n    parserLanguage: normalizeGrammar(Ruby as unknown as GrammarModule),\n    functionNodeTypes: rubyFunctionNodes,\n    decisionNodeTypes: rubyDecisionNodes,\n    nestingNodeTypes: rubyDecisionNodes,\n    ncssNodeTypes: rubyNcssNodes,\n    ncssContainerNodeTypes: rubyNcssContainers,\n  },\n  {\n    name: 'c',\n    parserLanguage: normalizeGrammar(C as unknown as GrammarModule),\n    functionNodeTypes: cFunctionNodes,\n    decisionNodeTypes: cDecisionNodes,\n    nestingNodeTypes: cDecisionNodes,\n    ncssNodeTypes: cNcssNodes,\n  },\n  {\n    name: 'cpp',\n    aliases: ['c++', 'cxx'],\n    parserLanguage: normalizeGrammar(Cpp as unknown as GrammarModule),\n    functionNodeTypes: cFunctionNodes,\n    decisionNodeTypes: cppDecisionNodes,\n    nestingNodeTypes: cppDecisionNodes,\n    ncssNodeTypes: cppNcssNodes,\n  },\n].map((language) => ({\n  functionNodeTypes: commonFunctionNodes,\n  decisionNodeTypes: commonDecisionNodes,\n  nestingNodeTypes: commonNestingNodes,\n  ncssNodeTypes: jsNcssNodes,\n  ...language,\n}));\n\nexport function createLanguageRegistry(\n  languages: readonly LanguageDefinition[] = defaultLanguages\n): Map<LanguageName, LanguageDefinition> {\n  const registry = new Map<LanguageName, LanguageDefinition>();\n\n  for (const language of languages) {\n    registry.set(language.name, language);\n    for (const alias of language.aliases ?? []) {\n      registry.set(alias, language);\n    }\n  }\n\n  return registry;\n}\n\nexport const supportedLanguages = defaultLanguages.map((language) => language.name);\n"],"mappings":"ohBAaA,MAAM,EAAsB,CAC1B,WACA,uBACA,sBACA,sBACA,gBACA,0BACA,sBACA,eACA,qBACA,oBACA,cACA,iBACA,qBACA,iCACA,SACA,oBACA,oBACF,EAEM,EAAsB,CAC1B,eACA,cACA,iBACA,gBACA,mBACA,kBACA,eACA,eACA,gBACA,cACA,cACA,YACA,yBACA,qBACA,gBACA,mBACA,iBACA,iBACF,EAGM,EAAkB,CAAC,GAAG,EAAqB,kBAAmB,YAAa,oBAAoB,EAI/F,EAAqB,CAAC,GAAG,EAAqB,gBAAgB,EAC9D,EAAiB,CAAC,GAAG,EAAiB,cAAc,EAEpD,EAAoB,CACxB,GAAG,EACH,0BACA,iCACF,EACM,EAAoB,CACxB,GAAG,EACH,yBACA,+BACA,aACF,EAKM,EAAoB,CAAC,SAAU,mBAAoB,SAAU,QAAS,UAAU,EAChF,EAAoB,CACxB,KACA,QACA,SACA,QACA,QACA,MACA,OACA,YACA,SACA,cACA,cACA,kBACA,iBACA,iBACA,iBACF,EAIM,EAAiB,CAAC,sBAAuB,mBAAmB,EAC5D,EAAiB,CAAC,GAAG,EAAqB,gBAAgB,EAC1D,EAAmB,CAAC,GAAG,EAAgB,gBAAgB,EAMvD,EAAc,iqBAsCpB,EAEM,EAAkB,gfA8BxB,EAEM,EAAc,kgBAmCpB,EAEM,EAAgB,CACpB,kBACA,2BACA,mBACA,WACA,aACA,cACA,cACA,YACA,aACA,aACA,YACA,gBACA,0BACA,YACA,kBACA,mBACA,oBACA,kBACA,uBACA,cACA,WACF,EAEM,EAAgB,k1BA2CtB,EAIM,EAAgB,CAAC,QAAS,OAAQ,OAAQ,YAAa,SAAU,QAAQ,EACzE,EAAqB,CACzB,UACA,iBACA,OACA,OACA,KACA,aACA,QACA,QACF,EAEM,EAAa,CACjB,kBACA,cACA,uBACA,cACA,kBACA,oBACA,sBAEA,mBACA,iBACA,kBACA,uBACA,eACA,cACA,mBACA,iBACA,gBACA,kBACA,eACA,mBACA,kBACA,qBACA,iBACA,mBACF,EACM,EAAe,CACnB,GAAG,EACH,kBACA,uBACA,oBACA,oBACA,6BACA,qBACA,4BACA,iBACA,eACA,kBACA,sBACA,oBACF,EAEA,SAAS,EAAiB,EAAuC,CAK/D,OAJI,EAAiB,EAAQ,SAAS,EAC7B,EAAO,QAGT,CACT,CAEA,SAAS,EAAqB,EAA4C,CAExE,OAAOA,EAAAA,QAAS,EAClB,CAEA,SAAS,EAAiB,EAAsB,EAA6D,CAK3G,OAJI,OAAO,GAAU,WAAY,GAAkB,EAAE,KAAO,GACnD,GAGF,EAAS,EAAyC,EAC3D,CAEA,MAAa,EAAkD,CAC7D,CACE,KAAM,aACN,QAAS,CAAC,KAAM,MAAO,KAAK,EAC5B,eAAgB,EAAiBC,EAAAA,OAAsC,CACzE,EACA,CACE,KAAM,MACN,eAAgB,EAAiBA,EAAAA,OAAsC,CACzE,EACA,CACE,KAAM,aACN,QAAS,CAAC,IAAI,EACd,eAAgB,EAAqB,YAAY,CACnD,EACA,CACE,KAAM,MACN,eAAgB,EAAqB,KAAK,CAC5C,EACA,CACE,KAAM,SACN,QAAS,CAAC,IAAI,EACd,eAAgB,EAAiBC,EAAAA,OAAkC,EACnE,cAAe,CACjB,EACA,CACE,KAAM,KACN,eAAgB,EAAiBC,EAAAA,OAA8B,EAC/D,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,CACjB,EACA,CACE,KAAM,OACN,QAAS,CAAC,IAAI,EACd,eAAgB,EAAiBC,EAAAA,OAAgC,EACjE,cAAe,EAGf,uBAAwB,CAAC,QAAS,aAAa,CACjD,EACA,CACE,KAAM,OACN,eAAgB,EAAiBC,EAAAA,OAAgC,EACjE,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,CACjB,EACA,CACE,KAAM,OACN,QAAS,CAAC,IAAI,EACd,eAAgB,EAAiBC,EAAAA,OAAgC,EACjE,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,EACf,uBAAwB,CAC1B,EACA,CACE,KAAM,IACN,eAAgB,EAAiBC,EAAAA,OAA6B,EAC9D,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,CACjB,EACA,CACE,KAAM,MACN,QAAS,CAAC,MAAO,KAAK,EACtB,eAAgB,EAAiBC,EAAAA,OAA+B,EAChE,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,CACjB,CACF,CAAC,CAAC,IAAK,IAAc,CACnB,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,EACf,GAAG,CACL,EAAE,EAEF,SAAgB,EACd,EAA2C,EACJ,CACvC,IAAM,EAAW,IAAI,IAErB,IAAK,IAAM,KAAY,EAAW,CAChC,EAAS,IAAI,EAAS,KAAM,CAAQ,EACpC,IAAK,IAAM,KAAS,EAAS,SAAW,CAAC,EACvC,EAAS,IAAI,EAAO,CAAQ,CAEhC,CAEA,OAAO,CACT,CAEA,MAAa,EAAqB,EAAiB,IAAK,GAAa,EAAS,IAAI"}
+{"version":3,"file":"languages.cjs","names":[],"sources":["../src/languages.ts"],"sourcesContent":["import type { LanguageDefinition, LanguageName } from './types.js';\n\n/**\n * The built-in languages. Grammars and per-language node-type configuration live in the Rust\n * addon (native/src/languages.rs); this list only names the languages and their aliases so the\n * CLI and API can resolve and enumerate them without crossing the N-API boundary.\n */\nexport const defaultLanguages: readonly LanguageDefinition[] = [\n  { name: 'javascript', aliases: ['js', 'mjs', 'cjs'] },\n  { name: 'jsx' },\n  { name: 'typescript', aliases: ['ts'] },\n  { name: 'tsx' },\n  { name: 'python', aliases: ['py'] },\n  { name: 'go' },\n  { name: 'rust', aliases: ['rs'] },\n  { name: 'java' },\n  { name: 'ruby', aliases: ['rb'] },\n  { name: 'c' },\n  { name: 'cpp', aliases: ['c++', 'cxx'] },\n];\n\nexport function createLanguageRegistry(\n  languages: readonly LanguageDefinition[] = defaultLanguages\n): Map<LanguageName, LanguageDefinition> {\n  const registry = new Map<LanguageName, LanguageDefinition>();\n\n  for (const language of languages) {\n    registry.set(language.name, language);\n    for (const alias of language.aliases ?? []) {\n      registry.set(alias, language);\n    }\n  }\n\n  return registry;\n}\n\nexport const supportedLanguages = defaultLanguages.map((language) => language.name);\n"],"mappings":"aAOA,MAAa,EAAkD,CAC7D,CAAE,KAAM,aAAc,QAAS,CAAC,KAAM,MAAO,KAAK,CAAE,EACpD,CAAE,KAAM,KAAM,EACd,CAAE,KAAM,aAAc,QAAS,CAAC,IAAI,CAAE,EACtC,CAAE,KAAM,KAAM,EACd,CAAE,KAAM,SAAU,QAAS,CAAC,IAAI,CAAE,EAClC,CAAE,KAAM,IAAK,EACb,CAAE,KAAM,OAAQ,QAAS,CAAC,IAAI,CAAE,EAChC,CAAE,KAAM,MAAO,EACf,CAAE,KAAM,OAAQ,QAAS,CAAC,IAAI,CAAE,EAChC,CAAE,KAAM,GAAI,EACZ,CAAE,KAAM,MAAO,QAAS,CAAC,MAAO,KAAK,CAAE,CACzC,EAEA,SAAgB,EACd,EAA2C,EACJ,CACvC,IAAM,EAAW,IAAI,IAErB,IAAK,IAAM,KAAY,EAAW,CAChC,EAAS,IAAI,EAAS,KAAM,CAAQ,EACpC,IAAK,IAAM,KAAS,EAAS,SAAW,CAAC,EACvC,EAAS,IAAI,EAAO,CAAQ,CAEhC,CAEA,OAAO,CACT,CAEA,MAAa,EAAqB,EAAiB,IAAK,GAAa,EAAS,IAAI"}

