code-gauge 4.2.1 → 4.2.2

package/dist/crossFileDuplication.cjs

@@ -1,2 +1,2 @@
-"use strict";const e=require("./duplicateSelection.cjs"),t=require("./duplication.cjs");function n(n,c){let{minTokens:l,maxGapTokens:u}=t.resolveDuplicationOptions(c),d=n.flatMap(({file:e,candidates:t},n)=>t.map(t=>({...t,regionBucket:n,file:e})));for(let e of r(n,l))d.push(e);let f=e.selectMaximalGroups(d,i,(e,t)=>e.regionBucket-t.regionBucket||e.startIndex-t.startIndex),p=a(n,u);return s(o([...f.values()],p,u),n,p)}function r(e,n){let r=[],i=[];for(let[n,{tokens:a,containerStatements:o}]of e.entries())a&&o&&(r.push(n),i.push({tokens:a,literalCountPrefix:t.buildLiteralCountPrefix(a),containers:o}));return i.length<2?[]:t.collectSequenceWindowCandidates(i,n,!0).flatMap(({candidate:t,contextIndex:n})=>{let i=r[n],a=i===void 0?void 0:e[i];return i===void 0||a===void 0?[]:[{...t,regionBucket:i,file:a.file}]})}function i(e){return e.length>=2&&new Set(e.map(e=>e.regionBucket)).size>=2}function a(e,t){let n=[],r=0;for(let{tokens:i,candidates:a}of e){n.push(r);let e=i?.length??0;if(!i)for(let t of a)e=Math.max(e,t.endTokenIndex);r+=e+t+1}return n}function o(e,n,r){let i=e.map(e=>e.map(e=>{let t=e.startTokenIndex+(n[e.regionBucket]??0),r=e.endTokenIndex+(n[e.regionBucket]??0);return{file:e.file,segments:[{startTokenIndex:t,endTokenIndex:r}],tokenCount:e.tokenCount,startTokenIndex:t,endTokenIndex:r,startIndex:e.startIndex,endIndex:e.endIndex,startLine:e.startLine,endLine:e.endLine}}).toSorted((e,t)=>e.startTokenIndex-t.startTokenIndex));return t.mergeAdjacentGroups(i,r)}function s(e,n,r){let i=[],a=new Map,o=new Map(n.map((e,t)=>[e.file,{tokens:e.tokens,codeLineNumbers:e.codeLineNumbers,offset:r[t]??0}])),s=new Map,l=0;for(let n of e){l+=t.countRedundantFragments(n);for(let e of n)c(e,o,s);let e=n.map(({file:e,startLine:t,endLine:n})=>({file:e,startLine:t,endLine:n})).toSorted((e,t)=>e.file.localeCompare(t.file)||e.startLine-t.startLine),r=[...new Set(e.map(({file:e})=>e))];for(let e of r)a.set(e,(a.get(e)??0)+1);i.push({files:r,occurrences:e,tokenCount:n[0]?.tokenCount??0})}return i.sort((e,t)=>t.tokenCount-e.tokenCount||(e.occurrences[0]?.file??``).localeCompare(t.occurrences[0]?.file??``)||(e.occurrences[0]?.startLine??0)-(t.occurrences[0]?.startLine??0)),{duplicateBlockCount:l,duplicateBlockGroupCountByFile:Object.fromEntries(a),duplicateLineNumbersByFile:Object.fromEntries([...s].map(([e,t])=>[e,[...t].toSorted((e,t)=>e-t)])),groups:i}}function c(e,n,r){let i=n.get(e.file);if(!i?.tokens)return;let a=r.get(e.file);a||(a=new Set,r.set(e.file,a));for(let n of e.segments)t.collectSegmentLines({startTokenIndex:n.startTokenIndex-i.offset,endTokenIndex:n.endTokenIndex-i.offset},i.tokens,i.codeLineNumbers,a)}exports.measureCrossFileDuplication=n;
+"use strict";const e=require("./duplicateSelection.cjs"),t=require("./duplication.cjs");function n(n,i){let{minTokens:l,maxGapTokens:u}=t.resolveDuplicationOptions(i),d=n.flatMap(({file:e,candidates:t},n)=>t.map(t=>({...t,regionBucket:n,file:e})));for(let e of r(n,l))d.push(e);let f=e.selectMaximalGroups(d,a,(e,t)=>e.regionBucket-t.regionBucket||e.startIndex-t.startIndex),p=o(n,u);return c(s([...f.values()],p,u),n,p)}function r(e,n){let r=[],i=[];for(let[n,{tokens:a,containerStatements:o}]of e.entries())a&&o&&(r.push(n),i.push({tokens:a,literalCountPrefix:t.buildLiteralCountPrefix(a),containers:o}));return i.length<2?[]:t.collectSequenceWindowCandidates(i,n,!0).flatMap(({candidate:t,contextIndex:n})=>{let i=r[n],a=i===void 0?void 0:e[i];return i===void 0||a===void 0?[]:[{...t,regionBucket:i,file:a.file}]})}function i(e){return new Set(e.map(e=>e.file)).size>=2}function a(e){return e.length>=2&&new Set(e.map(e=>e.regionBucket)).size>=2}function o(e,t){let n=[],r=0;for(let{tokens:i,candidates:a}of e){n.push(r);let e=i?.length??0;if(!i)for(let t of a)e=Math.max(e,t.endTokenIndex);r+=e+t+1}return n}function s(e,n,r){let a=e.map(e=>e.map(e=>{let t=e.startTokenIndex+(n[e.regionBucket]??0),r=e.endTokenIndex+(n[e.regionBucket]??0);return{file:e.file,spanCountedElsewhere:e.nestedInLargerGroup,nestedInLargerGroup:e.nestedInLargerGroup,segments:[{startTokenIndex:t,endTokenIndex:r}],tokenCount:e.tokenCount,startTokenIndex:t,endTokenIndex:r,startIndex:e.startIndex,endIndex:e.endIndex,startLine:e.startLine,endLine:e.endLine}}).toSorted((e,t)=>e.startTokenIndex-t.startTokenIndex));return t.mergeAdjacentGroups(a,r,i)}function c(e,n,r){let i=[],a=new Map,o=new Map(n.map((e,t)=>[e.file,{tokens:e.tokens,codeLineNumbers:e.codeLineNumbers,offset:r[t]??0}])),s=new Map,c=0;for(let n of e){c+=t.countRedundantFragments(n);for(let e of n)l(e,o,s);let e=n.map(({file:e,startLine:t,endLine:n})=>({file:e,startLine:t,endLine:n})).toSorted((e,t)=>e.file.localeCompare(t.file)||e.startLine-t.startLine),r=[...new Set(e.map(({file:e})=>e))];for(let e of r)a.set(e,(a.get(e)??0)+1);i.push({files:r,occurrences:e,tokenCount:n[0]?.tokenCount??0})}return i.sort((e,t)=>t.tokenCount-e.tokenCount||(e.occurrences[0]?.file??``).localeCompare(t.occurrences[0]?.file??``)||(e.occurrences[0]?.startLine??0)-(t.occurrences[0]?.startLine??0)),{duplicateBlockCount:c,duplicateBlockGroupCountByFile:Object.fromEntries(a),duplicateLineNumbersByFile:Object.fromEntries([...s].map(([e,t])=>[e,[...t].toSorted((e,t)=>e-t)])),groups:i}}function l(e,n,r){let i=n.get(e.file);if(!i?.tokens)return;let a=r.get(e.file);a||(a=new Set,r.set(e.file,a));for(let n of e.segments)t.collectSegmentLines({startTokenIndex:n.startTokenIndex-i.offset,endTokenIndex:n.endTokenIndex-i.offset},i.tokens,i.codeLineNumbers,a)}exports.measureCrossFileDuplication=n;
 //# sourceMappingURL=crossFileDuplication.cjs.map