package/dist/languages.d.ts

@@ -1,4 +1,9 @@
 import type { LanguageDefinition, LanguageName } from './types.js';
+/**
+ * The built-in languages. Grammars and per-language node-type configuration live in the Rust
+ * addon (native/src/languages.rs); this list only names the languages and their aliases so the
+ * CLI and API can resolve and enumerate them without crossing the N-API boundary.
+ */
 export declare const defaultLanguages: readonly LanguageDefinition[];
 export declare function createLanguageRegistry(languages?: readonly LanguageDefinition[]): Map<LanguageName, LanguageDefinition>;
 export declare const supportedLanguages: LanguageName[];

package/dist/languages.js

@@ -1,2 +1,2 @@
-import e from"tree-sitter-c";import t from"tree-sitter-cpp";import n from"tree-sitter-go";import r from"tree-sitter-java";import i from"tree-sitter-javascript";import a from"tree-sitter-python";import o from"tree-sitter-ruby";import s from"tree-sitter-rust";import c from"tree-sitter-typescript";const l=[`function`,`function_declaration`,`function_definition`,`function_expression`,`function_item`,`function_signature_item`,`function_declarator`,`func_literal`,`method_declaration`,`method_definition`,`method_spec`,`arrow_function`,`generator_function`,`generator_function_declaration`,`lambda`,`lambda_expression`,`closure_expression`],u=[`if_statement`,`elif_clause`,`else_if_clause`,`for_statement`,`for_in_statement`,`while_statement`,`do_statement`,`catch_clause`,`except_clause`,`case_clause`,`switch_case`,`match_arm`,`conditional_expression`,`ternary_expression`,`if_expression`,`while_expression`,`for_expression`,`loop_expression`],d=[...u,`expression_case`,`type_case`,`communication_case`],f=[...u,`switch_default`],p=[...d,`default_case`],m=[...l,`constructor_declaration`,`compact_constructor_declaration`],h=[...u,`enhanced_for_statement`,`switch_block_statement_group`,`switch_rule`],g=[`method`,`singleton_method`,`lambda`,`block`,`do_block`],_=[`if`,`elsif`,`unless`,`while`,`until`,`for`,`when`,`in_clause`,`rescue`,`conditional`,`if_modifier`,`unless_modifier`,`while_modifier`,`until_modifier`,`rescue_modifier`],v=[`function_definition`,`lambda_expression`],y=[...u,`case_statement`],b=[...y,`for_range_loop`],x=`import_statement.export_statement.lexical_declaration.variable_declaration.function_declaration.function_signature.generator_function_declaration.class_declaration.abstract_class_declaration.module.method_definition.abstract_method_signature.class_static_block.field_definition.public_field_definition.type_alias_declaration.interface_declaration.enum_declaration.expression_statement.if_statement.else_clause.switch_statement.switch_case.switch_default.for_statement.for_in_statement.while_statement.do_statement.catch_clause.finally_clause.labeled_statement.return_statement.break_statement.continue_statement.throw_statement.debugger_statement.with_statement`.split(`.`),S=`import_statement.import_from_statement.future_import_statement.print_statement.exec_statement.assert_statement.expression_statement.return_statement.delete_statement.raise_statement.pass_statement.break_statement.continue_statement.global_statement.nonlocal_statement.if_statement.elif_clause.else_clause.for_statement.while_statement.except_clause.except_group_clause.finally_clause.with_statement.match_statement.case_clause.function_definition.class_definition.type_alias_statement`.split(`.`),C=`package_clause.import_spec.type_spec.type_alias.const_spec.var_spec.function_declaration.method_declaration.short_var_declaration.expression_statement.send_statement.inc_statement.dec_statement.assignment_statement.if_statement.for_statement.expression_switch_statement.type_switch_statement.select_statement.expression_case.type_case.communication_case.default_case.return_statement.break_statement.continue_statement.goto_statement.fallthrough_statement.defer_statement.go_statement.labeled_statement`.split(`.`),w=[`use_declaration`,`extern_crate_declaration`,`foreign_mod_item`,`mod_item`,`const_item`,`static_item`,`struct_item`,`enum_item`,`union_item`,`trait_item`,`impl_item`,`function_item`,`function_signature_item`,`type_item`,`associated_type`,`macro_definition`,`field_declaration`,`let_declaration`,`expression_statement`,`else_clause`,`match_arm`],T=`package_declaration.import_declaration.module_declaration.requires_module_directive.exports_module_directive.opens_module_directive.uses_module_directive.provides_module_directive.class_declaration.interface_declaration.enum_declaration.annotation_type_declaration.annotation_type_element_declaration.record_declaration.field_declaration.constant_declaration.method_declaration.constructor_declaration.compact_constructor_declaration.static_initializer.explicit_constructor_invocation.local_variable_declaration.expression_statement.if_statement.while_statement.do_statement.for_statement.enhanced_for_statement.switch_statement.switch_expression.switch_label.break_statement.continue_statement.return_statement.throw_statement.assert_statement.synchronized_statement.labeled_statement.yield_statement.resource.catch_clause.finally_clause`.split(`.`),E=[`elsif`,`else`,`when`,`in_clause`,`rescue`,`ensure`],D=[`program`,`body_statement`,`then`,`else`,`do`,`block_body`,`begin`,`ensure`],O=[`preproc_include`,`preproc_def`,`preproc_function_def`,`declaration`,`type_definition`,`field_declaration`,`function_definition`,`struct_specifier`,`enum_specifier`,`union_specifier`,`expression_statement`,`if_statement`,`else_clause`,`switch_statement`,`case_statement`,`for_statement`,`while_statement`,`do_statement`,`return_statement`,`break_statement`,`continue_statement`,`goto_statement`,`labeled_statement`],k=[...O,`class_specifier`,`namespace_definition`,`using_declaration`,`alias_declaration`,`namespace_alias_definition`,`concept_definition`,`static_assert_declaration`,`for_range_loop`,`catch_clause`,`throw_statement`,`co_return_statement`,`co_yield_statement`];function A(e){return M(e,`default`)?e.default:e}function j(e){return c[e]}function M(e,t){return typeof e!=`object`||!e||!(t in e)?!1:!!e[t]}const N=[{name:`javascript`,aliases:[`js`,`mjs`,`cjs`],parserLanguage:A(i)},{name:`jsx`,parserLanguage:A(i)},{name:`typescript`,aliases:[`ts`],parserLanguage:j(`typescript`)},{name:`tsx`,parserLanguage:j(`tsx`)},{name:`python`,aliases:[`py`],parserLanguage:A(a),ncssNodeTypes:S},{name:`go`,parserLanguage:A(n),decisionNodeTypes:d,nestingNodeTypes:p,ncssNodeTypes:C},{name:`rust`,aliases:[`rs`],parserLanguage:A(s),ncssNodeTypes:w,ncssContainerNodeTypes:[`block`,`else_clause`]},{name:`java`,parserLanguage:A(r),functionNodeTypes:m,decisionNodeTypes:h,nestingNodeTypes:h,ncssNodeTypes:T},{name:`ruby`,aliases:[`rb`],parserLanguage:A(o),functionNodeTypes:g,decisionNodeTypes:_,nestingNodeTypes:_,ncssNodeTypes:E,ncssContainerNodeTypes:D},{name:`c`,parserLanguage:A(e),functionNodeTypes:v,decisionNodeTypes:y,nestingNodeTypes:y,ncssNodeTypes:O},{name:`cpp`,aliases:[`c++`,`cxx`],parserLanguage:A(t),functionNodeTypes:v,decisionNodeTypes:b,nestingNodeTypes:b,ncssNodeTypes:k}].map(e=>({functionNodeTypes:l,decisionNodeTypes:u,nestingNodeTypes:f,ncssNodeTypes:x,...e}));function P(e=N){let t=new Map;for(let n of e){t.set(n.name,n);for(let e of n.aliases??[])t.set(e,n)}return t}const F=N.map(e=>e.name);export{P as createLanguageRegistry,N as defaultLanguages,F as supportedLanguages};
+const e=[{name:`javascript`,aliases:[`js`,`mjs`,`cjs`]},{name:`jsx`},{name:`typescript`,aliases:[`ts`]},{name:`tsx`},{name:`python`,aliases:[`py`]},{name:`go`},{name:`rust`,aliases:[`rs`]},{name:`java`},{name:`ruby`,aliases:[`rb`]},{name:`c`},{name:`cpp`,aliases:[`c++`,`cxx`]}];function t(t=e){let n=new Map;for(let e of t){n.set(e.name,e);for(let t of e.aliases??[])n.set(t,e)}return n}const n=e.map(e=>e.name);export{t as createLanguageRegistry,e as defaultLanguages,n as supportedLanguages};
 //# sourceMappingURL=languages.js.map

package/dist/languages.js.map

@@ -1 +1 @@
-{"version":3,"file":"languages.js","names":["grammars"],"sources":["../src/languages.ts"],"sourcesContent":["import C from 'tree-sitter-c';\nimport Cpp from 'tree-sitter-cpp';\nimport Go from 'tree-sitter-go';\nimport Java from 'tree-sitter-java';\nimport JavaScript from 'tree-sitter-javascript';\nimport Python from 'tree-sitter-python';\nimport Ruby from 'tree-sitter-ruby';\nimport Rust from 'tree-sitter-rust';\nimport TypeScript from 'tree-sitter-typescript';\nimport type { LanguageDefinition, LanguageName, ParserLanguage } from './types.js';\n\ntype GrammarModule = unknown;\n\nconst commonFunctionNodes = [\n  'function',\n  'function_declaration',\n  'function_definition',\n  'function_expression',\n  'function_item',\n  'function_signature_item',\n  'function_declarator',\n  'func_literal',\n  'method_declaration',\n  'method_definition',\n  'method_spec',\n  'arrow_function',\n  'generator_function',\n  'generator_function_declaration',\n  'lambda',\n  'lambda_expression',\n  'closure_expression',\n] as const;\n\nconst commonDecisionNodes = [\n  'if_statement',\n  'elif_clause',\n  'else_if_clause',\n  'for_statement',\n  'for_in_statement',\n  'while_statement',\n  'do_statement',\n  'catch_clause',\n  'except_clause',\n  'case_clause',\n  'switch_case',\n  'match_arm',\n  'conditional_expression',\n  'ternary_expression',\n  'if_expression',\n  'while_expression',\n  'for_expression',\n  'loop_expression',\n] as const;\n\n// Go `switch`/`select` branches are `*_case` nodes, not the `case_clause`/`switch_case` of other grammars.\nconst goDecisionNodes = [...commonDecisionNodes, 'expression_case', 'type_case', 'communication_case'] as const;\n\n// Default switch branches add no decision, but their contents are nested inside the switch like\n// any other arm, so they appear in the nesting sets only.\nconst commonNestingNodes = [...commonDecisionNodes, 'switch_default'] as const;\nconst goNestingNodes = [...goDecisionNodes, 'default_case'] as const;\n\nconst javaFunctionNodes = [\n  ...commonFunctionNodes,\n  'constructor_declaration',\n  'compact_constructor_declaration',\n] as const;\nconst javaDecisionNodes = [\n  ...commonDecisionNodes,\n  'enhanced_for_statement',\n  'switch_block_statement_group',\n  'switch_rule',\n] as const;\n\n// Ruby node types are keyword-like (`if`, `while`, ...), so they must stay Ruby-specific: the same\n// strings appear as anonymous keyword tokens in other grammars and would be double-counted there.\n// `block`/`do_block` are Ruby's closures (`items.map { ... }`), the analog of JS callbacks.\nconst rubyFunctionNodes = ['method', 'singleton_method', 'lambda', 'block', 'do_block'] as const;\nconst rubyDecisionNodes = [\n  'if',\n  'elsif',\n  'unless',\n  'while',\n  'until',\n  'for',\n  'when',\n  'in_clause',\n  'rescue',\n  'conditional',\n  'if_modifier',\n  'unless_modifier',\n  'while_modifier',\n  'until_modifier',\n  'rescue_modifier',\n] as const;\n\n// `function_declarator` must stay out: it is nested inside every `function_definition` (which\n// would double-count) and also appears in body-less prototypes.\nconst cFunctionNodes = ['function_definition', 'lambda_expression'] as const;\nconst cDecisionNodes = [...commonDecisionNodes, 'case_statement'] as const;\nconst cppDecisionNodes = [...cDecisionNodes, 'for_range_loop'] as const;\n\n// NCSS node sets: every listed type counts as one non-commenting source statement. The Java set is\n// calibrated against PMD's NcssCount rule (`try` counts 0; `else`, `case`/`default` labels,\n// `catch`, `finally`, and try-with-resources resources count 1 each); the other languages follow\n// the same conventions with their grammar's node types.\nconst jsNcssNodes = [\n  'import_statement',\n  'export_statement',\n  'lexical_declaration',\n  'variable_declaration',\n  'function_declaration',\n  'function_signature',\n  'generator_function_declaration',\n  'class_declaration',\n  'abstract_class_declaration',\n  'module',\n  'method_definition',\n  'abstract_method_signature',\n  'class_static_block',\n  'field_definition',\n  'public_field_definition',\n  'type_alias_declaration',\n  'interface_declaration',\n  'enum_declaration',\n  'expression_statement',\n  'if_statement',\n  'else_clause',\n  'switch_statement',\n  'switch_case',\n  'switch_default',\n  'for_statement',\n  'for_in_statement',\n  'while_statement',\n  'do_statement',\n  'catch_clause',\n  'finally_clause',\n  'labeled_statement',\n  'return_statement',\n  'break_statement',\n  'continue_statement',\n  'throw_statement',\n  'debugger_statement',\n  'with_statement',\n] as const;\n\nconst pythonNcssNodes = [\n  'import_statement',\n  'import_from_statement',\n  'future_import_statement',\n  'print_statement',\n  'exec_statement',\n  'assert_statement',\n  'expression_statement',\n  'return_statement',\n  'delete_statement',\n  'raise_statement',\n  'pass_statement',\n  'break_statement',\n  'continue_statement',\n  'global_statement',\n  'nonlocal_statement',\n  'if_statement',\n  'elif_clause',\n  'else_clause',\n  'for_statement',\n  'while_statement',\n  'except_clause',\n  'except_group_clause',\n  'finally_clause',\n  'with_statement',\n  'match_statement',\n  'case_clause',\n  'function_definition',\n  'class_definition',\n  'type_alias_statement',\n] as const;\n\nconst goNcssNodes = [\n  'package_clause',\n  'import_spec',\n  'type_spec',\n  'type_alias',\n  'const_spec',\n  'var_spec',\n  'function_declaration',\n  'method_declaration',\n  // Struct fields and interface members count contextually (see ncss.ts): only inside a named\n  // `type T struct/interface { ... }`, not in inline anonymous types, which are part of one\n  // declaration.