package/dist/crossFileDuplication.cjs.map

@@ -1 +1 @@
-{"version":3,"file":"crossFileDuplication.cjs","names":["resolveDuplicationOptions","selectMaximalGroups","buildLiteralCountPrefix","collectSequenceWindowCandidates","mergeAdjacentGroups","countRedundantFragments"],"sources":["../src/crossFileDuplication.ts"],"sourcesContent":["import { selectMaximalGroups } from './duplicateSelection.js';\nimport {\n  buildLiteralCountPrefix,\n  collectSegmentLines,\n  collectSequenceWindowCandidates,\n  countRedundantFragments,\n  mergeAdjacentGroups,\n  resolveDuplicationOptions,\n  type CountedOccurrence,\n  type CrossFileDuplicateCandidate,\n  type CrossFileDuplicationFileData,\n  type SequenceWindowContext,\n} from './duplication.js';\nimport type { DuplicationOptions } from './types.js';\n\nexport interface CrossFileDuplicationSourceFile extends Partial<CrossFileDuplicationFileData> {\n  file: string;\n  candidates: CrossFileDuplicateCandidate[];\n}\n\nexport interface CrossFileDuplicateOccurrence {\n  endLine: number;\n  file: string;\n  startLine: number;\n}\n\nexport interface CrossFileDuplicateBlockGroup {\n  files: string[];\n  occurrences: CrossFileDuplicateOccurrence[];\n  /** Matched token count of one occurrence (all occurrences share it; gaps are not counted). */\n  tokenCount: number;\n}\n\nexport interface CrossFileDuplicationMetrics {\n  /** Number of redundant copies across all groups, counted per matched fragment like within-file. */\n  duplicateBlockCount: number;\n  /** Groups the file participates in, keyed by the file name passed in. */\n  duplicateBlockGroupCountByFile: Record<string, number>;\n  /**\n   * Per file, the 1-based code lines covered by matched tokens of its cross-file occurrences,\n   * sorted ascending. Exact like within-file duplicateLineNumbers: the unmatched gap of a merged\n   * clone and comment/blank lines inside an occurrence's bounding range are excluded (blank rows\n   * inside multi-row tokens only when the file supplied codeLineNumbers). A file that supplied\n   * only candidates (no `tokens`) has no entry — without its token stream the matched lines are\n   * unknowable, and an approximate bounding range would break this field's exactness.\n   */\n  duplicateLineNumbersByFile: Record<string, number[]>;\n  groups: CrossFileDuplicateBlockGroup[];\n}\n\ninterface SelectableCandidate extends CrossFileDuplicateCandidate {\n  regionBucket: number;\n  file: string;\n}\n\n/** A cross-file occurrence: a within-file occurrence in the project-wide token index space. */\ninterface CrossFileOccurrence extends CountedOccurrence {\n  file: string;\n}\n\n/**\n * Detects code regions duplicated across files. Per-file candidates (whole block subtrees and full\n * container runs, fingerprinted with the same normalization as within-file duplication) are joined\n * by a project-level window index over per-statement fingerprint sequences (CPD-style), so a\n * copy-pasted partial statement run embedded in different surrounding code is matched even though\n * no single file can know it repeats elsewhere. Candidates are grouped by fingerprint, and only\n * maximal, non-overlapping regions whose group spans at least two files are counted. Groups that\n * shrink to a single file during selection are shed — a within-file repeat is already reported by\n * that file's own duplication metrics. Groups separated by a small token gap within each file then\n * merge into gapped (Type-3) clone groups under `maxGapTokens`, exactly like within-file merging.\n */\nexport function measureCrossFileDuplication(\n  files: CrossFileDuplicationSourceFile[],\n  options?: DuplicationOptions\n): CrossFileDuplicationMetrics {\n  const { minTokens, maxGapTokens } = resolveDuplicationOptions(options);\n  const candidates: SelectableCandidate[] = files.flatMap(({ file, candidates }, fileIndex) =>\n    candidates.map((candidate) => ({ ...candidate, regionBucket: fileIndex, file }))\n  );\n  // Pushed one by one: spreading the project-scale window-candidate array as call arguments\n  // overflows V8's argument limit (~124k) and crashes on Node, though Bun/JSC tolerates it.\n  for (const candidate of collectWindowCandidates(files, minTokens)) {\n    candidates.push(candidate);\n  }\n  const counted = selectMaximalGroups(\n    candidates,\n    spansMultipleFiles,\n    // File index and position break coverage ties deterministically.\n    (left, right) => left.regionBucket - right.regionBucket || left.startIndex - right.startIndex\n  );\n  const tokenOffsets = computeTokenOffsets(files, maxGapTokens);\n  return summarize(mergeGapAdjacentGroups([...counted.values()], tokenOffsets, maxGapTokens), files, tokenOffsets);\n}\n\n/** Repeated sub-windows of sibling statements matched across the whole project's files. */\nfunction collectWindowCandidates(files: CrossFileDuplicationSourceFile[], minTokens: number): SelectableCandidate[] {\n  const fileIndexByContext: number[] = [];\n  const contexts: SequenceWindowContext[] = [];\n  for (const [fileIndex, { tokens, containerStatements }] of files.entries()) {\n    if (tokens && containerStatements) {\n      fileIndexByContext.push(fileIndex);\n      contexts.push({ tokens, literalCountPrefix: buildLiteralCountPrefix(tokens), containers: containerStatements });\n    }\n  }\n  if (contexts.length < 2) {\n    return [];\n  }\n  return collectSequenceWindowCandidates(contexts, minTokens, true).flatMap(({ candidate, contextIndex }) => {\n    const fileIndex = fileIndexByContext[contextIndex];\n    const file = fileIndex === undefined ? undefined : files[fileIndex];\n    return fileIndex === undefined || file === undefined\n      ? []\n      : [{ ...candidate, regionBucket: fileIndex, file: file.file }];\n  });\n}\n\nfunction spansMultipleFiles(group: SelectableCandidate[]): boolean {\n  return group.length >= 2 && new Set(group.map((candidate) => candidate.regionBucket)).size >= 2;\n}\n\n/**\n * Per-file token offsets that map every file into one project-wide token index space: each file's\n * tokens are offset by more than `maxGapTokens` past the previous file's, so occurrences in\n * different files are never gap-adjacent and merged pairs always stay within one file.\n */\nfunction computeTokenOffsets(files: CrossFileDuplicationSourceFile[], maxGapTokens: number): number[] {\n  const tokenOffsets: number[] = [];\n  let offset = 0;\n  for (const { tokens, candidates } of files) {\n    tokenOffsets.push(offset);\n    // Accumulated in a loop: spreading a project-scale candidate array as call arguments would\n    // overflow V8's argument limit (~124k) and crash on Node.\n    let tokenCount = tokens?.length ?? 0;\n    if (!tokens) {\n      for (const candidate of candidates) {\n        tokenCount = Math.max(tokenCount, candidate.endTokenIndex);\n      }\n    }\n    offset += tokenCount + maxGapTokens + 1;\n  }\n  return tokenOffsets;\n}\n\n/** Reuses the within-file gapped (Type-3) merging in the project-wide token index space. */\nfunction mergeGapAdjacentGroups(\n  groups: SelectableCandidate[][],\n  tokenOffsets: number[],\n  maxGapTokens: number\n): CrossFileOccurrence[][] {\n  const occurrenceGroups = groups.map((group) =>\n    group\n      .map((candidate): CrossFileOccurrence => {\n        const start = candidate.startTokenIndex + (tokenOffsets[candidate.regionBucket] ?? 0);\n        const end = candidate.endTokenIndex + (tokenOffsets[candidate.regionBucket] ?? 0);\n        return {\n          file: candidate.file,\n          segments: [{ startTokenIndex: start, endTokenIndex: end }],\n          tokenCount: candidate.tokenCount,\n          startTokenIndex: start,\n          endTokenIndex: end,\n          startIndex: candidate.startIndex,\n          endIndex: candidate.endIndex,\n          startLine: candidate.startLine,\n          endLine: candidate.endLine,\n        };\n      })\n      .toSorted((left, right) => left.startTokenIndex - right.startTokenIndex)\n  );\n  return mergeAdjacentGroups(occurrenceGroups, maxGapTokens);\n}\n\nfunction summarize(\n  groups: CrossFileOccurrence[][],\n  files: CrossFileDuplicationSourceFile[],\n  tokenOffsets: number[]\n): CrossFileDuplicationMetrics {\n  const reported: CrossFileDuplicateBlockGroup[] = [];\n  // Accumulated in Maps: file names are arbitrary strings, and a plain object would read\n  // inherited properties for names like \"constructor\".\n  const groupCountByFile = new Map<string, number>();\n  const fileDataByName = new Map(\n    files.map((file, index) => [\n      file.file,\n      { tokens: file.tokens, codeLineNumbers: file.codeLineNumbers, offset: tokenOffsets[index] ?? 0 },\n    ])\n  );\n  const lineNumbersByFile = new Map<string, Set<number>>();\n  let duplicateBlockCount = 0;\n  for (const group of groups) {\n    // Mirrors within-file counting: each redundant occurrence contributes one count per matched\n    // fragment, gapped merging consolidates the grouping without halving the count, and spans a\n    // partial merge shares between a retained group and the merged group count once.\n    duplicateBlockCount += countRedundantFragments(group);\n    for (const occurrence of group) {\n      collectOccurrenceLines(occurrence, fileDataByName, lineNumbersByFile);\n    }\n    const occurrences = group\n      .map(({ file, startLine, endLine }) => ({ file, startLine, endLine }))\n      .toSorted((left, right) => left.file.localeCompare(right.file) || left.startLine - right.startLine);\n    const files = [...new Set(occurrences.map(({ file }) => file))];\n    for (const file of files) {\n      groupCountByFile.set(file, (groupCountByFile.get(file) ?? 0) + 1);\n    }\n    reported.push({ files, occurrences, tokenCount: group[0]?.tokenCount ?? 0 });\n  }\n  reported.sort(\n    (left, right) =>\n      right.tokenCount - left.tokenCount ||\n      (left.occurrences[0]?.file ?? '').localeCompare(right.occurrences[0]?.file ?? '') ||\n      (left.occurrences[0]?.startLine ?? 0) - (right.occurrences[0]?.startLine ?? 0)\n  );\n  return {\n    duplicateBlockCount,\n    duplicateBlockGroupCountByFile: Object.fromEntries(groupCountByFile),\n    duplicateLineNumbersByFile: Object.fromEntries(\n      [...lineNumbersByFile].map(([file, lines]) => [file, [...lines].toSorted((left, right) => left - right)])\n    ),\n    groups: reported,\n  };\n}\n\n/**\n * Adds the code lines an occurrence's matched tokens cover to its file's line set, mapping the\n * project-wide token segments back into the file's own token stream. A file that supplied only\n * candidates (no token stream) is skipped rather than approximated from the bounding line range,\n * which would include gap and comment/blank lines and break the field's exactness contract.\n */\nfunction collectOccurrenceLines(\n  occurrence: CrossFileOccurrence,\n  fileDataByName: Map<\n    string,\n    { tokens?: CrossFileDuplicationSourceFile['tokens']; codeLineNumbers?: Set<number>; offset: number }\n  >,\n  lineNumbersByFile: Map<string, Set<number>>\n): void {\n  const fileData = fileDataByName.get(occurrence.file);\n  if (!fileData?.tokens) {\n    return;\n  }\n  let lines = lineNumbersByFile.get(occurrence.file);\n  if (!lines) {\n    lines = new Set();\n    lineNumbersByFile.set(occurrence.file, lines);\n  }\n  for (const segment of occurrence.segments) {\n    collectSegmentLines(\n      {\n        startTokenIndex: segment.startTokenIndex - fileData.offset,\n        endTokenIndex: segment.endTokenIndex - fileData.offset,\n      },\n      fileData.tokens,\n      fileData.codeLineNumbers,\n      lines\n    );\n  }\n}\n"],"mappings":"wFAuEA,SAAgB,EACd,EACA,EAC6B,CAC7B,GAAM,CAAE,YAAW,gBAAiBA,EAAAA,0BAA0B,CAAO,EAC/D,EAAoC,EAAM,SAAS,CAAE,OAAM,cAAc,IAC7E,EAAW,IAAK,IAAe,CAAE,GAAG,EAAW,aAAc,EAAW,MAAK,EAAE,CACjF,EAGA,IAAK,IAAM,KAAa,EAAwB,EAAO,CAAS,EAC9D,EAAW,KAAK,CAAS,EAE3B,IAAM,EAAUC,EAAAA,oBACd,EACA,GAEC,EAAM,IAAU,EAAK,aAAe,EAAM,cAAgB,EAAK,WAAa,EAAM,UACrF,EACM,EAAe,EAAoB,EAAO,CAAY,EAC5D,OAAO,EAAU,EAAuB,CAAC,GAAG,EAAQ,OAAO,CAAC,EAAG,EAAc,CAAY,EAAG,EAAO,CAAY,CACjH,CAGA,SAAS,EAAwB,EAAyC,EAA0C,CAClH,IAAM,EAA+B,CAAC,EAChC,EAAoC,CAAC,EAC3C,IAAK,GAAM,CAAC,EAAW,CAAE,SAAQ,0BAA0B,EAAM,QAAQ,EACnE,GAAU,IACZ,EAAmB,KAAK,CAAS,EACjC,EAAS,KAAK,CAAE,SAAQ,mBAAoBC,EAAAA,wBAAwB,CAAM,EAAG,WAAY,CAAoB,CAAC,GAMlH,OAHI,EAAS,OAAS,EACb,CAAC,EAEHC,EAAAA,gCAAgC,EAAU,EAAW,EAAI,CAAC,CAAC,SAAS,CAAE,YAAW,kBAAmB,CACzG,IAAM,EAAY,EAAmB,GAC/B,EAAO,IAAc,IAAA,GAAY,IAAA,GAAY,EAAM,GACzD,OAAO,IAAc,IAAA,IAAa,IAAS,IAAA,GACvC,CAAC,EACD,CAAC,CAAE,GAAG,EAAW,aAAc,EAAW,KAAM,EAAK,IAAK,CAAC,CACjE,CAAC,CACH,CAEA,SAAS,EAAmB,EAAuC,CACjE,OAAO,EAAM,QAAU,GAAK,IAAI,IAAI,EAAM,IAAK,GAAc,EAAU,YAAY,CAAC,CAAC,CAAC,MAAQ,CAChG,CAOA,SAAS,EAAoB,EAAyC,EAAgC,CACpG,IAAM,EAAyB,CAAC,EAC5B,EAAS,EACb,IAAK,GAAM,CAAE,SAAQ,gBAAgB,EAAO,CAC1C,EAAa,KAAK,CAAM,EAGxB,IAAI,EAAa,GAAQ,QAAU,EACnC,GAAI,CAAC,EACH,IAAK,IAAM,KAAa,EACtB,EAAa,KAAK,IAAI,EAAY,EAAU,aAAa,EAG7D,GAAU,EAAa,EAAe,CACxC,CACA,OAAO,CACT,CAGA,SAAS,EACP,EACA,EACA,EACyB,CACzB,IAAM,EAAmB,EAAO,IAAK,GACnC,EACG,IAAK,GAAmC,CACvC,IAAM,EAAQ,EAAU,iBAAmB,EAAa,EAAU,eAAiB,GAC7E,EAAM,EAAU,eAAiB,EAAa,EAAU,eAAiB,GAC/E,MAAO,CACL,KAAM,EAAU,KAChB,SAAU,CAAC,CAAE,gBAAiB,EAAO,cAAe,CAAI,CAAC,EACzD,WAAY,EAAU,WACtB,gBAAiB,EACjB,cAAe,EACf,WAAY,EAAU,WACtB,SAAU,EAAU,SACpB,UAAW,EAAU,UACrB,QAAS,EAAU,OACrB,CACF,CAAC,CAAC,CACD,UAAU,EAAM,IAAU,EAAK,gBAAkB,EAAM,eAAe,CAC3E,EACA,OAAOC,EAAAA,oBAAoB,EAAkB,CAAY,CAC3D,CAEA,SAAS,EACP,EACA,EACA,EAC6B,CAC7B,IAAM,EAA2C,CAAC,EAG5C,EAAmB,IAAI,IACvB,EAAiB,IAAI,IACzB,EAAM,KAAK,EAAM,IAAU,CACzB,EAAK,KACL,CAAE,OAAQ,EAAK,OAAQ,gBAAiB,EAAK,gBAAiB,OAAQ,EAAa,IAAU,CAAE,CACjG,CAAC,CACH,EACM,EAAoB,IAAI,IAC1B,EAAsB,EAC1B,IAAK,IAAM,KAAS,EAAQ,CAI1B,GAAuBC,EAAAA,wBAAwB,CAAK,EACpD,IAAK,IAAM,KAAc,EACvB,EAAuB,EAAY,EAAgB,CAAiB,EAEtE,IAAM,EAAc,EACjB,KAAK,CAAE,OAAM,YAAW,cAAe,CAAE,OAAM,YAAW,SAAQ,EAAE,CAAC,CACrE,UAAU,EAAM,IAAU,EAAK,KAAK,cAAc,EAAM,IAAI,GAAK,EAAK,UAAY,EAAM,SAAS,EAC9F,EAAQ,CAAC,GAAG,IAAI,IAAI,EAAY,KAAK,CAAE,UAAW,CAAI,CAAC,CAAC,EAC9D,IAAK,IAAM,KAAQ,EACjB,EAAiB,IAAI,GAAO,EAAiB,IAAI,CAAI,GAAK,GAAK,CAAC,EAElE,EAAS,KAAK,CAAE,QAAO,cAAa,WAAY,EAAM,EAAE,EAAE,YAAc,CAAE,CAAC,CAC7E,CAOA,OANA,EAAS,MACN,EAAM,IACL,EAAM,WAAa,EAAK,aACvB,EAAK,YAAY,EAAE,EAAE,MAAQ,GAAA,CAAI,cAAc,EAAM,YAAY,EAAE,EAAE,MAAQ,EAAE,IAC/E,EAAK,YAAY,EAAE,EAAE,WAAa,IAAM,EAAM,YAAY,EAAE,EAAE,WAAa,EAChF,EACO,CACL,sBACA,+BAAgC,OAAO,YAAY,CAAgB,EACnE,2BAA4B,OAAO,YACjC,CAAC,GAAG,CAAiB,CAAC,CAAC,KAAK,CAAC,EAAM,KAAW,CAAC,EAAM,CAAC,GAAG,CAAK,CAAC,CAAC,UAAU,EAAM,IAAU,EAAO,CAAK,CAAC,CAAC,CAC1G,EACA,OAAQ,CACV,CACF,CAQA,SAAS,EACP,EACA,EAIA,EACM,CACN,IAAM,EAAW,EAAe,IAAI,EAAW,IAAI,EACnD,GAAI,CAAC,GAAU,OACb,OAEF,IAAI,EAAQ,EAAkB,IAAI,EAAW,IAAI,EAC5C,IACH,EAAQ,IAAI,IACZ,EAAkB,IAAI,EAAW,KAAM,CAAK,GAE9C,IAAK,IAAM,KAAW,EAAW,SAC/B,EAAA,oBACE,CACE,gBAAiB,EAAQ,gBAAkB,EAAS,OACpD,cAAe,EAAQ,cAAgB,EAAS,MAClD,EACA,EAAS,OACT,EAAS,gBACT,CACF,CAEJ"}