\n  'short_var_declaration',\n  'expression_statement',\n  'send_statement',\n  'inc_statement',\n  'dec_statement',\n  'assignment_statement',\n  'if_statement',\n  'for_statement',\n  'expression_switch_statement',\n  'type_switch_statement',\n  'select_statement',\n  'expression_case',\n  'type_case',\n  'communication_case',\n  'default_case',\n  'return_statement',\n  'break_statement',\n  'continue_statement',\n  'goto_statement',\n  'fallthrough_statement',\n  'defer_statement',\n  'go_statement',\n  'labeled_statement',\n] as const;\n\nconst rustNcssNodes = [\n  'use_declaration',\n  'extern_crate_declaration',\n  'foreign_mod_item',\n  'mod_item',\n  'const_item',\n  'static_item',\n  'struct_item',\n  'enum_item',\n  'union_item',\n  'trait_item',\n  'impl_item',\n  'function_item',\n  'function_signature_item',\n  'type_item',\n  'associated_type',\n  'macro_definition',\n  'field_declaration',\n  'let_declaration',\n  'expression_statement',\n  'else_clause',\n  'match_arm',\n] as const;\n\nconst javaNcssNodes = [\n  'package_declaration',\n  'import_declaration',\n  'module_declaration',\n  'requires_module_directive',\n  'exports_module_directive',\n  'opens_module_directive',\n  'uses_module_directive',\n  'provides_module_directive',\n  'class_declaration',\n  'interface_declaration',\n  'enum_declaration',\n  'annotation_type_declaration',\n  'annotation_type_element_declaration',\n  'record_declaration',\n  'field_declaration',\n  'constant_declaration',\n  'method_declaration',\n  'constructor_declaration',\n  'compact_constructor_declaration',\n  'static_initializer',\n  'explicit_constructor_invocation',\n  'local_variable_declaration',\n  'expression_statement',\n  'if_statement',\n  'while_statement',\n  'do_statement',\n  'for_statement',\n  'enhanced_for_statement',\n  'switch_statement',\n  'switch_expression',\n  'switch_label',\n  'break_statement',\n  'continue_statement',\n  'return_statement',\n  'throw_statement',\n  'assert_statement',\n  'synchronized_statement',\n  'labeled_statement',\n  'yield_statement',\n  'resource',\n  'catch_clause',\n  'finally_clause',\n] as const;\n\n// Ruby statements have no wrapper node types, so bodies are counted positionally (see\n// ncssContainerNodeTypes); only clause nodes hanging off non-container parents need listing.\nconst rubyNcssNodes = ['elsif', 'else', 'when', 'in_clause', 'rescue', 'ensure'] as const;\nconst rubyNcssContainers = [\n  'program',\n  'body_statement',\n  'then',\n  'else',\n  'do',\n  'block_body',\n  'begin',\n  'ensure',\n] as const;\n\nconst cNcssNodes = [\n  'preproc_include',\n  'preproc_def',\n  'preproc_function_def',\n  'declaration',\n  'type_definition',\n  'field_declaration',\n  'function_definition',\n  // Counted only when they carry a body (see bodylessNcssSpecifierTypes in ncss.ts).\n  'struct_specifier',\n  'enum_specifier',\n  'union_specifier',\n  'expression_statement',\n  'if_statement',\n  'else_clause',\n  'switch_statement',\n  'case_statement',\n  'for_statement',\n  'while_statement',\n  'do_statement',\n  'return_statement',\n  'break_statement',\n  'continue_statement',\n  'goto_statement',\n  'labeled_statement',\n] as const;\nconst cppNcssNodes = [\n  ...cNcssNodes,\n  'class_specifier',\n  'namespace_definition',\n  'using_declaration',\n  'alias_declaration',\n  'namespace_alias_definition',\n  'concept_definition',\n  'static_assert_declaration',\n  'for_range_loop',\n  'catch_clause',\n  'throw_statement',\n  'co_return_statement',\n  'co_yield_statement',\n] as const;\n\nfunction normalizeGrammar(module: GrammarModule): ParserLanguage {\n  if (isGrammarWrapper(module, 'default')) {\n    return module.default;\n  }\n\n  return module;\n}\n\nfunction getTypeScriptGrammar(name: 'typescript' | 'tsx'): ParserLanguage {\n  const grammars = TypeScript as unknown as Record<string, GrammarModule>;\n  return grammars[name];\n}\n\nfunction isGrammarWrapper(value: GrammarModule, key: 'default'): value is Record<typeof key, ParserLanguage> {\n  if (typeof value !== 'object' || value === null || !(key in value)) {\n    return false;\n  }\n\n  return Boolean((value as Record<string, ParserLanguage>)[key]);\n}\n\nexport const defaultLanguages: readonly LanguageDefinition[] = [\n  {\n    name: 'javascript',\n    aliases: ['js', 'mjs', 'cjs'],\n    parserLanguage: normalizeGrammar(JavaScript as unknown as GrammarModule),\n  },\n  {\n    name: 'jsx',\n    parserLanguage: normalizeGrammar(JavaScript as unknown as GrammarModule),\n  },\n  {\n    name: 'typescript',\n    aliases: ['ts'],\n    parserLanguage: getTypeScriptGrammar('typescript'),\n  },\n  {\n    name: 'tsx',\n    parserLanguage: getTypeScriptGrammar('tsx'),\n  },\n  {\n    name: 'python',\n    aliases: ['py'],\n    parserLanguage: normalizeGrammar(Python as unknown as GrammarModule),\n    ncssNodeTypes: pythonNcssNodes,\n  },\n  {\n    name: 'go',\n    parserLanguage: normalizeGrammar(Go as unknown as GrammarModule),\n    decisionNodeTypes: goDecisionNodes,\n    nestingNodeTypes: goNestingNodes,\n    ncssNodeTypes: goNcssNodes,\n  },\n  {\n    name: 'rust',\n    aliases: ['rs'],\n    parserLanguage: normalizeGrammar(Rust as unknown as GrammarModule),\n    ncssNodeTypes: rustNcssNodes,\n    // `else_clause` is a container so `else if` chains count the nested if_expression; a plain\n    // `else { ... }` is unaffected because its only child is a `block`, itself a container.\n    ncssContainerNodeTypes: ['block', 'else_clause'],\n  },\n  {\n    name: 'java',\n    parserLanguage: normalizeGrammar(Java as unknown as GrammarModule),\n    functionNodeTypes: javaFunctionNodes,\n    decisionNodeTypes: javaDecisionNodes,\n    nestingNodeTypes: javaDecisionNodes,\n    ncssNodeTypes: javaNcssNodes,\n  },\n  {\n    name: 'ruby',\n    aliases: ['rb'],\n    parserLanguage: normalizeGrammar(Ruby as unknown as GrammarModule),\n    functionNodeTypes: rubyFunctionNodes,\n    decisionNodeTypes: rubyDecisionNodes,\n    nestingNodeTypes: rubyDecisionNodes,\n    ncssNodeTypes: rubyNcssNodes,\n    ncssContainerNodeTypes: rubyNcssContainers,\n  },\n  {\n    name: 'c',\n    parserLanguage: normalizeGrammar(C as unknown as GrammarModule),\n    functionNodeTypes: cFunctionNodes,\n    decisionNodeTypes: cDecisionNodes,\n    nestingNodeTypes: cDecisionNodes,\n    ncssNodeTypes: cNcssNodes,\n  },\n  {\n    name: 'cpp',\n    aliases: ['c++', 'cxx'],\n    parserLanguage: normalizeGrammar(Cpp as unknown as GrammarModule),\n    functionNodeTypes: cFunctionNodes,\n    decisionNodeTypes: cppDecisionNodes,\n    nestingNodeTypes: cppDecisionNodes,\n    ncssNodeTypes: cppNcssNodes,\n  },\n].map((language) => ({\n  functionNodeTypes: commonFunctionNodes,\n  decisionNodeTypes: commonDecisionNodes,\n  nestingNodeTypes: commonNestingNodes,\n  ncssNodeTypes: jsNcssNodes,\n  ...language,\n}));\n\nexport function createLanguageRegistry(\n  languages: readonly LanguageDefinition[] = defaultLanguages\n): Map<LanguageName, LanguageDefinition> {\n  const registry = new Map<LanguageName, LanguageDefinition>();\n\n  for (const language of languages) {\n    registry.set(language.name, language);\n    for (const alias of language.aliases ?? []) {\n      registry.set(alias, language);\n    }\n  }\n\n  return registry;\n}\n\nexport const supportedLanguages = defaultLanguages.map((language) => language.name);\n"],"mappings":"wSAaA,MAAM,EAAsB,CAC1B,WACA,uBACA,sBACA,sBACA,gBACA,0BACA,sBACA,eACA,qBACA,oBACA,cACA,iBACA,qBACA,iCACA,SACA,oBACA,oBACF,EAEM,EAAsB,CAC1B,eACA,cACA,iBACA,gBACA,mBACA,kBACA,eACA,eACA,gBACA,cACA,cACA,YACA,yBACA,qBACA,gBACA,mBACA,iBACA,iBACF,EAGM,EAAkB,CAAC,GAAG,EAAqB,kBAAmB,YAAa,oBAAoB,EAI/F,EAAqB,CAAC,GAAG,EAAqB,gBAAgB,EAC9D,EAAiB,CAAC,GAAG,EAAiB,cAAc,EAEpD,EAAoB,CACxB,GAAG,EACH,0BACA,iCACF,EACM,EAAoB,CACxB,GAAG,EACH,yBACA,+BACA,aACF,EAKM,EAAoB,CAAC,SAAU,mBAAoB,SAAU,QAAS,UAAU,EAChF,EAAoB,CACxB,KACA,QACA,SACA,QACA,QACA,MACA,OACA,YACA,SACA,cACA,cACA,kBACA,iBACA,iBACA,iBACF,EAIM,EAAiB,CAAC,sBAAuB,mBAAmB,EAC5D,EAAiB,CAAC,GAAG,EAAqB,gBAAgB,EAC1D,EAAmB,CAAC,GAAG,EAAgB,gBAAgB,EAMvD,EAAc,iqBAsCpB,EAEM,EAAkB,gfA8BxB,EAEM,EAAc,kgBAmCpB,EAEM,EAAgB,CACpB,kBACA,2BACA,mBACA,WACA,aACA,cACA,cACA,YACA,aACA,aACA,YACA,gBACA,0BACA,YACA,kBACA,mBACA,oBACA,kBACA,uBACA,cACA,WACF,EAEM,EAAgB,k1BA2CtB,EAIM,EAAgB,CAAC,QAAS,OAAQ,OAAQ,YAAa,SAAU,QAAQ,EACzE,EAAqB,CACzB,UACA,iBACA,OACA,OACA,KACA,aACA,QACA,QACF,EAEM,EAAa,CACjB,kBACA,cACA,uBACA,cACA,kBACA,oBACA,sBAEA,mBACA,iBACA,kBACA,uBACA,eACA,cACA,mBACA,iBACA,gBACA,kBACA,eACA,mBACA,kBACA,qBACA,iBACA,mBACF,EACM,EAAe,CACnB,GAAG,EACH,kBACA,uBACA,oBACA,oBACA,6BACA,qBACA,4BACA,iBACA,eACA,kBACA,sBACA,oBACF,EAEA,SAAS,EAAiB,EAAuC,CAK/D,OAJI,EAAiB,EAAQ,SAAS,EAC7B,EAAO,QAGT,CACT,CAEA,SAAS,EAAqB,EAA4C,CAExE,OAAOA,EAAS,EAClB,CAEA,SAAS,EAAiB,EAAsB,EAA6D,CAK3G,OAJI,OAAO,GAAU,WAAY,GAAkB,EAAE,KAAO,GACnD,GAGF,EAAS,EAAyC,EAC3D,CAEA,MAAa,EAAkD,CAC7D,CACE,KAAM,aACN,QAAS,CAAC,KAAM,MAAO,KAAK,EAC5B,eAAgB,EAAiB,CAAsC,CACzE,EACA,CACE,KAAM,MACN,eAAgB,EAAiB,CAAsC,CACzE,EACA,CACE,KAAM,aACN,QAAS,CAAC,IAAI,EACd,eAAgB,EAAqB,YAAY,CACnD,EACA,CACE,KAAM,MACN,eAAgB,EAAqB,KAAK,CAC5C,EACA,CACE,KAAM,SACN,QAAS,CAAC,IAAI,EACd,eAAgB,EAAiB,CAAkC,EACnE,cAAe,CACjB,EACA,CACE,KAAM,KACN,eAAgB,EAAiB,CAA8B,EAC/D,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,CACjB,EACA,CACE,KAAM,OACN,QAAS,CAAC,IAAI,EACd,eAAgB,EAAiB,CAAgC,EACjE,cAAe,EAGf,uBAAwB,CAAC,QAAS,aAAa,CACjD,EACA,CACE,KAAM,OACN,eAAgB,EAAiB,CAAgC,EACjE,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,CACjB,EACA,CACE,KAAM,OACN,QAAS,CAAC,IAAI,EACd,eAAgB,EAAiB,CAAgC,EACjE,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,EACf,uBAAwB,CAC1B,EACA,CACE,KAAM,IACN,eAAgB,EAAiB,CAA6B,EAC9D,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,CACjB,EACA,CACE,KAAM,MACN,QAAS,CAAC,MAAO,KAAK,EACtB,eAAgB,EAAiB,CAA+B,EAChE,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,CACjB,CACF,CAAC,CAAC,IAAK,IAAc,CACnB,kBAAmB,EACnB,kBAAmB,EACnB,iBAAkB,EAClB,cAAe,EACf,GAAG,CACL,EAAE,EAEF,SAAgB,EACd,EAA2C,EACJ,CACvC,IAAM,EAAW,IAAI,IAErB,IAAK,IAAM,KAAY,EAAW,CAChC,EAAS,IAAI,EAAS,KAAM,CAAQ,EACpC,IAAK,IAAM,KAAS,EAAS,SAAW,CAAC,EACvC,EAAS,IAAI,EAAO,CAAQ,CAEhC,CAEA,OAAO,CACT,CAEA,MAAa,EAAqB,EAAiB,IAAK,GAAa,EAAS,IAAI"}
+{"version":3,"file":"languages.js","names":[],"sources":["../src/languages.ts"],"sourcesContent":["import type { LanguageDefinition, LanguageName } from './types.js';\n\n/**\n * The built-in languages. Grammars and per-language node-type configuration live in the Rust\n * addon (native/src/languages.rs); this list only names the languages and their aliases so the\n * CLI and API can resolve and enumerate them without crossing the N-API boundary.\n */\nexport const defaultLanguages: readonly LanguageDefinition[] = [\n  { name: 'javascript', aliases: ['js', 'mjs', 'cjs'] },\n  { name: 'jsx' },\n  { name: 'typescript', aliases: ['ts'] },\n  { name: 'tsx' },\n  { name: 'python', aliases: ['py'] },\n  { name: 'go' },\n  { name: 'rust', aliases: ['rs'] },\n  { name: 'java' },\n  { name: 'ruby', aliases: ['rb'] },\n  { name: 'c' },\n  { name: 'cpp', aliases: ['c++', 'cxx'] },\n];\n\nexport function createLanguageRegistry(\n  languages: readonly LanguageDefinition[] = defaultLanguages\n): Map<LanguageName, LanguageDefinition> {\n  const registry = new Map<LanguageName, LanguageDefinition>();\n\n  for (const language of languages) {\n    registry.set(language.name, language);\n    for (const alias of language.aliases ?? []) {\n      registry.set(alias, language);\n    }\n  }\n\n  return registry;\n}\n\nexport const supportedLanguages = defaultLanguages.map((language) => language.name);\n"],"mappings":"AAOA,MAAa,EAAkD,CAC7D,CAAE,KAAM,aAAc,QAAS,CAAC,KAAM,MAAO,KAAK,CAAE,EACpD,CAAE,KAAM,KAAM,EACd,CAAE,KAAM,aAAc,QAAS,CAAC,IAAI,CAAE,EACtC,CAAE,KAAM,KAAM,EACd,CAAE,KAAM,SAAU,QAAS,CAAC,IAAI,CAAE,EAClC,CAAE,KAAM,IAAK,EACb,CAAE,KAAM,OAAQ,QAAS,CAAC,IAAI,CAAE,EAChC,CAAE,KAAM,MAAO,EACf,CAAE,KAAM,OAAQ,QAAS,CAAC,IAAI,CAAE,EAChC,CAAE,KAAM,GAAI,EACZ,CAAE,KAAM,MAAO,QAAS,CAAC,MAAO,KAAK,CAAE,CACzC,EAEA,SAAgB,EACd,EAA2C,EACJ,CACvC,IAAM,EAAW,IAAI,IAErB,IAAK,IAAM,KAAY,EAAW,CAChC,EAAS,IAAI,EAAS,KAAM,CAAQ,EACpC,IAAK,IAAM,KAAS,EAAS,SAAW,CAAC,EACvC,EAAS,IAAI,EAAO,CAAQ,CAEhC,CAEA,OAAO,CACT,CAEA,MAAa,EAAqB,EAAiB,IAAK,GAAa,EAAS,IAAI"}