+{"version":3,"file":"crossFileDuplication.cjs","names":["resolveDuplicationOptions","selectMaximalGroups","buildLiteralCountPrefix","collectSequenceWindowCandidates","mergeAdjacentGroups","countRedundantFragments"],"sources":["../src/crossFileDuplication.ts"],"sourcesContent":["import { selectMaximalGroups, type SelectableRegion } from './duplicateSelection.js';\nimport {\n  buildLiteralCountPrefix,\n  collectSegmentLines,\n  collectSequenceWindowCandidates,\n  countRedundantFragments,\n  mergeAdjacentGroups,\n  resolveDuplicationOptions,\n  type CountedOccurrence,\n  type CrossFileDuplicateCandidate,\n  type CrossFileDuplicationFileData,\n  type SequenceWindowContext,\n} from './duplication.js';\nimport type { DuplicationOptions } from './types.js';\n\nexport interface CrossFileDuplicationSourceFile extends Partial<CrossFileDuplicationFileData> {\n  file: string;\n  candidates: CrossFileDuplicateCandidate[];\n}\n\nexport interface CrossFileDuplicateOccurrence {\n  endLine: number;\n  file: string;\n  startLine: number;\n}\n\nexport interface CrossFileDuplicateBlockGroup {\n  files: string[];\n  occurrences: CrossFileDuplicateOccurrence[];\n  /** Matched token count of one occurrence (all occurrences share it; gaps are not counted). */\n  tokenCount: number;\n}\n\nexport interface CrossFileDuplicationMetrics {\n  /** Number of redundant copies across all groups, counted per matched fragment like within-file. */\n  duplicateBlockCount: number;\n  /** Groups the file participates in, keyed by the file name passed in. */\n  duplicateBlockGroupCountByFile: Record<string, number>;\n  /**\n   * Per file, the 1-based code lines covered by matched tokens of its cross-file occurrences,\n   * sorted ascending. Exact like within-file duplicateLineNumbers: the unmatched gap of a merged\n   * clone and comment/blank lines inside an occurrence's bounding range are excluded (blank rows\n   * inside multi-row tokens only when the file supplied codeLineNumbers). A file that supplied\n   * only candidates (no `tokens`) has no entry — without its token stream the matched lines are\n   * unknowable, and an approximate bounding range would break this field's exactness.\n   */\n  duplicateLineNumbersByFile: Record<string, number[]>;\n  groups: CrossFileDuplicateBlockGroup[];\n}\n\ninterface SelectableCandidate extends CrossFileDuplicateCandidate, SelectableRegion {\n  regionBucket: number;\n  file: string;\n}\n\n/** A cross-file occurrence: a within-file occurrence in the project-wide token index space. */\ninterface CrossFileOccurrence extends CountedOccurrence {\n  file: string;\n}\n\n/**\n * Detects code regions duplicated across files. Per-file candidates (whole block subtrees and full\n * container runs, fingerprinted with the same normalization as within-file duplication) are joined\n * by a project-level window index over per-statement fingerprint sequences (CPD-style), so a\n * copy-pasted partial statement run embedded in different surrounding code is matched even though\n * no single file can know it repeats elsewhere. Candidates are grouped by fingerprint, and only\n * maximal, non-overlapping regions whose group spans at least two files are counted. Groups that\n * shrink to a single file during selection are shed — a within-file repeat is already reported by\n * that file's own duplication metrics. A copy nested inside a larger group's region (two files share\n * a whole function, a third file only a block of it) is reported with its group, so the third\n * file's copy still shows what it duplicates. Groups separated by a small token gap within each file then\n * merge into gapped (Type-3) clone groups under `maxGapTokens`, exactly like within-file merging.\n */\nexport function measureCrossFileDuplication(\n  files: CrossFileDuplicationSourceFile[],\n  options?: DuplicationOptions\n): CrossFileDuplicationMetrics {\n  const { minTokens, maxGapTokens } = resolveDuplicationOptions(options);\n  const candidates: SelectableCandidate[] = files.flatMap(({ file, candidates }, fileIndex) =>\n    candidates.map((candidate) => ({ ...candidate, regionBucket: fileIndex, file }))\n  );\n  // Pushed one by one: spreading the project-scale window-candidate array as call arguments\n  // overflows V8's argument limit (~124k) and crashes on Node, though Bun/JSC tolerates it.\n  for (const candidate of collectWindowCandidates(files, minTokens)) {\n    candidates.push(candidate);\n  }\n  const counted = selectMaximalGroups(\n    candidates,\n    spansMultipleFiles,\n    // File index and position break coverage ties deterministically.\n    (left, right) => left.regionBucket - right.regionBucket || left.startIndex - right.startIndex\n  );\n  const tokenOffsets = computeTokenOffsets(files, maxGapTokens);\n  return summarize(mergeGapAdjacentGroups([...counted.values()], tokenOffsets, maxGapTokens), files, tokenOffsets);\n}\n\n/** Repeated sub-windows of sibling statements matched across the whole project's files. */\nfunction collectWindowCandidates(files: CrossFileDuplicationSourceFile[], minTokens: number): SelectableCandidate[] {\n  const fileIndexByContext: number[] = [];\n  const contexts: SequenceWindowContext[] = [];\n  for (const [fileIndex, { tokens, containerStatements }] of files.entries()) {\n    if (tokens && containerStatements) {\n      fileIndexByContext.push(fileIndex);\n      contexts.push({ tokens, literalCountPrefix: buildLiteralCountPrefix(tokens), containers: containerStatements });\n    }\n  }\n  if (contexts.length < 2) {\n    return [];\n  }\n  return collectSequenceWindowCandidates(contexts, minTokens, true).flatMap(({ candidate, contextIndex }) => {\n    const fileIndex = fileIndexByContext[contextIndex];\n    const file = fileIndex === undefined ? undefined : files[fileIndex];\n    return fileIndex === undefined || file === undefined\n      ? []\n      : [{ ...candidate, regionBucket: fileIndex, file: file.file }];\n  });\n}\n\n/** A merged group is reported only while it still covers more than one file. */\nfunction spansMultipleFilesAfterMerge(group: CrossFileOccurrence[]): boolean {\n  return new Set(group.map((occurrence) => occurrence.file)).size >= 2;\n}\n\nfunction spansMultipleFiles(group: SelectableCandidate[]): boolean {\n  return group.length >= 2 && new Set(group.map((candidate) => candidate.regionBucket)).size >= 2;\n}\n\n/**\n * Per-file token offsets that map every file into one project-wide token index space: each file's\n * tokens are offset by more than `maxGapTokens` past the previous file's, so occurrences in\n * different files are never gap-adjacent and merged pairs always stay within one file.\n */\nfunction computeTokenOffsets(files: CrossFileDuplicationSourceFile[], maxGapTokens: number): number[] {\n  const tokenOffsets: number[] = [];\n  let offset = 0;\n  for (const { tokens, candidates } of files) {\n    tokenOffsets.push(offset);\n    // Accumulated in a loop: spreading a project-scale candidate array as call arguments would\n    // overflow V8's argument limit (~124k) and crash on Node.\n    let tokenCount = tokens?.length ?? 0;\n    if (!tokens) {\n      for (const candidate of candidates) {\n        tokenCount = Math.max(tokenCount, candidate.endTokenIndex);\n      }\n    }\n    offset += tokenCount + maxGapTokens + 1;\n  }\n  return tokenOffsets;\n}\n\n/** Reuses the within-file gapped (Type-3) merging in the project-wide token index space. */\nfunction mergeGapAdjacentGroups(\n  groups: SelectableCandidate[][],\n  tokenOffsets: number[],\n  maxGapTokens: number\n): CrossFileOccurrence[][] {\n  const occurrenceGroups = groups.map((group) =>\n    group\n      .map((candidate): CrossFileOccurrence => {\n        const start = candidate.startTokenIndex + (tokenOffsets[candidate.regionBucket] ?? 0);\n        const end = candidate.endTokenIndex + (tokenOffsets[candidate.regionBucket] ?? 0);\n        return {\n          file: candidate.file,\n          spanCountedElsewhere: candidate.nestedInLargerGroup,\n          nestedInLargerGroup: candidate.nestedInLargerGroup,\n          segments: [{ startTokenIndex: start, endTokenIndex: end }],\n          tokenCount: candidate.tokenCount,\n          startTokenIndex: start,\n          endTokenIndex: end,\n          startIndex: candidate.startIndex,\n          endIndex: candidate.endIndex,\n          startLine: candidate.startLine,\n          endLine: candidate.endLine,\n        };\n      })\n      .toSorted((left, right) => left.startTokenIndex - right.startTokenIndex)\n  );\n  return mergeAdjacentGroups(occurrenceGroups, maxGapTokens, spansMultipleFilesAfterMerge);\n}\n\nfunction summarize(\n  groups: CrossFileOccurrence[][],\n  files: CrossFileDuplicationSourceFile[],\n  tokenOffsets: number[]\n): CrossFileDuplicationMetrics {\n  const reported: CrossFileDuplicateBlockGroup[] = [];\n  // Accumulated in Maps: file names are arbitrary strings, and a plain object would read\n  // inherited properties for names like \"constructor\".\n  const groupCountByFile = new Map<string, number>();\n  const fileDataByName = new Map(\n    files.map((file, index) => [\n      file.file,\n      { tokens: file.tokens, codeLineNumbers: file.codeLineNumbers, offset: tokenOffsets[index] ?? 0 },\n    ])\n  );\n  const lineNumbersByFile = new Map<string, Set<number>>();\n  let duplicateBlockCount = 0;\n  for (const group of groups) {\n    // Mirrors within-file counting: each redundant occurrence contributes one count per matched\n    // fragment, gapped merging consolidates the grouping without halving the count, and spans a\n    // partial merge shares between a retained group and the merged group count once.\n    duplicateBlockCount += countRedundantFragments(group);\n    for (const occurrence of group) {\n      collectOccurrenceLines(occurrence, fileDataByName, lineNumbersByFile);\n    }\n    const occurrences = group\n      .map(({ file, startLine, endLine }) => ({ file, startLine, endLine }))\n      .toSorted((left, right) => left.file.localeCompare(right.file) || left.startLine - right.startLine);\n    const files = [...new Set(occurrences.map(({ file }) => file))];\n    for (const file of files) {\n      groupCountByFile.set(file, (groupCountByFile.get(file) ?? 0) + 1);\n    }\n    reported.push({ files, occurrences, tokenCount: group[0]?.tokenCount ?? 0 });\n  }\n  reported.sort(\n    (left, right) =>\n      right.tokenCount - left.tokenCount ||\n      (left.occurrences[0]?.file ?? '').localeCompare(right.occurrences[0]?.file ?? '') ||\n      (left.occurrences[0]?.startLine ?? 0) - (right.occurrences[0]?.startLine ?? 0)\n  );\n  return {\n    duplicateBlockCount,\n    duplicateBlockGroupCountByFile: Object.fromEntries(groupCountByFile),\n    duplicateLineNumbersByFile: Object.fromEntries(\n      [...lineNumbersByFile].map(([file, lines]) => [file, [...lines].toSorted((left, right) => left - right)])\n    ),\n    groups: reported,\n  };\n}\n\n/**\n * Adds the code lines an occurrence's matched tokens cover to its file's line set, mapping the\n * project-wide token segments back into the file's own token stream. A file that supplied only\n * candidates (no token stream) is skipped rather than approximated from the bounding line range,\n * which would include gap and comment/blank lines and break the field's exactness contract.\n */\nfunction collectOccurrenceLines(\n  occurrence: CrossFileOccurrence,\n  fileDataByName: Map<\n    string,\n    { tokens?: CrossFileDuplicationSourceFile['tokens']; codeLineNumbers?: Set<number>; offset: number }\n  >,\n  lineNumbersByFile: Map<string, Set<number>>\n): void {\n  const fileData = fileDataByName.get(occurrence.file);\n  if (!fileData?.tokens) {\n    return;\n  }\n  let lines = lineNumbersByFile.get(occurrence.file);\n  if (!lines) {\n    lines = new Set();\n    lineNumbersByFile.set(occurrence.file, lines);\n  }\n  for (const segment of occurrence.segments) {\n    collectSegmentLines(\n      {\n        startTokenIndex: segment.startTokenIndex - fileData.offset,\n        endTokenIndex: segment.endTokenIndex - fileData.offset,\n      },\n      fileData.tokens,\n      fileData.codeLineNumbers,\n      lines\n    );\n  }\n}\n"],"mappings":"wFAyEA,SAAgB,EACd,EACA,EAC6B,CAC7B,GAAM,CAAE,YAAW,gBAAiBA,EAAAA,0BAA0B,CAAO,EAC/D,EAAoC,EAAM,SAAS,CAAE,OAAM,cAAc,IAC7E,EAAW,IAAK,IAAe,CAAE,GAAG,EAAW,aAAc,EAAW,MAAK,EAAE,CACjF,EAGA,IAAK,IAAM,KAAa,EAAwB,EAAO,CAAS,EAC9D,EAAW,KAAK,CAAS,EAE3B,IAAM,EAAUC,EAAAA,oBACd,EACA,GAEC,EAAM,IAAU,EAAK,aAAe,EAAM,cAAgB,EAAK,WAAa,EAAM,UACrF,EACM,EAAe,EAAoB,EAAO,CAAY,EAC5D,OAAO,EAAU,EAAuB,CAAC,GAAG,EAAQ,OAAO,CAAC,EAAG,EAAc,CAAY,EAAG,EAAO,CAAY,CACjH,CAGA,SAAS,EAAwB,EAAyC,EAA0C,CAClH,IAAM,EAA+B,CAAC,EAChC,EAAoC,CAAC,EAC3C,IAAK,GAAM,CAAC,EAAW,CAAE,SAAQ,0BAA0B,EAAM,QAAQ,EACnE,GAAU,IACZ,EAAmB,KAAK,CAAS,EACjC,EAAS,KAAK,CAAE,SAAQ,mBAAoBC,EAAAA,wBAAwB,CAAM,EAAG,WAAY,CAAoB,CAAC,GAMlH,OAHI,EAAS,OAAS,EACb,CAAC,EAEHC,EAAAA,gCAAgC,EAAU,EAAW,EAAI,CAAC,CAAC,SAAS,CAAE,YAAW,kBAAmB,CACzG,IAAM,EAAY,EAAmB,GAC/B,EAAO,IAAc,IAAA,GAAY,IAAA,GAAY,EAAM,GACzD,OAAO,IAAc,IAAA,IAAa,IAAS,IAAA,GACvC,CAAC,EACD,CAAC,CAAE,GAAG,EAAW,aAAc,EAAW,KAAM,EAAK,IAAK,CAAC,CACjE,CAAC,CACH,CAGA,SAAS,EAA6B,EAAuC,CAC3E,OAAO,IAAI,IAAI,EAAM,IAAK,GAAe,EAAW,IAAI,CAAC,CAAC,CAAC,MAAQ,CACrE,CAEA,SAAS,EAAmB,EAAuC,CACjE,OAAO,EAAM,QAAU,GAAK,IAAI,IAAI,EAAM,IAAK,GAAc,EAAU,YAAY,CAAC,CAAC,CAAC,MAAQ,CAChG,CAOA,SAAS,EAAoB,EAAyC,EAAgC,CACpG,IAAM,EAAyB,CAAC,EAC5B,EAAS,EACb,IAAK,GAAM,CAAE,SAAQ,gBAAgB,EAAO,CAC1C,EAAa,KAAK,CAAM,EAGxB,IAAI,EAAa,GAAQ,QAAU,EACnC,GAAI,CAAC,EACH,IAAK,IAAM,KAAa,EACtB,EAAa,KAAK,IAAI,EAAY,EAAU,aAAa,EAG7D,GAAU,EAAa,EAAe,CACxC,CACA,OAAO,CACT,CAGA,SAAS,EACP,EACA,EACA,EACyB,CACzB,IAAM,EAAmB,EAAO,IAAK,GACnC,EACG,IAAK,GAAmC,CACvC,IAAM,EAAQ,EAAU,iBAAmB,EAAa,EAAU,eAAiB,GAC7E,EAAM,EAAU,eAAiB,EAAa,EAAU,eAAiB,GAC/E,MAAO,CACL,KAAM,EAAU,KAChB,qBAAsB,EAAU,oBAChC,oBAAqB,EAAU,oBAC/B,SAAU,CAAC,CAAE,gBAAiB,EAAO,cAAe,CAAI,CAAC,EACzD,WAAY,EAAU,WACtB,gBAAiB,EACjB,cAAe,EACf,WAAY,EAAU,WACtB,SAAU,EAAU,SACpB,UAAW,EAAU,UACrB,QAAS,EAAU,OACrB,CACF,CAAC,CAAC,CACD,UAAU,EAAM,IAAU,EAAK,gBAAkB,EAAM,eAAe,CAC3E,EACA,OAAOC,EAAAA,oBAAoB,EAAkB,EAAc,CAA4B,CACzF,CAEA,SAAS,EACP,EACA,EACA,EAC6B,CAC7B,IAAM,EAA2C,CAAC,EAG5C,EAAmB,IAAI,IACvB,EAAiB,IAAI,IACzB,EAAM,KAAK,EAAM,IAAU,CACzB,EAAK,KACL,CAAE,OAAQ,EAAK,OAAQ,gBAAiB,EAAK,gBAAiB,OAAQ,EAAa,IAAU,CAAE,CACjG,CAAC,CACH,EACM,EAAoB,IAAI,IAC1B,EAAsB,EAC1B,IAAK,IAAM,KAAS,EAAQ,CAI1B,GAAuBC,EAAAA,wBAAwB,CAAK,EACpD,IAAK,IAAM,KAAc,EACvB,EAAuB,EAAY,EAAgB,CAAiB,EAEtE,IAAM,EAAc,EACjB,KAAK,CAAE,OAAM,YAAW,cAAe,CAAE,OAAM,YAAW,SAAQ,EAAE,CAAC,CACrE,UAAU,EAAM,IAAU,EAAK,KAAK,cAAc,EAAM,IAAI,GAAK,EAAK,UAAY,EAAM,SAAS,EAC9F,EAAQ,CAAC,GAAG,IAAI,IAAI,EAAY,KAAK,CAAE,UAAW,CAAI,CAAC,CAAC,EAC9D,IAAK,IAAM,KAAQ,EACjB,EAAiB,IAAI,GAAO,EAAiB,IAAI,CAAI,GAAK,GAAK,CAAC,EAElE,EAAS,KAAK,CAAE,QAAO,cAAa,WAAY,EAAM,EAAE,EAAE,YAAc,CAAE,CAAC,CAC7E,CAOA,OANA,EAAS,MACN,EAAM,IACL,EAAM,WAAa,EAAK,aACvB,EAAK,YAAY,EAAE,EAAE,MAAQ,GAAA,CAAI,cAAc,EAAM,YAAY,EAAE,EAAE,MAAQ,EAAE,IAC/E,EAAK,YAAY,EAAE,EAAE,WAAa,IAAM,EAAM,YAAY,EAAE,EAAE,WAAa,EAChF,EACO,CACL,sBACA,+BAAgC,OAAO,YAAY,CAAgB,EACnE,2BAA4B,OAAO,YACjC,CAAC,GAAG,CAAiB,CAAC,CAAC,KAAK,CAAC,EAAM,KAAW,CAAC,EAAM,CAAC,GAAG,CAAK,CAAC,CAAC,UAAU,EAAM,IAAU,EAAO,CAAK,CAAC,CAAC,CAC1G,EACA,OAAQ,CACV,CACF,CAQA,SAAS,EACP,EACA,EAIA,EACM,CACN,IAAM,EAAW,EAAe,IAAI,EAAW,IAAI,EACnD,GAAI,CAAC,GAAU,OACb,OAEF,IAAI,EAAQ,EAAkB,IAAI,EAAW,IAAI,EAC5C,IACH,EAAQ,IAAI,IACZ,EAAkB,IAAI,EAAW,KAAM,CAAK,GAE9C,IAAK,IAAM,KAAW,EAAW,SAC/B,EAAA,oBACE,CACE,gBAAiB,EAAQ,gBAAkB,EAAS,OACpD,cAAe,EAAQ,cAAgB,EAAS,MAClD,EACA,EAAS,OACT,EAAS,gBACT,CACF,CAEJ"}