package/dist/metrics.cjs

@@ -1,2 +1,2 @@
-"use strict";const e=require("./_virtual/_rolldown/runtime.cjs"),t=require("./duplication.cjs"),n=require("./languages.cjs"),r=require("./depDegree.cjs"),i=require("./ncss.cjs"),a=require("./nativeMetrics.cjs");let o=require("tree-sitter");o=e.__toESM(o,1);const s=new Set([`&&`,`||`,`and`,`or`]),c=new Set(`+,-,*,/,%,**,=,+=,-=,*=,/=,%=,==,!=,===,!==,<,<=,>,>=,!,~,&,|,^,++,--,<<,>>,>>>,=>,**=,<<=,>>=,>>>=,&=,|=,^=,&&=,||=,??=,??,?.,?,//,//=,@,@=,:=,<-,<=>,=~,..,...,..=,&&,||,!~,&^,&^=,&.,.,->,::,->*,.*,sizeof,alignof,defined?,as,bitand,bitor,xor,compl,and_eq,or_eq,xor_eq,not_eq,and,or,not,in,is,instanceof,typeof,new,delete,return,throw,raise,yield,await,co_await,co_yield,co_return,break,continue`.split(`,`)),l=new Set(`identifier.property_identifier.field_identifier.type_identifier.constant.instance_variable.class_variable.global_variable.simple_symbol.self.this.super.primitive_type.boolean_type.void_type.auto.number.integer.float.integer_literal.float_literal.int_literal.rune_literal.imaginary_literal.number_literal.decimal_integer_literal.hex_integer_literal.octal_integer_literal.binary_integer_literal.decimal_floating_point_literal.hex_floating_point_literal.string.string_literal.raw_string_literal.string_fragment.multiline_string_fragment.string_content.raw_string_content.template_string.character_literal.char_literal.character.true.false.null.null_literal.undefined.nil.none`.split(`.`)),u=new Set([`interpreted_string_literal`,`regex`,`user_defined_literal`,`integral_type`,`floating_point_type`,`sized_type_specifier`,`placeholder_type_specifier`]);var d=class{registry=n.createLanguageRegistry();registerLanguage(e){ee(e),i.invalidateNcssSetsCache(e),this.registry.set(e.name,e);for(let t of e.aliases??[])this.registry.set(t,e)}getSupportedLanguages(){return[...new Set([...this.registry.values()].map(e=>e.name))]}measure(e,n){let r=this.registry.get(n.language);if(!r)throw Error(`Unsupported language: ${n.language}`);let o=p(n)?a.measureWithNativeBackend(e,r,n.includeSyntaxTree??!1):void 0;if(o)return y(o,n.includeSyntaxTree??!1);let s=f(e,r),c=b(s,H(s,new Set(r.functionNodeTypes)).filter(e=>!E(e)&&T(e)),r,e),l=ne(s,r),{lines:u,codeLineNumbers:d}=U(e,s);return{language:r.name,bytes:Buffer.byteLength(e),lines:u,functions:c,cognitiveComplexity:l.cognitiveComplexity,maxCognitiveComplexity:ve(c),nestingDepth:l.nestingDepth,ncssCount:i.countNcss(s,r),duplication:t.measureDuplication(s,d,n.duplication),halstead:W(s,e),syntaxTree:n.includeSyntaxTree?s.toString():void 0}}collectDuplicationCandidates(e,t){return this.collectCrossFileDuplicationFileData(e,t).candidates}collectFunctionTokenSequences(e,t){let{language:n,root:r}=this.parse(e,t);return H(r,new Set(n.functionNodeTypes)).filter(e=>!E(e)&&T(e)).map(e=>le(e))}collectCrossFileDuplicationFileData(e,n){let{root:r}=this.parse(e,n);return{...t.collectCrossFileDuplicateCandidates(r,n.duplication),codeLineNumbers:U(e,r).codeLineNumbers}}parse(e,t){let n=this.registry.get(t.language);if(!n)throw Error(`Unsupported language: ${t.language}`);return{language:n,root:f(e,n)}}};function f(e,t){let n=new o.default;return n.setLanguage(t.parserLanguage),n.parse(e,void 0,{bufferSize:e.length+1}).rootNode}function p(e){let n=e.duplication;return(n?.minTokens??t.defaultDuplicationOptions.minTokens)===t.defaultDuplicationOptions.minTokens&&(n?.maxGapTokens??t.defaultDuplicationOptions.maxGapTokens)===t.defaultDuplicationOptions.maxGapTokens&&(n?.minSimilarityPercent??t.defaultDuplicationOptions.minSimilarityPercent)===t.defaultDuplicationOptions.minSimilarityPercent}const m=new d;function h(e,t){return m.measure(e,t)}function g(e,t){return m.collectDuplicationCandidates(e,t)}function _(e,t){return m.collectFunctionTokenSequences(e,t)}function v(e,t){return m.collectCrossFileDuplicationFileData(e,t)}function y(e,t){return{language:e.language,bytes:e.bytes,lines:e.lines,functions:e.functions.map(e=>({name:e.name,nodeType:e.nodeType,startLine:e.startLine,startColumn:e.startColumn,endLine:e.endLine,cognitiveComplexity:e.cognitiveComplexity,nestingDepth:e.nestingDepth,ncss:e.ncss,parameterCount:e.parameterCount,halstead:G(e.halsteadCounts),depDegree:e.depDegree})),cognitiveComplexity:e.cognitiveComplexity,maxCognitiveComplexity:e.maxCognitiveComplexity,nestingDepth:e.nestingDepth,ncssCount:e.ncssCount,duplication:e.duplication,halstead:G(e.halsteadCounts),syntaxTree:t?e.syntaxTree:void 0}}function b(e,t,n,i){let a=te(e,n),{functionNodes:o}=M(n);return t.map(e=>{let t=a.get(e.id);if(!t)throw Error(`missing body metrics for function node at line ${e.startPosition.row+1}`);return{name:ue(e),nodeType:e.type,startLine:e.startPosition.row+1,startColumn:e.startPosition.column,endLine:e.endPosition.row+1,cognitiveComplexity:t.cognitiveComplexity,nestingDepth:t.nestingDepth,ncss:t.ncss,parameterCount:x(e),halstead:W(e,i),depDegree:r.measureDepDegree(e,e=>D(e,o))}})}function x(e){if(e.childForFieldName(`parameter`))return 1;let t=C(e);if(!t)return 0;if(t.type===`identifier`)return 1;let n=new Set(Z(t,`locals`).map(e=>e.id)),r=0;for(let e of t.namedChildren)e.type===`comment`||e.type===`self_parameter`||e.type===`receiver_parameter`||e.type===`block_parameter`||e.type===`positional_separator`||e.type===`keyword_separator`||n.has(e.id)||S(e)||(r+=e.type===`parameter_declaration`?Math.max(1,Z(e,`name`).length):1);let i=t.children.filter(e=>!e.isNamed&&e.text===`...`).length;return r+i}function S(e){return e.type===`parameter_declaration`&&e.childForFieldName(`declarator`)===null&&e.childForFieldName(`type`)?.text===`void`}function C(e){let t=e.childForFieldName(`parameters`);if(t)return t;if(e.type===`compact_constructor_declaration`)return e.parent?.parent?.childForFieldName(`parameters`)??void 0;let n=e.childForFieldName(`declarator`);for(;n;){let e=n.childForFieldName(`parameters`);if(e)return e;n=J(n)}return e.namedChildren.find(e=>e.type===`formal_parameters`||e.type===`parameter_list`)}const w=new Set([`function_definition`,`constructor_declaration`,`compact_constructor_declaration`,`function_signature_item`]);function T(e){return!w.has(e.type)||e.childForFieldName(`body`)!==null||e.namedChildren.some(e=>e.type===`try_statement`)}function E(e){return(e.type===`block`||e.type===`do_block`)&&e.parent?.type===`lambda`}function D(e,t){return t.has(e.type)&&!E(e)}const O=new Set([`switch_statement`,`switch_expression`,`expression_switch_statement`,`type_switch_statement`,`select_statement`,`match_expression`,`match_statement`,`case`,`case_match`]),k=new Set([`case_clause`,`switch_case`,`switch_block_statement_group`,`switch_rule`,`case_statement`,`expression_case`,`type_case`,`communication_case`,`match_arm`,`when`,`in_clause`]),A=new Set([`if_statement`,`if_expression`,`if`,`unless`]),j=new WeakMap;function ee(e){j.delete(e)}function M(e){let t=j.get(e);return t||(t={functionNodes:new Set(e.functionNodeTypes),decisionNodes:new Set(e.decisionNodeTypes),nestingNodes:new Set(e.nestingNodeTypes)},j.set(e,t)),t}function N(e,t){return{cognitiveComplexity:0,nestingSensitiveCount:0,nestingDepth:0,ncss:0,hasOwnNcssContribution:!1,entryCognitiveNesting:e,entryStructuralNesting:t}}function te(e,t){let{functionNodes:n,decisionNodes:r,nestingNodes:a}=M(t),{countable:o,containers:s}=i.getNcssSets(t),c=new Map,l=[N(0,0)];function u(e,t,d,f,p,m){let h=e.type===`class_body`&&f;h&&(p=!0,d+=1);let g=D(e,n);g&&(f&&!m&&(d+=1),f=!0);let _=l.at(-1),v=t+d-_.entryCognitiveNesting,y=!p&&!g,b=e.isNamed&&r.has(e.type)&&!B(e),x=e.isNamed&&k.has(e.type),S=e.isNamed&&(a.has(e.type)||e.type===`else`&&(e.parent?.type===`case`||e.parent?.type===`case_match`)),C=b&&z(e);b&&!x&&(C?_.cognitiveComplexity+=1:(_.cognitiveComplexity+=1+v,_.nestingSensitiveCount+=1)),e.isNamed&&O.has(e.type)&&(_.cognitiveComplexity+=1+v,_.nestingSensitiveCount+=1),_.cognitiveComplexity+=P(e),F(e)&&(_.cognitiveComplexity+=1),V(e)&&I(e)&&(_.cognitiveComplexity+=1),R(e)&&(_.cognitiveComplexity+=1);let w=S&&!C?t+1:t;y&&(_.nestingDepth=Math.max(_.nestingDepth,w-_.entryStructuralNesting)),g&&l.push(N(w+d,w));let T=i.ncssContribution(e,o,s),E=l.at(-1);E.ncss+=T,g&&T>0&&(E.hasOwnNcssContribution=!0);for(let t of e.children)u(t,w,d,f,!g&&p,h);if(g){let t=l.pop();c.set(e.id,{cognitiveComplexity:t.cognitiveComplexity,nestingDepth:t.nestingDepth,ncss:t.ncss+ +!t.hasOwnNcssContribution});let n=l.at(-1);n.cognitiveComplexity+=t.cognitiveComplexity+t.nestingSensitiveCount*(t.entryCognitiveNesting-n.entryCognitiveNesting),n.nestingSensitiveCount+=t.nestingSensitiveCount,n.ncss+=t.ncss}}return u(e,0,0,!1,!1,!1),c}function ne(e,t){let n=0,r=0,{functionNodes:i,decisionNodes:a,nestingNodes:o}=M(t);function s(e,t,c,l,u){let d=e.type===`class_body`&&l;d&&(c+=1),D(e,i)&&(l&&!u&&(c+=1),l=!0);let f=t+c,p=e.isNamed&&a.has(e.type)&&!B(e),m=e.isNamed&&k.has(e.type),h=e.isNamed&&(o.has(e.type)||e.type===`else`&&(e.parent?.type===`case`||e.parent?.type===`case_match`)),g=p&&z(e);p&&!m&&(n+=g?1:1+f),e.isNamed&&O.has(e.type)&&(n+=1+f),n+=P(e),F(e)&&(n+=1),V(e)&&I(e)&&(n+=1),R(e)&&(n+=1);let _=h&&!g?t+1:t;r=Math.max(r,_);for(let t of e.children)s(t,_,c,l,d)}for(let t of e.children)s(t,0,0,!1,!1);return{cognitiveComplexity:n,nestingDepth:r}}function P(e){if(!e.isNamed)return 0;if(e.type===`else`)return e.parent?.type===`case`||e.parent?.type===`case_match`?0:1;if(e.type===`else_clause`)return+!e.namedChildren.some(e=>A.has(e.type));if(e.type!==`if_statement`&&e.type!==`if_expression`)return 0;let t=0;for(let n=0;n<e.childCount;n+=1){let r=e.child(n);r&&e.fieldNameForChild(n)===`alternative`&&r.type!==`else_clause`&&r.type!==`elif_clause`&&!A.has(r.type)&&(t+=1)}return t}function F(e){return e.isNamed?e.type===`goto_statement`?!0:e.type===`break_expression`||e.type===`continue_expression`?e.namedChildren.some(e=>e.type===`label`||e.type===`loop_label`):(e.type===`break_statement`||e.type===`continue_statement`)&&e.namedChildren.some(e=>!i.commentNodeTypes.has(e.type)):!1}const re=new Set([`parenthesized_expression`,`parenthesized_statements`]);function I(e){let t=e.parent;if(!t)return!0;let n=t.parent;for(;n&&re.has(n.type);)n=n.parent;return!n||n.type!==t.type||L(ie(n))!==L(e.text)}function L(e){return e===`and`?`&&`:e===`or`?`||`:e}function ie(e){let t=e.childForFieldName(`operator`);return t?t.text:e.children.find(e=>!e.isNamed&&s.has(e.text))?.text}function R(e){return e.isNamed?e.type===`guard`||e.type===`if_guard`||e.type===`unless_guard`||e.type===`if_clause`||e.type===`match_pattern`&&e.children.some(e=>!e.isNamed&&e.type===`if`):!1}function z(e){if(e.type===`elsif`||e.type===`elif_clause`)return!0;if(e.type!==`if_statement`&&e.type!==`if_expression`&&e.type!==`if`)return!1;let t=e.parent;return t?t.type===`else_clause`||t.childForFieldName(`alternative`)?.id===e.id:!1}function B(e){if(e.type===`case_statement`)return e.childForFieldName(`value`)===null;if(e.type===`switch_block_statement_group`||e.type===`switch_rule`){let t=e.namedChildren.find(e=>e.type===`switch_label`);return t!==void 0&&t.namedChildCount===0}if(e.type===`case_clause`||e.type===`match_arm`){let t=e.namedChildren.find(e=>e.type===`case_pattern`||e.type===`match_pattern`);if(!t)return!1;if(t.child(0)?.type===`_`&&(t.childCount===1||t.child(1)?.type===`if`))return!0;let n=t.namedChildCount===1?t.namedChild(0):void 0;return e.type===`case_clause`&&n?.type===`dotted_name`&&n.namedChildCount===1&&n.namedChild(0)?.type===`identifier`}return e.type===`in_clause`&&e.namedChild(0)?.type===`identifier`}const ae=new Set([`binary_expression`,`binary`,`boolean_operator`]);function V(e){if(e.isNamed||!s.has(e.text))return!1;let t=e.parent;return t!==null&&ae.has(t.type)}function H(e,t){let n=[];function r(e){t.has(e.type)&&n.push(e);for(let t of e.namedChildren)r(t)}return r(e),n}function U(e,t){let n=e.length===0?[]:e.split(/\r\n|\n|\r/),r=new Map;for(let e of oe(t)){let t=r.get(e.line)??[];t.push(e),r.set(e.line,t)}let i=0,a=0,o=new Set;for(let[e,t]of n.entries()){if(t.trim()===``){i+=1;continue}se(t,r.get(e)??[])?a+=1:o.add(e+1)}return{lines:{total:n.length,code:o.size,comment:a,blank:i},codeLineNumbers:o}}function oe(e){let t=[];function n(e){if(e.type===`comment`||e.type===`line_comment`||e.type===`block_comment`)for(let n=e.startPosition.row;n<=e.endPosition.row;n+=1)t.push({line:n,startColumn:n===e.startPosition.row?e.startPosition.column:0,endColumn:n===e.endPosition.row?e.endPosition.column:1/0});for(let t of e.namedChildren)n(t)}return n(e),t}function se(e,t){if(t.length===0)return!1;for(let n=0;n<e.length;n+=1)if(!/\s/u.test(e[n]??` `)&&!t.some(e=>e.startColumn<=n&&n<e.endColumn))return!1;return!0}function W(e,t){let n=new Map,r=new Map;function i(e){if(e.type!==`comment`&&e.type!==`line_comment`&&e.type!