package/dist/crossFileDuplication.d.ts

@@ -39,7 +39,9 @@ export interface CrossFileDuplicationMetrics {
  * no single file can know it repeats elsewhere. Candidates are grouped by fingerprint, and only
  * maximal, non-overlapping regions whose group spans at least two files are counted. Groups that
  * shrink to a single file during selection are shed — a within-file repeat is already reported by
- * that file's own duplication metrics. Groups separated by a small token gap within each file then
+ * that file's own duplication metrics. A copy nested inside a larger group's region (two files share
+ * a whole function, a third file only a block of it) is reported with its group, so the third
+ * file's copy still shows what it duplicates. Groups separated by a small token gap within each file then
  * merge into gapped (Type-3) clone groups under `maxGapTokens`, exactly like within-file merging.
  */
 export declare function measureCrossFileDuplication(files: CrossFileDuplicationSourceFile[], options?: DuplicationOptions): CrossFileDuplicationMetrics;

package/dist/crossFileDuplication.js

@@ -1,2 +1,2 @@
-import{selectMaximalGroups as e}from"./duplicateSelection.js";import{buildLiteralCountPrefix as t,collectSegmentLines as n,collectSequenceWindowCandidates as r,countRedundantFragments as i,mergeAdjacentGroups as a,resolveDuplicationOptions as o}from"./duplication.js";function s(t,n){let{minTokens:r,maxGapTokens:i}=o(n),a=t.flatMap(({file:e,candidates:t},n)=>t.map(t=>({...t,regionBucket:n,file:e})));for(let e of c(t,r))a.push(e);let s=e(a,l,(e,t)=>e.regionBucket-t.regionBucket||e.startIndex-t.startIndex),p=u(t,i);return f(d([...s.values()],p,i),t,p)}function c(e,n){let i=[],a=[];for(let[n,{tokens:r,containerStatements:o}]of e.entries())r&&o&&(i.push(n),a.push({tokens:r,literalCountPrefix:t(r),containers:o}));return a.length<2?[]:r(a,n,!0).flatMap(({candidate:t,contextIndex:n})=>{let r=i[n],a=r===void 0?void 0:e[r];return r===void 0||a===void 0?[]:[{...t,regionBucket:r,file:a.file}]})}function l(e){return e.length>=2&&new Set(e.map(e=>e.regionBucket)).size>=2}function u(e,t){let n=[],r=0;for(let{tokens:i,candidates:a}of e){n.push(r);let e=i?.length??0;if(!i)for(let t of a)e=Math.max(e,t.endTokenIndex);r+=e+t+1}return n}function d(e,t,n){let r=e.map(e=>e.map(e=>{let n=e.startTokenIndex+(t[e.regionBucket]??0),r=e.endTokenIndex+(t[e.regionBucket]??0);return{file:e.file,segments:[{startTokenIndex:n,endTokenIndex:r}],tokenCount:e.tokenCount,startTokenIndex:n,endTokenIndex:r,startIndex:e.startIndex,endIndex:e.endIndex,startLine:e.startLine,endLine:e.endLine}}).toSorted((e,t)=>e.startTokenIndex-t.startTokenIndex));return a(r,n)}function f(e,t,n){let r=[],a=new Map,o=new Map(t.map((e,t)=>[e.file,{tokens:e.tokens,codeLineNumbers:e.codeLineNumbers,offset:n[t]??0}])),s=new Map,c=0;for(let t of e){c+=i(t);for(let e of t)p(e,o,s);let e=t.map(({file:e,startLine:t,endLine:n})=>({file:e,startLine:t,endLine:n})).toSorted((e,t)=>e.file.localeCompare(t.file)||e.startLine-t.startLine),n=[...new Set(e.map(({file:e})=>e))];for(let e of n)a.set(e,(a.get(e)??0)+1);r.push({files:n,occurrences:e,tokenCount:t[0]?.tokenCount??0})}return r.sort((e,t)=>t.tokenCount-e.tokenCount||(e.occurrences[0]?.file??``).localeCompare(t.occurrences[0]?.file??``)||(e.occurrences[0]?.startLine??0)-(t.occurrences[0]?.startLine??0)),{duplicateBlockCount:c,duplicateBlockGroupCountByFile:Object.fromEntries(a),duplicateLineNumbersByFile:Object.fromEntries([...s].map(([e,t])=>[e,[...t].toSorted((e,t)=>e-t)])),groups:r}}function p(e,t,r){let i=t.get(e.file);if(!i?.tokens)return;let a=r.get(e.file);a||(a=new Set,r.set(e.file,a));for(let t of e.segments)n({startTokenIndex:t.startTokenIndex-i.offset,endTokenIndex:t.endTokenIndex-i.offset},i.tokens,i.codeLineNumbers,a)}export{s as measureCrossFileDuplication};
+import{selectMaximalGroups as e}from"./duplicateSelection.js";import{buildLiteralCountPrefix as t,collectSegmentLines as n,collectSequenceWindowCandidates as r,countRedundantFragments as i,mergeAdjacentGroups as a,resolveDuplicationOptions as o}from"./duplication.js";function s(t,n){let{minTokens:r,maxGapTokens:i}=o(n),a=t.flatMap(({file:e,candidates:t},n)=>t.map(t=>({...t,regionBucket:n,file:e})));for(let e of c(t,r))a.push(e);let s=e(a,u,(e,t)=>e.regionBucket-t.regionBucket||e.startIndex-t.startIndex),l=d(t,i);return p(f([...s.values()],l,i),t,l)}function c(e,n){let i=[],a=[];for(let[n,{tokens:r,containerStatements:o}]of e.entries())r&&o&&(i.push(n),a.push({tokens:r,literalCountPrefix:t(r),containers:o}));return a.length<2?[]:r(a,n,!0).flatMap(({candidate:t,contextIndex:n})=>{let r=i[n],a=r===void 0?void 0:e[r];return r===void 0||a===void 0?[]:[{...t,regionBucket:r,file:a.file}]})}function l(e){return new Set(e.map(e=>e.file)).size>=2}function u(e){return e.length>=2&&new Set(e.map(e=>e.regionBucket)).size>=2}function d(e,t){let n=[],r=0;for(let{tokens:i,candidates:a}of e){n.push(r);let e=i?.length??0;if(!i)for(let t of a)e=Math.max(e,t.endTokenIndex);r+=e+t+1}return n}function f(e,t,n){let r=e.map(e=>e.map(e=>{let n=e.startTokenIndex+(t[e.regionBucket]??0),r=e.endTokenIndex+(t[e.regionBucket]??0);return{file:e.file,spanCountedElsewhere:e.nestedInLargerGroup,nestedInLargerGroup:e.nestedInLargerGroup,segments:[{startTokenIndex:n,endTokenIndex:r}],tokenCount:e.tokenCount,startTokenIndex:n,endTokenIndex:r,startIndex:e.startIndex,endIndex:e.endIndex,startLine:e.startLine,endLine:e.endLine}}).toSorted((e,t)=>e.startTokenIndex-t.startTokenIndex));return a(r,n,l)}function p(e,t,n){let r=[],a=new Map,o=new Map(t.map((e,t)=>[e.file,{tokens:e.tokens,codeLineNumbers:e.codeLineNumbers,offset:n[t]??0}])),s=new Map,c=0;for(let t of e){c+=i(t);for(let e of t)m(e,o,s);let e=t.map(({file:e,startLine:t,endLine:n})=>({file:e,startLine:t,endLine:n})).toSorted((e,t)=>e.file.localeCompare(t.file)||e.startLine-t.startLine),n=[...new Set(e.map(({file:e})=>e))];for(let e of n)a.set(e,(a.get(e)??0)+1);r.push({files:n,occurrences:e,tokenCount:t[0]?.tokenCount??0})}return r.sort((e,t)=>t.tokenCount-e.tokenCount||(e.occurrences[0]?.file??``).localeCompare(t.occurrences[0]?.file??``)||(e.occurrences[0]?.startLine??0)-(t.occurrences[0]?.startLine??0)),{duplicateBlockCount:c,duplicateBlockGroupCountByFile:Object.fromEntries(a),duplicateLineNumbersByFile:Object.fromEntries([...s].map(([e,t])=>[e,[...t].toSorted((e,t)=>e-t)])),groups:r}}function m(e,t,r){let i=t.get(e.file);if(!i?.tokens)return;let a=r.get(e.file);a||(a=new Set,r.set(e.file,a));for(let t of e.segments)n({startTokenIndex:t.startTokenIndex-i.offset,endTokenIndex:t.endTokenIndex-i.offset},i.tokens,i.codeLineNumbers,a)}export{s as measureCrossFileDuplication};
 //# sourceMappingURL=crossFileDuplication.js.map