==`block_comment`){if(u.has(e.type)){Q(r,t.slice(e.startIndex,e.endIndex));return}if(e.childCount===0){let i=t.slice(e.startIndex,e.endIndex);l.has(e.type)?Q(r,i):(c.has(i)||c.has(e.type))&&_e(e,i)&&Q(n,i||e.type);return}for(let t of e.children)i(t)}}return i(e),G({distinctOperators:n.size,distinctOperands:r.size,totalOperators:$(n.values()),totalOperands:$(r.values())})}function G(e){let{distinctOperators:t,distinctOperands:n,totalOperators:r,totalOperands:i}=e,a=t+n,o=r+i,s=a===0?0:o*Math.log2(a);return{distinctOperators:t,distinctOperands:n,totalOperators:r,totalOperands:i,vocabulary:a,length:o,volume:s,effort:(n===0?0:t/2*(i/n))*s}}const ce=new Set([`identifier`,`property_identifier`,`field_identifier`,`type_identifier`,`constant`,`instance_variable`,`class_variable`,`global_variable`]);function le(e){let t=[];return K(e,t,new Map),Int32Array.from(t)}function K(e,n,r){if(e.type!==`comment`&&e.type!==`line_comment`&&e.type!==`block_comment`){if(u.has(e.type)){n.push(t.hashText(e.type));return}if(e.childCount>0){for(let t of e.children)K(t,n,r);return}if(ce.has(e.type)){let i=r.get(e.text);i===void 0&&(i=r.size,r.set(e.text,i)),n.push(t.hashText(`id${i}`));return}n.push(t.hashText(l.has(e.type)?e.type:e.text))}}function ue(e){let t=me(e);if(t)return t;let n=e.childForFieldName(`name`);if(n)return n.text;let r=de(e);if(r)return r;let i=e.parent;if(i){if(e.type===`closure_expression`&&i.type===`let_declaration`){let e=i.childForFieldName(`pattern`);return e?.type===`identifier`?e.text:void 0}return e.type===`lambda_expression`&&i.type===`init_declarator`?q(i.childForFieldName(`declarator`)):e.type===`func_literal`&&i.type===`expression_list`?pe(e,i):e.type===`lambda`&&i.type===`assignment`?Y(i):(e.type===`block`||e.type===`do_block`)&&fe(i)?i.parent?.type===`assignment`?Y(i.parent):void 0:i.childForFieldName(`name`)?.text}}function de(e){return q(e.childForFieldName(`declarator`))}function q(e){let t=e;for(;t;)switch(t.type){case`identifier`:case`field_identifier`:case`type_identifier`:case`destructor_name`:case`operator_name`:return t.text;case`operator_cast`:return`operator ${t.childForFieldName(`type`)?.text??``}`.trimEnd();case`template_function`:case`qualified_identifier`:t=t.childForFieldName(`name`);break;default:t=J(t)}}function J(e){let t=e.childForFieldName(`declarator`);if(t)return t;if(e.type===`reference_declarator`||e.type===`parenthesized_declarator`)return e.namedChild(0)??void 0}function Y(e){let t=e.childForFieldName(`left`);return t?.type===`identifier`||t?.type===`constant`?t.text:void 0}function fe(e){if(e.type!==`call`||e.childForFieldName(`receiver`))return!1;let t=e.childForFieldName(`method`);return t?.type===`identifier`&&(t.text===`lambda`||t.text===`proc`)}function pe(e,t){let n=t.parent,r=t.namedChildren.filter(e=>e.type!==`comment`).findIndex(t=>t.id===e.id);if(!(!n||r===-1)){if(n.type===`short_var_declaration`){let e=n.childForFieldName(`left`)?.namedChildren.filter(e=>e.type!==`comment`);return X(e?.[r])}if(n.type===`var_spec`){let e=Z(n,`name`)[r];return X(e)}}}function X(e){return e?.type===`identifier`&&e.text!==`_`?e.text:void 0}function me(e){let t=e;for(;t;){let e=t.parent,n=e?.parent;if(e?.type!==`arguments`||n?.type!==`call_expression`||!he(n))return;let r=n.parent;if(r?.type===`variable_declarator`)return r.childForFieldName(`name`)?.text;t=n}}function he(e){let t=e.childForFieldName(`function`)??e.namedChild(0);return t?.text===`memo`||t?.text===`React.memo`||t?.text===`forwardRef`||t?.text===`React.forwardRef`}const ge=new Set([`ternary_expression`,`conditional_expression`,`conditional`,`try_expression`,`conditional_type`]);function _e(e,t){if(t===`@`){let t=e.parent?.type;return t===`binary_operator`||t===`augmented_assignment`}if(t!==`?`)return!0;let n=e.parent?.type;return n!==void 0&&ge.has(n)}function Z(e,t){let n=[];for(let r=0;r<e.childCount;r+=1){let i=e.child(r);i&&e.fieldNameForChild(r)===t&&n.push(i)}return n}function Q(e,t){e.set(t,(e.get(t)??0)+1)}function ve(e){return e.length===0?0:Math.max(...e.map(e=>e.cognitiveComplexity))}function $(e){let t=0;for(let n of e)t+=n;return t}exports.TreeMeasurer=d,exports.collectCrossFileDuplicationFileData=v,exports.collectDuplicationCandidates=g,exports.collectFunctionTokenSequences=_,exports.defaultMeasurer=m,exports.measureCode=h;
+"use strict";const e=require("./languages.cjs"),t=require("./nativeMetrics.cjs");var n=class{registry=e.createLanguageRegistry();getSupportedLanguages(){return[...new Set([...this.registry.values()].map(e=>e.name))]}measure(e,n){let i=this.resolveLanguage(n.language),a=n.includeSyntaxTree??!1;return r(t.measureCodeNative(e,i.name,a,n.duplication),a)}collectDuplicationCandidates(e,t){return this.collectCrossFileDuplicationFileData(e,t).candidates}collectCrossFileDuplicationFileData(e,n){let r=this.resolveLanguage(n.language),i=t.collectCrossFileDataNative(e,r.name,n.duplication?.minTokens);return{candidates:i.candidates,tokens:i.tokens,containerStatements:i.containerStatements,codeLineNumbers:new Set(i.codeLineNumbers)}}collectFunctionTokenSequences(e,n){let r=this.resolveLanguage(n.language);return t.collectFunctionTokenSequencesNative(e,r.name)}resolveLanguage(e){let t=this.registry.get(e);if(!t)throw Error(`Unsupported language: ${e}`);return t}};function r(e,t){return{language:e.language,bytes:e.bytes,lines:e.lines,functions:e.functions.map(e=>({name:e.name,nodeType:e.nodeType,startLine:e.startLine,startColumn:e.startColumn,endLine:e.endLine,cognitiveComplexity:e.cognitiveComplexity,nestingDepth:e.nestingDepth,ncss:e.ncss,parameterCount:e.parameterCount,halstead:i(e.halsteadCounts),depDegree:e.depDegree})),cognitiveComplexity:e.cognitiveComplexity,maxCognitiveComplexity:e.maxCognitiveComplexity,nestingDepth:e.nestingDepth,ncssCount:e.ncssCount,duplication:e.duplication,halstead:i(e.halsteadCounts),syntaxTree:t?e.syntaxTree:void 0}}function i(e){let{distinctOperators:t,distinctOperands:n,totalOperators:r,totalOperands:i}=e,a=t+n,o=r+i,s=a===0?0:o*Math.log2(a);return{distinctOperators:t,distinctOperands:n,totalOperators:r,totalOperands:i,vocabulary:a,length:o,volume:s,effort:(n===0?0:t/2*(i/n))*s}}const a=new n;function o(e,t){return a.measure(e,t)}function s(e,t){return a.collectDuplicationCandidates(e,t)}function c(e,t){return a.collectFunctionTokenSequences(e,t)}function l(e,t){return a.collectCrossFileDuplicationFileData(e,t)}exports.TreeMeasurer=n,exports.collectCrossFileDuplicationFileData=l,exports.collectDuplicationCandidates=s,exports.collectFunctionTokenSequences=c,exports.defaultMeasurer=a,exports.measureCode=o;
 //# sourceMappingURL=metrics.cjs.map