package/dist/crossFileDuplication.js.map

@@ -1 +1 @@
-{"version":3,"file":"crossFileDuplication.js","names":[],"sources":["../src/crossFileDuplication.ts"],"sourcesContent":["import { selectMaximalGroups } from './duplicateSelection.js';\nimport {\n  buildLiteralCountPrefix,\n  collectSegmentLines,\n  collectSequenceWindowCandidates,\n  countRedundantFragments,\n  mergeAdjacentGroups,\n  resolveDuplicationOptions,\n  type CountedOccurrence,\n  type CrossFileDuplicateCandidate,\n  type CrossFileDuplicationFileData,\n  type SequenceWindowContext,\n} from './duplication.js';\nimport type { DuplicationOptions } from './types.js';\n\nexport interface CrossFileDuplicationSourceFile extends Partial<CrossFileDuplicationFileData> {\n  file: string;\n  candidates: CrossFileDuplicateCandidate[];\n}\n\nexport interface CrossFileDuplicateOccurrence {\n  endLine: number;\n  file: string;\n  startLine: number;\n}\n\nexport interface CrossFileDuplicateBlockGroup {\n  files: string[];\n  occurrences: CrossFileDuplicateOccurrence[];\n  /** Matched token count of one occurrence (all occurrences share it; gaps are not counted). */\n  tokenCount: number;\n}\n\nexport interface CrossFileDuplicationMetrics {\n  /** Number of redundant copies across all groups, counted per matched fragment like within-file. */\n  duplicateBlockCount: number;\n  /** Groups the file participates in, keyed by the file name passed in. */\n  duplicateBlockGroupCountByFile: Record<string, number>;\n  /**\n   * Per file, the 1-based code lines covered by matched tokens of its cross-file occurrences,\n   * sorted ascending. Exact like within-file duplicateLineNumbers: the unmatched gap of a merged\n   * clone and comment/blank lines inside an occurrence's bounding range are excluded (blank rows\n   * inside multi-row tokens only when the file supplied codeLineNumbers). A file that supplied\n   * only candidates (no `tokens`) has no entry — without its token stream the matched lines are\n   * unknowable, and an approximate bounding range would break this field's exactness.\n   */\n  duplicateLineNumbersByFile: Record<string, number[]>;\n  groups: CrossFileDuplicateBlockGroup[];\n}\n\ninterface SelectableCandidate extends CrossFileDuplicateCandidate {\n  regionBucket: number;\n  file: string;\n}\n\n/** A cross-file occurrence: a within-file occurrence in the project-wide token index space. */\ninterface CrossFileOccurrence extends CountedOccurrence {\n  file: string;\n}\n\n/**\n * Detects code regions duplicated across files. Per-file candidates (whole block subtrees and full\n * container runs, fingerprinted with the same normalization as within-file duplication) are joined\n * by a project-level window index over per-statement fingerprint sequences (CPD-style), so a\n * copy-pasted partial statement run embedded in different surrounding code is matched even though\n * no single file can know it repeats elsewhere. Candidates are grouped by fingerprint, and only\n * maximal, non-overlapping regions whose group spans at least two files are counted. Groups that\n * shrink to a single file during selection are shed — a within-file repeat is already reported by\n * that file's own duplication metrics. Groups separated by a small token gap within each file then\n * merge into gapped (Type-3) clone groups under `maxGapTokens`, exactly like within-file merging.\n */\nexport function measureCrossFileDuplication(\n  files: CrossFileDuplicationSourceFile[],\n  options?: DuplicationOptions\n): CrossFileDuplicationMetrics {\n  const { minTokens, maxGapTokens } = resolveDuplicationOptions(options);\n  const candidates: SelectableCandidate[] = files.flatMap(({ file, candidates }, fileIndex) =>\n    candidates.map((candidate) => ({ ...candidate, regionBucket: fileIndex, file }))\n  );\n  // Pushed one by one: spreading the project-scale window-candidate array as call arguments\n  // overflows V8's argument limit (~124k) and crashes on Node, though Bun/JSC tolerates it.\n  for (const candidate of collectWindowCandidates(files, minTokens)) {\n    candidates.push(candidate);\n  }\n  const counted = selectMaximalGroups(\n    candidates,\n    spansMultipleFiles,\n    // File index and position break coverage ties deterministically.\n    (left, right) => left.regionBucket - right.regionBucket || left.startIndex - right.startIndex\n  );\n  const tokenOffsets = computeTokenOffsets(files, maxGapTokens);\n  return summarize(mergeGapAdjacentGroups([...counted.values()], tokenOffsets, maxGapTokens), files, tokenOffsets);\n}\n\n/** Repeated sub-windows of sibling statements matched across the whole project's files. */\nfunction collectWindowCandidates(files: CrossFileDuplicationSourceFile[], minTokens: number): SelectableCandidate[] {\n  const fileIndexByContext: number[] = [];\n  const contexts: SequenceWindowContext[] = [];\n  for (const [fileIndex, { tokens, containerStatements }] of files.entries()) {\n    if (tokens && containerStatements) {\n      fileIndexByContext.push(fileIndex);\n      contexts.push({ tokens, literalCountPrefix: buildLiteralCountPrefix(tokens), containers: containerStatements });\n    }\n  }\n  if (contexts.length < 2) {\n    return [];\n  }\n  return collectSequenceWindowCandidates(contexts, minTokens, true).flatMap(({ candidate, contextIndex }) => {\n    const fileIndex = fileIndexByContext[contextIndex];\n    const file = fileIndex === undefined ? undefined : files[fileIndex];\n    return fileIndex === undefined || file === undefined\n      ? []\n      : [{ ...candidate, regionBucket: fileIndex, file: file.file }];\n  });\n}\n\nfunction spansMultipleFiles(group: SelectableCandidate[]): boolean {\n  return group.length >= 2 && new Set(group.map((candidate) => candidate.regionBucket)).size >= 2;\n}\n\n/**\n * Per-file token offsets that map every file into one project-wide token index space: each file's\n * tokens are offset by more than `maxGapTokens` past the previous file's, so occurrences in\n * different files are never gap-adjacent and merged pairs always stay within one file.\n */\nfunction computeTokenOffsets(files: CrossFileDuplicationSourceFile[], maxGapTokens: number): number[] {\n  const tokenOffsets: number[] = [];\n  let offset = 0;\n  for (const { tokens, candidates } of files) {\n    tokenOffsets.push(offset);\n    // Accumulated in a loop: spreading a project-scale candidate array as call arguments would\n    // overflow V8's argument limit (~124k) and crash on Node.\n    let tokenCount = tokens?.length ?? 0;\n    if (!tokens) {\n      for (const candidate of candidates) {\n        tokenCount = Math.max(tokenCount, candidate.endTokenIndex);\n      }\n    }\n    offset += tokenCount + maxGapTokens + 1;\n  }\n  return tokenOffsets;\n}\n\n/** Reuses the within-file gapped (Type-3) merging in the project-wide token index space. */\nfunction mergeGapAdjacentGroups(\n  groups: SelectableCandidate[][],\n  tokenOffsets: number[],\n  maxGapTokens: number\n): CrossFileOccurrence[][] {\n  const occurrenceGroups = groups.map((group) =>\n    group\n      .map((candidate): CrossFileOccurrence => {\n        const start = candidate.startTokenIndex + (tokenOffsets[candidate.regionBucket] ?? 0);\n        const end = candidate.endTokenIndex + (tokenOffsets[candidate.regionBucket] ?? 0);\n        return {\n          file: candidate.file,\n          segments: [{ startTokenIndex: start, endTokenIndex: end }],\n          tokenCount: candidate.tokenCount,\n          startTokenIndex: start,\n          endTokenIndex: end,\n          startIndex: candidate.startIndex,\n          endIndex: candidate.endIndex,\n          startLine: candidate.startLine,\n          endLine: candidate.endLine,\n        };\n      })\n      .toSorted((left, right) => left.startTokenIndex - right.startTokenIndex)\n  );\n  return mergeAdjacentGroups(occurrenceGroups, maxGapTokens);\n}\n\nfunction summarize(\n  groups: CrossFileOccurrence[][],\n  files: CrossFileDuplicationSourceFile[],\n  tokenOffsets: number[]\n): CrossFileDuplicationMetrics {\n  const reported: CrossFileDuplicateBlockGroup[] = [];\n  // Accumulated in Maps: file names are arbitrary strings, and a plain object would read\n  // inherited properties for names like \"constructor\".\n  const groupCountByFile = new Map<string, number>();\n  const fileDataByName = new Map(\n    files.map((file, index) => [\n      file.file,\n      { tokens: file.tokens, codeLineNumbers: file.codeLineNumbers, offset: tokenOffsets[index] ?? 0 },\n    ])\n  );\n  const lineNumbersByFile = new Map<string, Set<number>>();\n  let duplicateBlockCount = 0;\n  for (const group of groups) {\n    // Mirrors within-file counting: each redundant occurrence contributes one count per matched\n    // fragment, gapped merging consolidates the grouping without halving the count, and spans a\n    // partial merge shares between a retained group and the merged group count once.\n    duplicateBlockCount += countRedundantFragments(group);\n    for (const occurrence of group) {\n      collectOccurrenceLines(occurrence, fileDataByName, lineNumbersByFile);\n    }\n    const occurrences = group\n      .map(({ file, startLine, endLine }) => ({ file, startLine, endLine }))\n      .toSorted((left, right) => left.file.localeCompare(right.file) || left.startLine - right.startLine);\n    const files = [...new Set(occurrences.map(({ file }) => file))];\n    for (const file of files) {\n      groupCountByFile.set(file, (groupCountByFile.get(file) ?? 0) + 1);\n    }\n    reported.push({ files, occurrences, tokenCount: group[0]?.tokenCount ?? 0 });\n  }\n  reported.sort(\n    (left, right) =>\n      right.tokenCount - left.tokenCount ||\n      (left.occurrences[0]?.file ?? '').localeCompare(right.occurrences[0]?.file ?? '') ||\n      (left.occurrences[0]?.startLine ?? 0) - (right.occurrences[0]?.startLine ?? 0)\n  );\n  return {\n    duplicateBlockCount,\n    duplicateBlockGroupCountByFile: Object.fromEntries(groupCountByFile),\n    duplicateLineNumbersByFile: Object.fromEntries(\n      [...lineNumbersByFile].map(([file, lines]) => [file, [...lines].toSorted((left, right) => left - right)])\n    ),\n    groups: reported,\n  };\n}\n\n/**\n * Adds the code lines an occurrence's matched tokens cover to its file's line set, mapping the\n * project-wide token segments back into the file's own token stream. A file that supplied only\n * candidates (no token stream) is skipped rather than approximated from the bounding line range,\n * which would include gap and comment/blank lines and break the field's exactness contract.\n */\nfunction collectOccurrenceLines(\n  occurrence: CrossFileOccurrence,\n  fileDataByName: Map<\n    string,\n    { tokens?: CrossFileDuplicationSourceFile['tokens']; codeLineNumbers?: Set<number>; offset: number }\n  >,\n  lineNumbersByFile: Map<string, Set<number>>\n): void {\n  const fileData = fileDataByName.get(occurrence.file);\n  if (!fileData?.tokens) {\n    return;\n  }\n  let lines = lineNumbersByFile.get(occurrence.file);\n  if (!lines) {\n    lines = new Set();\n    lineNumbersByFile.set(occurrence.file, lines);\n  }\n  for (const segment of occurrence.segments) {\n    collectSegmentLines(\n      {\n        startTokenIndex: segment.startTokenIndex - fileData.offset,\n        endTokenIndex: segment.endTokenIndex - fileData.offset,\n      },\n      fileData.tokens,\n      fileData.codeLineNumbers,\n      lines\n    );\n  }\n}\n"],"mappings":"4QAuEA,SAAgB,EACd,EACA,EAC6B,CAC7B,GAAM,CAAE,YAAW,gBAAiB,EAA0B,CAAO,EAC/D,EAAoC,EAAM,SAAS,CAAE,OAAM,cAAc,IAC7E,EAAW,IAAK,IAAe,CAAE,GAAG,EAAW,aAAc,EAAW,MAAK,EAAE,CACjF,EAGA,IAAK,IAAM,KAAa,EAAwB,EAAO,CAAS,EAC9D,EAAW,KAAK,CAAS,EAE3B,IAAM,EAAU,EACd,EACA,GAEC,EAAM,IAAU,EAAK,aAAe,EAAM,cAAgB,EAAK,WAAa,EAAM,UACrF,EACM,EAAe,EAAoB,EAAO,CAAY,EAC5D,OAAO,EAAU,EAAuB,CAAC,GAAG,EAAQ,OAAO,CAAC,EAAG,EAAc,CAAY,EAAG,EAAO,CAAY,CACjH,CAGA,SAAS,EAAwB,EAAyC,EAA0C,CAClH,IAAM,EAA+B,CAAC,EAChC,EAAoC,CAAC,EAC3C,IAAK,GAAM,CAAC,EAAW,CAAE,SAAQ,0BAA0B,EAAM,QAAQ,EACnE,GAAU,IACZ,EAAmB,KAAK,CAAS,EACjC,EAAS,KAAK,CAAE,SAAQ,mBAAoB,EAAwB,CAAM,EAAG,WAAY,CAAoB,CAAC,GAMlH,OAHI,EAAS,OAAS,EACb,CAAC,EAEH,EAAgC,EAAU,EAAW,EAAI,CAAC,CAAC,SAAS,CAAE,YAAW,kBAAmB,CACzG,IAAM,EAAY,EAAmB,GAC/B,EAAO,IAAc,IAAA,GAAY,IAAA,GAAY,EAAM,GACzD,OAAO,IAAc,IAAA,IAAa,IAAS,IAAA,GACvC,CAAC,EACD,CAAC,CAAE,GAAG,EAAW,aAAc,EAAW,KAAM,EAAK,IAAK,CAAC,CACjE,CAAC,CACH,CAEA,SAAS,EAAmB,EAAuC,CACjE,OAAO,EAAM,QAAU,GAAK,IAAI,IAAI,EAAM,IAAK,GAAc,EAAU,YAAY,CAAC,CAAC,CAAC,MAAQ,CAChG,CAOA,SAAS,EAAoB,EAAyC,EAAgC,CACpG,IAAM,EAAyB,CAAC,EAC5B,EAAS,EACb,IAAK,GAAM,CAAE,SAAQ,gBAAgB,EAAO,CAC1C,EAAa,KAAK,CAAM,EAGxB,IAAI,EAAa,GAAQ,QAAU,EACnC,GAAI,CAAC,EACH,IAAK,IAAM,KAAa,EACtB,EAAa,KAAK,IAAI,EAAY,EAAU,aAAa,EAG7D,GAAU,EAAa,EAAe,CACxC,CACA,OAAO,CACT,CAGA,SAAS,EACP,EACA,EACA,EACyB,CACzB,IAAM,EAAmB,EAAO,IAAK,GACnC,EACG,IAAK,GAAmC,CACvC,IAAM,EAAQ,EAAU,iBAAmB,EAAa,EAAU,eAAiB,GAC7E,EAAM,EAAU,eAAiB,EAAa,EAAU,eAAiB,GAC/E,MAAO,CACL,KAAM,EAAU,KAChB,SAAU,CAAC,CAAE,gBAAiB,EAAO,cAAe,CAAI,CAAC,EACzD,WAAY,EAAU,WACtB,gBAAiB,EACjB,cAAe,EACf,WAAY,EAAU,WACtB,SAAU,EAAU,SACpB,UAAW,EAAU,UACrB,QAAS,EAAU,OACrB,CACF,CAAC,CAAC,CACD,UAAU,EAAM,IAAU,EAAK,gBAAkB,EAAM,eAAe,CAC3E,EACA,OAAO,EAAoB,EAAkB,CAAY,CAC3D,CAEA,SAAS,EACP,EACA,EACA,EAC6B,CAC7B,IAAM,EAA2C,CAAC,EAG5C,EAAmB,IAAI,IACvB,EAAiB,IAAI,IACzB,EAAM,KAAK,EAAM,IAAU,CACzB,EAAK,KACL,CAAE,OAAQ,EAAK,OAAQ,gBAAiB,EAAK,gBAAiB,OAAQ,EAAa,IAAU,CAAE,CACjG,CAAC,CACH,EACM,EAAoB,IAAI,IAC1B,EAAsB,EAC1B,IAAK,IAAM,KAAS,EAAQ,CAI1B,GAAuB,EAAwB,CAAK,EACpD,IAAK,IAAM,KAAc,EACvB,EAAuB,EAAY,EAAgB,CAAiB,EAEtE,IAAM,EAAc,EACjB,KAAK,CAAE,OAAM,YAAW,cAAe,CAAE,OAAM,YAAW,SAAQ,EAAE,CAAC,CACrE,UAAU,EAAM,IAAU,EAAK,KAAK,cAAc,EAAM,IAAI,GAAK,EAAK,UAAY,EAAM,SAAS,EAC9F,EAAQ,CAAC,GAAG,IAAI,IAAI,EAAY,KAAK,CAAE,UAAW,CAAI,CAAC,CAAC,EAC9D,IAAK,IAAM,KAAQ,EACjB,EAAiB,IAAI,GAAO,EAAiB,IAAI,CAAI,GAAK,GAAK,CAAC,EAElE,EAAS,KAAK,CAAE,QAAO,cAAa,WAAY,EAAM,EAAE,EAAE,YAAc,CAAE,CAAC,CAC7E,CAOA,OANA,EAAS,MACN,EAAM,IACL,EAAM,WAAa,EAAK,aACvB,EAAK,YAAY,EAAE,EAAE,MAAQ,GAAA,CAAI,cAAc,EAAM,YAAY,EAAE,EAAE,MAAQ,EAAE,IAC/E,EAAK,YAAY,EAAE,EAAE,WAAa,IAAM,EAAM,YAAY,EAAE,EAAE,WAAa,EAChF,EACO,CACL,sBACA,+BAAgC,OAAO,YAAY,CAAgB,EACnE,2BAA4B,OAAO,YACjC,CAAC,GAAG,CAAiB,CAAC,CAAC,KAAK,CAAC,EAAM,KAAW,CAAC,EAAM,CAAC,GAAG,CAAK,CAAC,CAAC,UAAU,EAAM,IAAU,EAAO,CAAK,CAAC,CAAC,CAC1G,EACA,OAAQ,CACV,CACF,CAQA,SAAS,EACP,EACA,EAIA,EACM,CACN,IAAM,EAAW,EAAe,IAAI,EAAW,IAAI,EACnD,GAAI,CAAC,GAAU,OACb,OAEF,IAAI,EAAQ,EAAkB,IAAI,EAAW,IAAI,EAC5C,IACH,EAAQ,IAAI,IACZ,EAAkB,IAAI,EAAW,KAAM,CAAK,GAE9C,IAAK,IAAM,KAAW,EAAW,SAC/B,EACE,CACE,gBAAiB,EAAQ,gBAAkB,EAAS,OACpD,cAAe,EAAQ,cAAgB,EAAS,MAClD,EACA,EAAS,OACT,EAAS,gBACT,CACF,CAEJ"}
+{"version":3,"file":"crossFileDuplication.js","names":[],"sources":["../src/crossFileDuplication.ts"],"sourcesContent":["import { selectMaximalGroups, type SelectableRegion } from './duplicateSelection.js';\nimport {\n  buildLiteralCountPrefix,\n  collectSegmentLines,\n  collectSequenceWindowCandidates,\n  countRedundantFragments,\n  mergeAdjacentGroups,\n  resolveDuplicationOptions,\n  type CountedOccurrence,\n  type CrossFileDuplicateCandidate,\n  type CrossFileDuplicationFileData,\n  type SequenceWindowContext,\n} from './duplication.js';\nimport type { DuplicationOptions } from './types.js';\n\nexport interface CrossFileDuplicationSourceFile extends Partial<CrossFileDuplicationFileData> {\n  file: string;\n  candidates: CrossFileDuplicateCandidate[];\n}\n\nexport interface CrossFileDuplicateOccurrence {\n  endLine: number;\n  file: string;\n  startLine: number;\n}\n\nexport interface CrossFileDuplicateBlockGroup {\n  files: string[];\n  occurrences: CrossFileDuplicateOccurrence[];\n  /** Matched token count of one occurrence (all occurrences share it; gaps are not counted). */\n  tokenCount: number;\n}\n\nexport interface CrossFileDuplicationMetrics {\n  /** Number of redundant copies across all groups, counted per matched fragment like within-file. */\n  duplicateBlockCount: number;\n  /** Groups the file participates in, keyed by the file name passed in. */\n  duplicateBlockGroupCountByFile: Record<string, number>;\n  /**\n   * Per file, the 1-based code lines covered by matched tokens of its cross-file occurrences,\n   * sorted ascending. Exact like within-file duplicateLineNumbers: the unmatched gap of a merged\n   * clone and comment/blank lines inside an occurrence's bounding range are excluded (blank rows\n   * inside multi-row tokens only when the file supplied codeLineNumbers). A file that supplied\n   * only candidates (no `tokens`) has no entry — without its token stream the matched lines are\n   * unknowable, and an approximate bounding range would break this field's exactness.\n   */\n  duplicateLineNumbersByFile: Record<string, number[]>;\n  groups: CrossFileDuplicateBlockGroup[];\n}\n\ninterface SelectableCandidate extends CrossFileDuplicateCandidate, SelectableRegion {\n  regionBucket: number;\n  file: string;\n}\n\n/** A cross-file occurrence: a within-file occurrence in the project-wide token index space. */\ninterface CrossFileOccurrence extends CountedOccurrence {\n  file: string;\n}\n\n/**\n * Detects code regions duplicated across files. Per-file candidates (whole block subtrees and full\n * container runs, fingerprinted with the same normalization as within-file duplication) are joined\n * by a project-level window index over per-statement fingerprint sequences (CPD-style), so a\n * copy-pasted partial statement run embedded in different surrounding code is matched even though\n * no single file can know it repeats elsewhere. Candidates are grouped by fingerprint, and only\n * maximal, non-overlapping regions whose group spans at least two files are counted. Groups that\n * shrink to a single file during selection are shed — a within-file repeat is already reported by\n * that file's own duplication metrics. A copy nested inside a larger group's region (two files share\n * a whole function, a third file only a block of it) is reported with its group, so the third\n * file's copy still shows what it duplicates. Groups separated by a small token gap within each file then\n * merge into gapped (Type-3) clone groups under `maxGapTokens`, exactly like within-file merging.\n */\nexport function measureCrossFileDuplication(\n  files: CrossFileDuplicationSourceFile[],\n  options?: DuplicationOptions\n): CrossFileDuplicationMetrics {\n  const { minTokens, maxGapTokens } = resolveDuplicationOptions(options);\n  const candidates: SelectableCandidate[] = files.flatMap(({ file, candidates }, fileIndex) =>\n    candidates.map((candidate) => ({ ...candidate, regionBucket: fileIndex, file }))\n  );\n  // Pushed one by one: spreading the project-scale window-candidate array as call arguments\n  // overflows V8's argument limit (~124k) and crashes on Node, though Bun/JSC tolerates it.\n  for (const candidate of collectWindowCandidates(files, minTokens)) {\n    candidates.push(candidate);\n  }\n  const counted = selectMaximalGroups(\n    candidates,\n    spansMultipleFiles,\n    // File index and position break coverage ties deterministically.\n    (left, right) => left.regionBucket - right.regionBucket || left.startIndex - right.startIndex\n  );\n  const tokenOffsets = computeTokenOffsets(files, maxGapTokens);\n  return summarize(mergeGapAdjacentGroups([...counted.values()], tokenOffsets, maxGapTokens), files, tokenOffsets);\n}\n\n/** Repeated sub-windows of sibling statements matched across the whole project's files. */\nfunction collectWindowCandidates(files: CrossFileDuplicationSourceFile[], minTokens: number): SelectableCandidate[] {\n  const fileIndexByContext: number[] = [];\n  const contexts: SequenceWindowContext[] = [];\n  for (const [fileIndex, { tokens, containerStatements }] of files.entries()) {\n    if (tokens && containerStatements) {\n      fileIndexByContext.push(fileIndex);\n      contexts.push({ tokens, literalCountPrefix: buildLiteralCountPrefix(tokens), containers: containerStatements });\n    }\n  }\n  if (contexts.length < 2) {\n    return [];\n  }\n  return collectSequenceWindowCandidates(contexts, minTokens, true).flatMap(({ candidate, contextIndex }) => {\n    const fileIndex = fileIndexByContext[contextIndex];\n    const file = fileIndex === undefined ? undefined : files[fileIndex];\n    return fileIndex === undefined || file === undefined\n      ? []\n      : [{ ...candidate, regionBucket: fileIndex, file: file.file }];\n  });\n}\n\n/** A merged group is reported only while it still covers more than one file. */\nfunction spansMultipleFilesAfterMerge(group: CrossFileOccurrence[]): boolean {\n  return new Set(group.map((occurrence) => occurrence.file)).size >= 2;\n}\n\nfunction spansMultipleFiles(group: SelectableCandidate[]): boolean {\n  return group.length >= 2 && new Set(group.map((candidate) => candidate.regionBucket)).size >= 2;\n}\n\n/**\n * Per-file token offsets that map every file into one project-wide token index space: each file's\n * tokens are offset by more than `maxGapTokens` past the previous file's, so occurrences in\n * different files are never gap-adjacent and merged pairs always stay within one file.\n */\nfunction computeTokenOffsets(files: CrossFileDuplicationSourceFile[], maxGapTokens: number): number[] {\n  const tokenOffsets: number[] = [];\n  let offset = 0;\n  for (const { tokens, candidates } of files) {\n    tokenOffsets.push(offset);\n    // Accumulated in a loop: spreading a project-scale candidate array as call arguments would\n    // overflow V8's argument limit (~124k) and crash on Node.\n    let tokenCount = tokens?.length ?? 0;\n    if (!tokens) {\n      for (const candidate of candidates) {\n        tokenCount = Math.max(tokenCount, candidate.endTokenIndex);\n      }\n    }\n    offset += tokenCount + maxGapTokens + 1;\n  }\n  return tokenOffsets;\n}\n\n/** Reuses the within-file gapped (Type-3) merging in the project-wide token index space. */\nfunction mergeGapAdjacentGroups(\n  groups: SelectableCandidate[][],\n  tokenOffsets: number[],\n  maxGapTokens: number\n): CrossFileOccurrence[][] {\n  const occurrenceGroups = groups.map((group) =>\n    group\n      .map((candidate): CrossFileOccurrence => {\n        const start = candidate.startTokenIndex + (tokenOffsets[candidate.regionBucket] ?? 0);\n        const end = candidate.endTokenIndex + (tokenOffsets[candidate.regionBucket] ?? 0);\n        return {\n          file: candidate.file,\n          spanCountedElsewhere: candidate.nestedInLargerGroup,\n          nestedInLargerGroup: candidate.nestedInLargerGroup,\n          segments: [{ startTokenIndex: start, endTokenIndex: end }],\n          tokenCount: candidate.tokenCount,\n          startTokenIndex: start,\n          endTokenIndex: end,\n          startIndex: candidate.startIndex,\n          endIndex: candidate.endIndex,\n          startLine: candidate.startLine,\n          endLine: candidate.endLine,\n        };\n      })\n      .toSorted((left, right) => left.startTokenIndex - right.startTokenIndex)\n  );\n  return mergeAdjacentGroups(occurrenceGroups, maxGapTokens, spansMultipleFilesAfterMerge);\n}\n\nfunction summarize(\n  groups: CrossFileOccurrence[][],\n  files: CrossFileDuplicationSourceFile[],\n  tokenOffsets: number[]\n): CrossFileDuplicationMetrics {\n  const reported: CrossFileDuplicateBlockGroup[] = [];\n  // Accumulated in Maps: file names are arbitrary strings, and a plain object would read\n  // inherited properties for names like \"constructor\".\n  const groupCountByFile = new Map<string, number>();\n  const fileDataByName = new Map(\n    files.map((file, index) => [\n      file.file,\n      { tokens: file.tokens, codeLineNumbers: file.codeLineNumbers, offset: tokenOffsets[index] ?? 0 },\n    ])\n  );\n  const lineNumbersByFile = new Map<string, Set<number>>();\n  let duplicateBlockCount = 0;\n  for (const group of groups) {\n    // Mirrors within-file counting: each redundant occurrence contributes one count per matched\n    // fragment, gapped merging consolidates the grouping without halving the count, and spans a\n    // partial merge shares between a retained group and the merged group count once.\n    duplicateBlockCount += countRedundantFragments(group);\n    for (const occurrence of group) {\n      collectOccurrenceLines(occurrence, fileDataByName, lineNumbersByFile);\n    }\n    const occurrences = group\n      .map(({ file, startLine, endLine }) => ({ file, startLine, endLine }))\n      .toSorted((left, right) => left.file.localeCompare(right.file) || left.startLine - right.startLine);\n    const files = [...new Set(occurrences.map(({ file }) => file))];\n    for (const file of files) {\n      groupCountByFile.set(file, (groupCountByFile.get(file) ?? 0) + 1);\n    }\n    reported.push({ files, occurrences, tokenCount: group[0]?.tokenCount ?? 0 });\n  }\n  reported.sort(\n    (left, right) =>\n      right.tokenCount - left.tokenCount ||\n      (left.occurrences[0]?.file ?? '').localeCompare(right.occurrences[0]?.file ?? '') ||\n      (left.occurrences[0]?.startLine ?? 0) - (right.occurrences[0]?.startLine ?? 0)\n  );\n  return {\n    duplicateBlockCount,\n    duplicateBlockGroupCountByFile: Object.fromEntries(groupCountByFile),\n    duplicateLineNumbersByFile: Object.fromEntries(\n      [...lineNumbersByFile].map(([file, lines]) => [file, [...lines].toSorted((left, right) => left - right)])\n    ),\n    groups: reported,\n  };\n}\n\n/**\n * Adds the code lines an occurrence's matched tokens cover to its file's line set, mapping the\n * project-wide token segments back into the file's own token stream. A file that supplied only\n * candidates (no token stream) is skipped rather than approximated from the bounding line range,\n * which would include gap and comment/blank lines and break the field's exactness contract.\n */\nfunction collectOccurrenceLines(\n  occurrence: CrossFileOccurrence,\n  fileDataByName: Map<\n    string,\n    { tokens?: CrossFileDuplicationSourceFile['tokens']; codeLineNumbers?: Set<number>; offset: number }\n  >,\n  lineNumbersByFile: Map<string, Set<number>>\n): void {\n  const fileData = fileDataByName.get(occurrence.file);\n  if (!fileData?.tokens) {\n    return;\n  }\n  let lines = lineNumbersByFile.get(occurrence.file);\n  if (!lines) {\n    lines = new Set();\n    lineNumbersByFile.set(occurrence.file, lines);\n  }\n  for (const segment of occurrence.segments) {\n    collectSegmentLines(\n      {\n        startTokenIndex: segment.startTokenIndex - fileData.offset,\n        endTokenIndex: segment.endTokenIndex - fileData.offset,\n      },\n      fileData.tokens,\n      fileData.codeLineNumbers,\n      lines\n    );\n  }\n}\n"],"mappings":"4QAyEA,SAAgB,EACd,EACA,EAC6B,CAC7B,GAAM,CAAE,YAAW,gBAAiB,EAA0B,CAAO,EAC/D,EAAoC,EAAM,SAAS,CAAE,OAAM,cAAc,IAC7E,EAAW,IAAK,IAAe,CAAE,GAAG,EAAW,aAAc,EAAW,MAAK,EAAE,CACjF,EAGA,IAAK,IAAM,KAAa,EAAwB,EAAO,CAAS,EAC9D,EAAW,KAAK,CAAS,EAE3B,IAAM,EAAU,EACd,EACA,GAEC,EAAM,IAAU,EAAK,aAAe,EAAM,cAAgB,EAAK,WAAa,EAAM,UACrF,EACM,EAAe,EAAoB,EAAO,CAAY,EAC5D,OAAO,EAAU,EAAuB,CAAC,GAAG,EAAQ,OAAO,CAAC,EAAG,EAAc,CAAY,EAAG,EAAO,CAAY,CACjH,CAGA,SAAS,EAAwB,EAAyC,EAA0C,CAClH,IAAM,EAA+B,CAAC,EAChC,EAAoC,CAAC,EAC3C,IAAK,GAAM,CAAC,EAAW,CAAE,SAAQ,0BAA0B,EAAM,QAAQ,EACnE,GAAU,IACZ,EAAmB,KAAK,CAAS,EACjC,EAAS,KAAK,CAAE,SAAQ,mBAAoB,EAAwB,CAAM,EAAG,WAAY,CAAoB,CAAC,GAMlH,OAHI,EAAS,OAAS,EACb,CAAC,EAEH,EAAgC,EAAU,EAAW,EAAI,CAAC,CAAC,SAAS,CAAE,YAAW,kBAAmB,CACzG,IAAM,EAAY,EAAmB,GAC/B,EAAO,IAAc,IAAA,GAAY,IAAA,GAAY,EAAM,GACzD,OAAO,IAAc,IAAA,IAAa,IAAS,IAAA,GACvC,CAAC,EACD,CAAC,CAAE,GAAG,EAAW,aAAc,EAAW,KAAM,EAAK,IAAK,CAAC,CACjE,CAAC,CACH,CAGA,SAAS,EAA6B,EAAuC,CAC3E,OAAO,IAAI,IAAI,EAAM,IAAK,GAAe,EAAW,IAAI,CAAC,CAAC,CAAC,MAAQ,CACrE,CAEA,SAAS,EAAmB,EAAuC,CACjE,OAAO,EAAM,QAAU,GAAK,IAAI,IAAI,EAAM,IAAK,GAAc,EAAU,YAAY,CAAC,CAAC,CAAC,MAAQ,CAChG,CAOA,SAAS,EAAoB,EAAyC,EAAgC,CACpG,IAAM,EAAyB,CAAC,EAC5B,EAAS,EACb,IAAK,GAAM,CAAE,SAAQ,gBAAgB,EAAO,CAC1C,EAAa,KAAK,CAAM,EAGxB,IAAI,EAAa,GAAQ,QAAU,EACnC,GAAI,CAAC,EACH,IAAK,IAAM,KAAa,EACtB,EAAa,KAAK,IAAI,EAAY,EAAU,aAAa,EAG7D,GAAU,EAAa,EAAe,CACxC,CACA,OAAO,CACT,CAGA,SAAS,EACP,EACA,EACA,EACyB,CACzB,IAAM,EAAmB,EAAO,IAAK,GACnC,EACG,IAAK,GAAmC,CACvC,IAAM,EAAQ,EAAU,iBAAmB,EAAa,EAAU,eAAiB,GAC7E,EAAM,EAAU,eAAiB,EAAa,EAAU,eAAiB,GAC/E,MAAO,CACL,KAAM,EAAU,KAChB,qBAAsB,EAAU,oBAChC,oBAAqB,EAAU,oBAC/B,SAAU,CAAC,CAAE,gBAAiB,EAAO,cAAe,CAAI,CAAC,EACzD,WAAY,EAAU,WACtB,gBAAiB,EACjB,cAAe,EACf,WAAY,EAAU,WACtB,SAAU,EAAU,SACpB,UAAW,EAAU,UACrB,QAAS,EAAU,OACrB,CACF,CAAC,CAAC,CACD,UAAU,EAAM,IAAU,EAAK,gBAAkB,EAAM,eAAe,CAC3E,EACA,OAAO,EAAoB,EAAkB,EAAc,CAA4B,CACzF,CAEA,SAAS,EACP,EACA,EACA,EAC6B,CAC7B,IAAM,EAA2C,CAAC,EAG5C,EAAmB,IAAI,IACvB,EAAiB,IAAI,IACzB,EAAM,KAAK,EAAM,IAAU,CACzB,EAAK,KACL,CAAE,OAAQ,EAAK,OAAQ,gBAAiB,EAAK,gBAAiB,OAAQ,EAAa,IAAU,CAAE,CACjG,CAAC,CACH,EACM,EAAoB,IAAI,IAC1B,EAAsB,EAC1B,IAAK,IAAM,KAAS,EAAQ,CAI1B,GAAuB,EAAwB,CAAK,EACpD,IAAK,IAAM,KAAc,EACvB,EAAuB,EAAY,EAAgB,CAAiB,EAEtE,IAAM,EAAc,EACjB,KAAK,CAAE,OAAM,YAAW,cAAe,CAAE,OAAM,YAAW,SAAQ,EAAE,CAAC,CACrE,UAAU,EAAM,IAAU,EAAK,KAAK,cAAc,EAAM,IAAI,GAAK,EAAK,UAAY,EAAM,SAAS,EAC9F,EAAQ,CAAC,GAAG,IAAI,IAAI,EAAY,KAAK,CAAE,UAAW,CAAI,CAAC,CAAC,EAC9D,IAAK,IAAM,KAAQ,EACjB,EAAiB,IAAI,GAAO,EAAiB,IAAI,CAAI,GAAK,GAAK,CAAC,EAElE,EAAS,KAAK,CAAE,QAAO,cAAa,WAAY,EAAM,EAAE,EAAE,YAAc,CAAE,CAAC,CAC7E,CAOA,OANA,EAAS,MACN,EAAM,IACL,EAAM,WAAa,EAAK,aACvB,EAAK,YAAY,EAAE,EAAE,MAAQ,GAAA,CAAI,cAAc,EAAM,YAAY,EAAE,EAAE,MAAQ,EAAE,IAC/E,EAAK,YAAY,EAAE,EAAE,WAAa,IAAM,EAAM,YAAY,EAAE,EAAE,WAAa,EAChF,EACO,CACL,sBACA,+BAAgC,OAAO,YAAY,CAAgB,EACnE,2BAA4B,OAAO,YACjC,CAAC,GAAG,CAAiB,CAAC,CAAC,KAAK,CAAC,EAAM,KAAW,CAAC,EAAM,CAAC,GAAG,CAAK,CAAC,CAAC,UAAU,EAAM,IAAU,EAAO,CAAK,CAAC,CAAC,CAC1G,EACA,OAAQ,CACV,CACF,CAQA,SAAS,EACP,EACA,EAIA,EACM,CACN,IAAM,EAAW,EAAe,IAAI,EAAW,IAAI,EACnD,GAAI,CAAC,GAAU,OACb,OAEF,IAAI,EAAQ,EAAkB,IAAI,EAAW,IAAI,EAC5C,IACH,EAAQ,IAAI,IACZ,EAAkB,IAAI,EAAW,KAAM,CAAK,GAE9C,IAAK,IAAM,KAAW,EAAW,SAC/B,EACE,CACE,gBAAiB,EAAQ,gBAAkB,EAAS,OACpD,cAAe,EAAQ,cAAgB,EAAS,MAClD,EACA,EAAS,OACT,EAAS,gBACT,CACF,CAEJ"}

package/dist/duplicateSelection.cjs

@@ -1,2 +1,2 @@
-"use strict";function e(e,n,r){let i=new Map;for(let t of e){let e=i.get(t.fingerprint)??[];e.push(t),i.set(t.fingerprint,e)}let a=[...i.values()].map(t).filter(n),o=new Map(a.map(e=>[e[0]?.fingerprint??``,e.length])),s=e=>e.tokenCount*(o.get(e.fingerprint)??1),c=a.flat();c.sort((e,t)=>s(t)-s(e)||(r?r(e,t):0));for(let e=0;;e+=1){let t=new Map,r=new Map;for(let e of c){let n=t.get(e.regionBucket??0)??[];if(n.some(t=>t.startIndex<e.endIndex&&e.startIndex<t.endIndex))continue;n.push(e),t.set(e.regionBucket??0,n);let i=r.get(e.fingerprint)??[];i.push(e),r.set(e.fingerprint,i)}let i,a=-1;for(let[e,t]of r){let r=t[0]?.tokenCount??0;!n(t)&&r>a&&(i=e,a=r)}if(i===void 0)return r;if(e>=20){for(let[e,t]of r)n(t)||r.delete(e);return r}c=c.filter(e=>e.fingerprint!==i)}}function t(e){let t=new Map;for(let n of e){let e=`${n.regionBucket??0}:${n.startIndex}:${n.endIndex}`,r=t.get(e);(!r||n.tokenCount>r.tokenCount)&&t.set(e,n)}return[...t.values()]}exports.dedupeByRegion=t,exports.selectMaximalGroups=e;
+"use strict";function e(e,r,i){let a=new Map;for(let t of e){let e=a.get(t.fingerprint)??[];e.push(t),a.set(t.fingerprint,e)}let o=[...a.values()].map(n).filter(r),s=new Map(o.map(e=>[e[0]?.fingerprint??``,e.length])),c=e=>e.tokenCount*(s.get(e.fingerprint)??1),l=o.flat();l.sort((e,t)=>c(t)-c(e)||(i?i(e,t):0));for(let e=0;;e+=1){let n=new Map,i=new Map,a=new Map;for(let e of l){let t=n.get(e.regionBucket??0)??[],r=!1,o=!1,s;for(let n of t)if(!(n.startIndex>=e.endIndex||e.startIndex>=n.endIndex)){if(n.startIndex<=e.startIndex&&e.endIndex<=n.endIndex){r=!0;break}if(e.startIndex<=n.startIndex&&n.endIndex<=e.endIndex)(s??=[]).push(n);else{o=!0;break}}if(r){let t=a.get(e.fingerprint)??[];t.push({...e,nestedInLargerGroup:!0}),a.set(e.fingerprint,t);continue}if(o)continue;let c=s;for(let e of c??[]){let t=i.get(e.fingerprint)??[],n=t.indexOf(e);n!==-1&&(t[n]={...e,nestedInLargerGroup:!0})}let l=c?t.filter(e=>!c.includes(e)):t;l.push(e),n.set(e.regionBucket??0,l);let u=i.get(e.fingerprint)??[];u.push(e),i.set(e.fingerprint,u)}for(let[e,t]of a)i.get(e)?.push(...t);for(let[e,t]of i)t.every(e=>e.nestedInLargerGroup)&&i.delete(e);let o,s=-1;for(let[e,t]of i){let n=t[0]?.tokenCount??0;!r(t)&&n>s&&(o=e,s=n)}if(o===void 0)return i;if(e>=20)return t(i,r),i;l=l.filter(e=>e.fingerprint!==o)}}function t(e,t){for(let n=!0;n;){n=!1;for(let[r,i]of e)(!t(i)||i.every(e=>e.nestedInLargerGroup))&&(e.delete(r),n=!0);let r=[...e.values()].flat().filter(e=>!e.nestedInLargerGroup);for(let[t,i]of e){let a=i.filter(e=>!e.nestedInLargerGroup||r.some(t=>(t.regionBucket??0)===(e.regionBucket??0)&&t.startIndex<=e.startIndex&&e.endIndex<=t.endIndex));a.length!==i.length&&(e.set(t,a),n=!0)}}}function n(e){let t=new Map;for(let n of e){let e=`${n.regionBucket??0}:${n.startIndex}:${n.endIndex}`,r=t.get(e);(!r||n.tokenCount>r.tokenCount)&&t.set(e,n)}return[...t.values()]}exports.dedupeByRegion=n,exports.selectMaximalGroups=e;
 //# sourceMappingURL=duplicateSelection.cjs.map

package/dist/duplicateSelection.cjs.map

@@ -1 +1 @@
-{"version":3,"file":"duplicateSelection.cjs","names":[],"sources":["../src/duplicateSelection.ts"],"sourcesContent":["/**\n * Maximal, non-overlapping duplicate-group selection shared by the within-file and cross-file\n * detectors. Candidates are grouped by fingerprint, ranked by total coverage, kept greedily\n * without overlapping a kept region, and groups that fall below the survivor requirement are shed\n * one at a time (largest first) so their regions stop blocking smaller groups.\n */\n\nexport interface SelectableRegion {\n  fingerprint: string;\n  tokenCount: number;\n  startIndex: number;\n  endIndex: number;\n  /**\n   * Regions can only overlap within the same bucket. The within-file detector uses one bucket;\n   * the cross-file detector buckets by file index.\n   */\n  regionBucket?: number;\n}\n\n/** Caps how often the maximal-region selection reruns after shedding failed duplicate groups. */\nconst maxSelectionRerunCount = 20;\n\n/**\n * @param isSurvivingGroup whether a selected group counts (e.g. at least two occurrences, or\n *   occurrences spanning at least two files); failing groups are shed and re-selected without.\n * @param compareTies optional deterministic tie-break applied after the coverage ranking.\n */\nexport function selectMaximalGroups<T extends SelectableRegion>(\n  candidates: T[],\n  isSurvivingGroup: (group: T[]) => boolean,\n  compareTies?: (left: T, right: T) => number\n): Map<string, T[]> {\n  const byFingerprint = new Map<string, T[]>();\n  for (const candidate of candidates) {\n    const group = byFingerprint.get(candidate.fingerprint) ?? [];\n    group.push(candidate);\n    byFingerprint.set(candidate.fingerprint, group);\n  }\n\n  const groups = [...byFingerprint.values()].map(dedupeByRegion).filter(isSurvivingGroup);\n  // Greedy order ranks by total coverage (region size × copies): a 3×3-statement group must beat\n  // a 2×4-statement group overlapping two of its copies, or the third copy is silently dropped\n  // and the reported duplication shrinks as more copies are added.\n  const groupSizeByFingerprint = new Map(groups.map((group) => [group[0]?.fingerprint ?? '', group.length]));\n  const coverage = (candidate: T): number =>\n    candidate.tokenCount * (groupSizeByFingerprint.get(candidate.fingerprint) ?? 1);\n  let duplicates = groups.flat();\n  duplicates.sort((left, right) => coverage(right) - coverage(left) || (compareTies ? compareTies(left, right) : 0));\n\n  // Greedy selection can keep a candidate whose group ends up below the survivor requirement;\n  // such an uncounted region must not block smaller groups, so the largest failed group is\n  // removed and the selection reruns. One group at a time: freeing a failed group's regions can\n  // rescue another. The rerun cap bounds degenerate inputs; past it the remaining failed groups\n  // are dropped, trading a sliver of recall on such files for bounded runtime.\n  for (let rerun = 0; ; rerun += 1) {\n    const keptRegionsByBucket = new Map<number, { startIndex: number; endIndex: number }[]>();\n    const counted = new Map<string, T[]>();\n    for (const candidate of duplicates) {\n      const keptRegions = keptRegionsByBucket.get(candidate.regionBucket ?? 0) ?? [];\n      if (\n        keptRegions.some((region) => region.startIndex < candidate.endIndex && candidate.startIndex < region.endIndex)\n      ) {\n        continue;\n      }\n      keptRegions.push(candidate);\n      keptRegionsByBucket.set(candidate.regionBucket ?? 0, keptRegions);\n      const group = counted.get(candidate.fingerprint) ?? [];\n      group.push(candidate);\n      counted.set(candidate.fingerprint, group);\n    }\n\n    let failedFingerprint: string | undefined;\n    let failedTokenCount = -1;\n    for (const [fingerprint, group] of counted) {\n      const tokenCount = group[0]?.tokenCount ?? 0;\n      if (!isSurvivingGroup(group) && tokenCount > failedTokenCount) {\n        failedFingerprint = fingerprint;\n        failedTokenCount = tokenCount;\n      }\n    }\n    // No failed fingerprint means every counted group met the survivor requirement.\n    if (failedFingerprint === undefined) {\n      return counted;\n    }\n    if (rerun >= maxSelectionRerunCount) {\n      for (const [fingerprint, group] of counted) {\n        if (!isSurvivingGroup(group)) {\n          counted.delete(fingerprint);\n        }\n      }\n      return counted;\n    }\n\n    duplicates = duplicates.filter((candidate) => candidate.fingerprint !== failedFingerprint);\n  }\n}\n\n/** Drops candidates covering the same source region (a block and the statement run spanning it). */\nexport function dedupeByRegion<T extends SelectableRegion>(group: T[]): T[] {\n  const byRegion = new Map<string, T>();\n  for (const candidate of group) {\n    const key = `${candidate.regionBucket ?? 0}:${candidate.startIndex}:${candidate.endIndex}`;\n    const existing = byRegion.get(key);\n    if (!existing || candidate.tokenCount > existing.tokenCount) {\n      byRegion.set(key, candidate);\n    }\n  }\n  return [...byRegion.values()];\n}\n"],"mappings":"aA2BA,SAAgB,EACd,EACA,EACA,EACkB,CAClB,IAAM,EAAgB,IAAI,IAC1B,IAAK,IAAM,KAAa,EAAY,CAClC,IAAM,EAAQ,EAAc,IAAI,EAAU,WAAW,GAAK,CAAC,EAC3D,EAAM,KAAK,CAAS,EACpB,EAAc,IAAI,EAAU,YAAa,CAAK,CAChD,CAEA,IAAM,EAAS,CAAC,GAAG,EAAc,OAAO,CAAC,CAAC,CAAC,IAAI,CAAc,CAAC,CAAC,OAAO,CAAgB,EAIhF,EAAyB,IAAI,IAAI,EAAO,IAAK,GAAU,CAAC,EAAM,EAAE,EAAE,aAAe,GAAI,EAAM,MAAM,CAAC,CAAC,EACnG,EAAY,GAChB,EAAU,YAAc,EAAuB,IAAI,EAAU,WAAW,GAAK,GAC3E,EAAa,EAAO,KAAK,EAC7B,EAAW,MAAM,EAAM,IAAU,EAAS,CAAK,EAAI,EAAS,CAAI,IAAM,EAAc,EAAY,EAAM,CAAK,EAAI,EAAE,EAOjH,IAAK,IAAI,EAAQ,GAAK,GAAS,EAAG,CAChC,IAAM,EAAsB,IAAI,IAC1B,EAAU,IAAI,IACpB,IAAK,IAAM,KAAa,EAAY,CAClC,IAAM,EAAc,EAAoB,IAAI,EAAU,cAAgB,CAAC,GAAK,CAAC,EAC7E,GACE,EAAY,KAAM,GAAW,EAAO,WAAa,EAAU,UAAY,EAAU,WAAa,EAAO,QAAQ,EAE7G,SAEF,EAAY,KAAK,CAAS,EAC1B,EAAoB,IAAI,EAAU,cAAgB,EAAG,CAAW,EAChE,IAAM,EAAQ,EAAQ,IAAI,EAAU,WAAW,GAAK,CAAC,EACrD,EAAM,KAAK,CAAS,EACpB,EAAQ,IAAI,EAAU,YAAa,CAAK,CAC1C,CAEA,IAAI,EACA,EAAmB,GACvB,IAAK,GAAM,CAAC,EAAa,KAAU,EAAS,CAC1C,IAAM,EAAa,EAAM,EAAE,EAAE,YAAc,EACvC,CAAC,EAAiB,CAAK,GAAK,EAAa,IAC3C,EAAoB,EACpB,EAAmB,EAEvB,CAEA,GAAI,IAAsB,IAAA,GACxB,OAAO,EAET,GAAI,GAAS,GAAwB,CACnC,IAAK,GAAM,CAAC,EAAa,KAAU,EAC5B,EAAiB,CAAK,GACzB,EAAQ,OAAO,CAAW,EAG9B,OAAO,CACT,CAEA,EAAa,EAAW,OAAQ,GAAc,EAAU,cAAgB,CAAiB,CAC3F,CACF,CAGA,SAAgB,EAA2C,EAAiB,CAC1E,IAAM,EAAW,IAAI,IACrB,IAAK,IAAM,KAAa,EAAO,CAC7B,IAAM,EAAM,GAAG,EAAU,cAAgB,EAAE,GAAG,EAAU,WAAW,GAAG,EAAU,WAC1E,EAAW,EAAS,IAAI,CAAG,GAC7B,CAAC,GAAY,EAAU,WAAa,EAAS,aAC/C,EAAS,IAAI,EAAK,CAAS,CAE/B,CACA,MAAO,CAAC,GAAG,EAAS,OAAO,CAAC,CAC9B"}
+{"version":3,"file":"duplicateSelection.cjs","names":[],"sources":["../src/duplicateSelection.ts"],"sourcesContent":["/**\n * Maximal, non-overlapping duplicate-group selection for the cross-file detector (the native\n * within-file detector mirrors its greedy ranking and shedding, but not the nested-copy retention\n * below, which is cross-file only). Candidates are grouped by fingerprint, ranked by total\n * coverage, kept greedily without overlapping a kept region, and groups that fall below the\n * survivor requirement are shed one at a time (largest first) so their regions stop blocking\n * smaller groups. A copy lying entirely inside a larger group's region stays with its group as a\n * nested copy (whichever group the greedy order kept first), so a standalone copy elsewhere is\n * still reported as duplicating it.\n */\n\nexport interface SelectableRegion {\n  fingerprint: string;\n  tokenCount: number;\n  startIndex: number;\n  endIndex: number;\n  /**\n   * Regions can only overlap within the same bucket. The within-file detector uses one bucket;\n   * the cross-file detector buckets by file index.\n   */\n  regionBucket?: number;\n  /**\n   * Set by selectMaximalGroups on a copy nested inside a larger group's region: it is reported with\n   * its group, but its span is already counted by that larger group.\n   */\n  nestedInLargerGroup?: boolean;\n}\n\n/** Caps how often the maximal-region selection reruns after shedding failed duplicate groups. */\nconst maxSelectionRerunCount = 20;\n\n/**\n * @param isSurvivingGroup whether a selected group counts (e.g. at least two occurrences, or\n *   occurrences spanning at least two files); failing groups are shed and re-selected without.\n * @param compareTies optional deterministic tie-break applied after the coverage ranking.\n */\nexport function selectMaximalGroups<T extends SelectableRegion>(\n  candidates: T[],\n  isSurvivingGroup: (group: T[]) => boolean,\n  compareTies?: (left: T, right: T) => number\n): Map<string, T[]> {\n  const byFingerprint = new Map<string, T[]>();\n  for (const candidate of candidates) {\n    const group = byFingerprint.get(candidate.fingerprint) ?? [];\n    group.push(candidate);\n    byFingerprint.set(candidate.fingerprint, group);\n  }\n\n  const groups = [...byFingerprint.values()].map(dedupeByRegion).filter(isSurvivingGroup);\n  // Greedy order ranks by total coverage (region size × copies): a 3×3-statement group must beat\n  // a 2×4-statement group overlapping two of its copies, or the third copy is silently dropped\n  // and the reported duplication shrinks as more copies are added.\n  const groupSizeByFingerprint = new Map(groups.map((group) => [group[0]?.fingerprint ?? '', group.length]));\n  const coverage = (candidate: T): number =>\n    candidate.tokenCount * (groupSizeByFingerprint.get(candidate.fingerprint) ?? 1);\n  let duplicates = groups.flat();\n  duplicates.sort((left, right) => coverage(right) - coverage(left) || (compareTies ? compareTies(left, right) : 0));\n\n  // Greedy selection can keep a candidate whose group ends up below the survivor requirement;\n  // such an uncounted region must not block smaller groups, so the largest failed group is\n  // removed and the selection reruns. One group at a time: freeing a failed group's regions can\n  // rescue another. The rerun cap bounds degenerate inputs; past it the remaining failed groups\n  // are dropped, trading a sliver of recall on such files for bounded runtime.\n  for (let rerun = 0; ; rerun += 1) {\n    const keptRegionsByBucket = new Map<number, T[]>();\n    const counted = new Map<string, T[]>();\n    const nestedByFingerprint = new Map<string, T[]>();\n    for (const candidate of duplicates) {\n      const keptRegions = keptRegionsByBucket.get(candidate.regionBucket ?? 0) ?? [];\n      // A plain loop: this runs once per candidate over every kept region of the bucket, so\n      // allocating a filtered array per candidate would dominate project-scale runs. Kept regions\n      // never overlap each other, so a candidate inside one cannot partially overlap another.\n      let containedInKept = false;\n      let partiallyOverlaps = false;\n      let enclosedKept: T[] | undefined;\n      for (const region of keptRegions) {\n        if (region.startIndex >= candidate.endIndex || candidate.startIndex >= region.endIndex) {\n          continue;\n        }\n        if (region.startIndex <= candidate.startIndex && candidate.endIndex <= region.endIndex) {\n          containedInKept = true;\n          break;\n        }\n        if (candidate.startIndex <= region.startIndex && region.endIndex <= candidate.endIndex) {\n          (enclosedKept ??= []).push(region);\n        } else {\n          partiallyOverlaps = true;\n          break;\n        }\n      }\n      if (containedInKept) {\n        const nested = nestedByFingerprint.get(candidate.fingerprint) ?? [];\n        nested.push({ ...candidate, nestedInLargerGroup: true });\n        nestedByFingerprint.set(candidate.fingerprint, nested);\n        continue;\n      }\n      if (partiallyOverlaps) {\n        continue;\n      }\n      // Containment must not depend on greedy order: a candidate enclosing kept copies of smaller\n      // groups occupies its region, and those copies become nested copies of their groups.\n      const enclosed = enclosedKept;\n      for (const inner of enclosed ?? []) {\n        const group = counted.get(inner.fingerprint) ?? [];\n        const index = group.indexOf(inner);\n        if (index !== -1) {\n          group[index] = { ...inner, nestedInLargerGroup: true };\n        }\n      }\n      // The enclosed regions give way to the enclosing one, keeping kept regions mutually\n      // non-overlapping: a later candidate inside this region must see it, not a region it\n      // swallowed (which the candidate could straddle instead).\n      const occupied = enclosed ? keptRegions.filter((region) => !enclosed.includes(region)) : keptRegions;\n      occupied.push(candidate);\n      keptRegionsByBucket.set(candidate.regionBucket ?? 0, occupied);\n      const group = counted.get(candidate.fingerprint) ?? [];\n      group.push(candidate);\n      counted.set(candidate.fingerprint, group);\n    }\n    // Nested copies join only a group that kept a standalone copy; on their own they would merely\n    // restate the larger group.\n    for (const [fingerprint, nested] of nestedByFingerprint) {\n      counted.get(fingerprint)?.push(...nested);\n    }\n    for (const [fingerprint, group] of counted) {\n      if (group.every((candidate) => candidate.nestedInLargerGroup)) {\n        counted.delete(fingerprint);\n      }\n    }\n\n    let failedFingerprint: string | undefined;\n    let failedTokenCount = -1;\n    for (const [fingerprint, group] of counted) {\n      const tokenCount = group[0]?.tokenCount ?? 0;\n      if (!isSurvivingGroup(group) && tokenCount > failedTokenCount) {\n        failedFingerprint = fingerprint;\n        failedTokenCount = tokenCount;\n      }\n    }\n    // No failed fingerprint means every counted group met the survivor requirement.\n    if (failedFingerprint === undefined) {\n      return counted;\n    }\n    if (rerun >= maxSelectionRerunCount) {\n      dropFailedGroups(counted, isSurvivingGroup);\n      return counted;\n    }\n\n    duplicates = duplicates.filter((candidate) => candidate.fingerprint !== failedFingerprint);\n  }\n}\n\n/**\n * Past the rerun cap, still-failing groups are dropped without another selection pass. A dropped\n * group's regions may have been what nested copies of surviving groups lay inside, and such a copy\n * would then be counted by no group at all, so those copies are dropped too and the shrunk groups\n * are re-checked until nothing changes.\n */\nfunction dropFailedGroups<T extends SelectableRegion>(\n  counted: Map<string, T[]>,\n  isSurvivingGroup: (group: T[]) => boolean\n): void {\n  for (let changed = true; changed;) {\n    changed = false;\n    for (const [fingerprint, group] of counted) {\n      if (!isSurvivingGroup(group) || group.every((candidate) => candidate.nestedInLargerGroup)) {\n        counted.delete(fingerprint);\n        changed = true;\n      }\n    }\n    const standalone = [...counted.values()].flat().filter((candidate) => !candidate.nestedInLargerGroup);\n    for (const [fingerprint, group] of counted) {\n      const kept = group.filter(\n        (candidate) =>\n          !candidate.nestedInLargerGroup ||\n          standalone.some(\n            (region) =>\n              (region.regionBucket ?? 0) === (candidate.regionBucket ?? 0) &&\n              region.startIndex <= candidate.startIndex &&\n              candidate.endIndex <= region.endIndex\n          )\n      );\n      if (kept.length !== group.length) {\n        counted.set(fingerprint, kept);\n        changed = true;\n      }\n    }\n  }\n}\n\n/** Drops candidates covering the same source region (a block and the statement run spanning it). */\nexport function dedupeByRegion<T extends SelectableRegion>(group: T[]): T[] {\n  const byRegion = new Map<string, T>();\n  for (const candidate of group) {\n    const key = `${candidate.regionBucket ?? 0}:${candidate.startIndex}:${candidate.endIndex}`;\n    const existing = byRegion.get(key);\n    if (!existing || candidate.tokenCount > existing.tokenCount) {\n      byRegion.set(key, candidate);\n    }\n  }\n  return [...byRegion.values()];\n}\n"],"mappings":"aAoCA,SAAgB,EACd,EACA,EACA,EACkB,CAClB,IAAM,EAAgB,IAAI,IAC1B,IAAK,IAAM,KAAa,EAAY,CAClC,IAAM,EAAQ,EAAc,IAAI,EAAU,WAAW,GAAK,CAAC,EAC3D,EAAM,KAAK,CAAS,EACpB,EAAc,IAAI,EAAU,YAAa,CAAK,CAChD,CAEA,IAAM,EAAS,CAAC,GAAG,EAAc,OAAO,CAAC,CAAC,CAAC,IAAI,CAAc,CAAC,CAAC,OAAO,CAAgB,EAIhF,EAAyB,IAAI,IAAI,EAAO,IAAK,GAAU,CAAC,EAAM,EAAE,EAAE,aAAe,GAAI,EAAM,MAAM,CAAC,CAAC,EACnG,EAAY,GAChB,EAAU,YAAc,EAAuB,IAAI,EAAU,WAAW,GAAK,GAC3E,EAAa,EAAO,KAAK,EAC7B,EAAW,MAAM,EAAM,IAAU,EAAS,CAAK,EAAI,EAAS,CAAI,IAAM,EAAc,EAAY,EAAM,CAAK,EAAI,EAAE,EAOjH,IAAK,IAAI,EAAQ,GAAK,GAAS,EAAG,CAChC,IAAM,EAAsB,IAAI,IAC1B,EAAU,IAAI,IACd,EAAsB,IAAI,IAChC,IAAK,IAAM,KAAa,EAAY,CAClC,IAAM,EAAc,EAAoB,IAAI,EAAU,cAAgB,CAAC,GAAK,CAAC,EAIzE,EAAkB,GAClB,EAAoB,GACpB,EACJ,IAAK,IAAM,KAAU,EACf,OAAO,YAAc,EAAU,UAAY,EAAU,YAAc,EAAO,UAG9E,IAAI,EAAO,YAAc,EAAU,YAAc,EAAU,UAAY,EAAO,SAAU,CACtF,EAAkB,GAClB,KACF,CACA,GAAI,EAAU,YAAc,EAAO,YAAc,EAAO,UAAY,EAAU,UAC3E,IAAiB,CAAC,EAAA,CAAG,KAAK,CAAM,MAC5B,CACL,EAAoB,GACpB,KACF,CANA,CAQF,GAAI,EAAiB,CACnB,IAAM,EAAS,EAAoB,IAAI,EAAU,WAAW,GAAK,CAAC,EAClE,EAAO,KAAK,CAAE,GAAG,EAAW,oBAAqB,EAAK,CAAC,EACvD,EAAoB,IAAI,EAAU,YAAa,CAAM,EACrD,QACF,CACA,GAAI,EACF,SAIF,IAAM,EAAW,EACjB,IAAK,IAAM,KAAS,GAAY,CAAC,EAAG,CAClC,IAAM,EAAQ,EAAQ,IAAI,EAAM,WAAW,GAAK,CAAC,EAC3C,EAAQ,EAAM,QAAQ,CAAK,EAC7B,IAAU,KACZ,EAAM,GAAS,CAAE,GAAG,EAAO,oBAAqB,EAAK,EAEzD,CAIA,IAAM,EAAW,EAAW,EAAY,OAAQ,GAAW,CAAC,EAAS,SAAS,CAAM,CAAC,EAAI,EACzF,EAAS,KAAK,CAAS,EACvB,EAAoB,IAAI,EAAU,cAAgB,EAAG,CAAQ,EAC7D,IAAM,EAAQ,EAAQ,IAAI,EAAU,WAAW,GAAK,CAAC,EACrD,EAAM,KAAK,CAAS,EACpB,EAAQ,IAAI,EAAU,YAAa,CAAK,CAC1C,CAGA,IAAK,GAAM,CAAC,EAAa,KAAW,EAClC,EAAQ,IAAI,CAAW,CAAC,EAAE,KAAK,GAAG,CAAM,EAE1C,IAAK,GAAM,CAAC,EAAa,KAAU,EAC7B,EAAM,MAAO,GAAc,EAAU,mBAAmB,GAC1D,EAAQ,OAAO,CAAW,EAI9B,IAAI,EACA,EAAmB,GACvB,IAAK,GAAM,CAAC,EAAa,KAAU,EAAS,CAC1C,IAAM,EAAa,EAAM,EAAE,EAAE,YAAc,EACvC,CAAC,EAAiB,CAAK,GAAK,EAAa,IAC3C,EAAoB,EACpB,EAAmB,EAEvB,CAEA,GAAI,IAAsB,IAAA,GACxB,OAAO,EAET,GAAI,GAAS,GAEX,OADA,EAAiB,EAAS,CAAgB,EACnC,EAGT,EAAa,EAAW,OAAQ,GAAc,EAAU,cAAgB,CAAiB,CAC3F,CACF,CAQA,SAAS,EACP,EACA,EACM,CACN,IAAK,IAAI,EAAU,GAAM,GAAU,CACjC,EAAU,GACV,IAAK,GAAM,CAAC,EAAa,KAAU,GAC7B,CAAC,EAAiB,CAAK,GAAK,EAAM,MAAO,GAAc,EAAU,mBAAmB,KACtF,EAAQ,OAAO,CAAW,EAC1B,EAAU,IAGd,IAAM,EAAa,CAAC,GAAG,EAAQ,OAAO,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,OAAQ,GAAc,CAAC,EAAU,mBAAmB,EACpG,IAAK,GAAM,CAAC,EAAa,KAAU,EAAS,CAC1C,IAAM,EAAO,EAAM,OAChB,GACC,CAAC,EAAU,qBACX,EAAW,KACR,IACE,EAAO,cAAgB,MAAQ,EAAU,cAAgB,IAC1D,EAAO,YAAc,EAAU,YAC/B,EAAU,UAAY,EAAO,QACjC,CACJ,EACI,EAAK,SAAW,EAAM,SACxB,EAAQ,IAAI,EAAa,CAAI,EAC7B,EAAU,GAEd,CACF,CACF,CAGA,SAAgB,EAA2C,EAAiB,CAC1E,IAAM,EAAW,IAAI,IACrB,IAAK,IAAM,KAAa,EAAO,CAC7B,IAAM,EAAM,GAAG,EAAU,cAAgB,EAAE,GAAG,EAAU,WAAW,GAAG,EAAU,WAC1E,EAAW,EAAS,IAAI,CAAG,GAC7B,CAAC,GAAY,EAAU,WAAa,EAAS,aAC/C,EAAS,IAAI,EAAK,CAAS,CAE/B,CACA,MAAO,CAAC,GAAG,EAAS,OAAO,CAAC,CAC9B"}

package/dist/duplicateSelection.d.ts

@@ -1,8 +1,12 @@
 /**
- * Maximal, non-overlapping duplicate-group selection shared by the within-file and cross-file
- * detectors. Candidates are grouped by fingerprint, ranked by total coverage, kept greedily
- * without overlapping a kept region, and groups that fall below the survivor requirement are shed
- * one at a time (largest first) so their regions stop blocking smaller groups.
+ * Maximal, non-overlapping duplicate-group selection for the cross-file detector (the native
+ * within-file detector mirrors its greedy ranking and shedding, but not the nested-copy retention
+ * below, which is cross-file only). Candidates are grouped by fingerprint, ranked by total
+ * coverage, kept greedily without overlapping a kept region, and groups that fall below the
+ * survivor requirement are shed one at a time (largest first) so their regions stop blocking
+ * smaller groups. A copy lying entirely inside a larger group's region stays with its group as a
+ * nested copy (whichever group the greedy order kept first), so a standalone copy elsewhere is
+ * still reported as duplicating it.
  */
 export interface SelectableRegion {
     fingerprint: string;
@@ -14,6 +18,11 @@ export interface SelectableRegion {
      * the cross-file detector buckets by file index.
      */
     regionBucket?: number;
+    /**
+     * Set by selectMaximalGroups on a copy nested inside a larger group's region: it is reported with
+     * its group, but its span is already counted by that larger group.
+     */
+    nestedInLargerGroup?: boolean;
 }
 /**
  * @param isSurvivingGroup whether a selected group counts (e.g. at least two occurrences, or

package/dist/duplicateSelection.js

@@ -1,2 +1,2 @@
-function e(e,n,r){let i=new Map;for(let t of e){let e=i.get(t.fingerprint)??[];e.push(t),i.set(t.fingerprint,e)}let a=[...i.values()].map(t).filter(n),o=new Map(a.map(e=>[e[0]?.fingerprint??``,e.length])),s=e=>e.tokenCount*(o.get(e.fingerprint)??1),c=a.flat();c.sort((e,t)=>s(t)-s(e)||(r?r(e,t):0));for(let e=0;;e+=1){let t=new Map,r=new Map;for(let e of c){let n=t.get(e.regionBucket??0)??[];if(n.some(t=>t.startIndex<e.endIndex&&e.startIndex<t.endIndex))continue;n.push(e),t.set(e.regionBucket??0,n);let i=r.get(e.fingerprint)??[];i.push(e),r.set(e.fingerprint,i)}let i,a=-1;for(let[e,t]of r){let r=t[0]?.tokenCount??0;!n(t)&&r>a&&(i=e,a=r)}if(i===void 0)return r;if(e>=20){for(let[e,t]of r)n(t)||r.delete(e);return r}c=c.filter(e=>e.fingerprint!==i)}}function t(e){let t=new Map;for(let n of e){let e=`${n.regionBucket??0}:${n.startIndex}:${n.endIndex}`,r=t.get(e);(!r||n.tokenCount>r.tokenCount)&&t.set(e,n)}return[...t.values()]}export{t as dedupeByRegion,e as selectMaximalGroups};
+function e(e,r,i){let a=new Map;for(let t of e){let e=a.get(t.fingerprint)??[];e.push(t),a.set(t.fingerprint,e)}let o=[...a.values()].map(n).filter(r),s=new Map(o.map(e=>[e[0]?.fingerprint??``,e.length])),c=e=>e.tokenCount*(s.get(e.fingerprint)??1),l=o.flat();l.sort((e,t)=>c(t)-c(e)||(i?i(e,t):0));for(let e=0;;e+=1){let n=new Map,i=new Map,a=new Map;for(let e of l){let t=n.get(e.regionBucket??0)??[],r=!1,o=!1,s;for(let n of t)if(!(n.startIndex>=e.endIndex||e.startIndex>=n.endIndex)){if(n.startIndex<=e.startIndex&&e.endIndex<=n.endIndex){r=!0;break}if(e.startIndex<=n.startIndex&&n.endIndex<=e.endIndex)(s??=[]).push(n);else{o=!0;break}}if(r){let t=a.get(e.fingerprint)??[];t.push({...e,nestedInLargerGroup:!0}),a.set(e.fingerprint,t);continue}if(o)continue;let c=s;for(let e of c??[]){let t=i.get(e.fingerprint)??[],n=t.indexOf(e);n!==-1&&(t[n]={...e,nestedInLargerGroup:!0})}let l=c?t.filter(e=>!c.includes(e)):t;l.push(e),n.set(e.regionBucket??0,l);let u=i.get(e.fingerprint)??[];u.push(e),i.set(e.fingerprint,u)}for(let[e,t]of a)i.get(e)?.push(...t);for(let[e,t]of i)t.every(e=>e.nestedInLargerGroup)&&i.delete(e);let o,s=-1;for(let[e,t]of i){let n=t[0]?.tokenCount??0;!r(t)&&n>s&&(o=e,s=n)}if(o===void 0)return i;if(e>=20)return t(i,r),i;l=l.filter(e=>e.fingerprint!==o)}}function t(e,t){for(let n=!0;n;){n=!1;for(let[r,i]of e)(!t(i)||i.every(e=>e.nestedInLargerGroup))&&(e.delete(r),n=!0);let r=[...e.values()].flat().filter(e=>!e.nestedInLargerGroup);for(let[t,i]of e){let a=i.filter(e=>!e.nestedInLargerGroup||r.some(t=>(t.regionBucket??0)===(e.regionBucket??0)&&t.startIndex<=e.startIndex&&e.endIndex<=t.endIndex));a.length!==i.length&&(e.set(t,a),n=!0)}}}function n(e){let t=new Map;for(let n of e){let e=`${n.regionBucket??0}:${n.startIndex}:${n.endIndex}`,r=t.get(e);(!r||n.tokenCount>r.tokenCount)&&t.set(e,n)}return[...t.values()]}export{n as dedupeByRegion,e as selectMaximalGroups};
 //# sourceMappingURL=duplicateSelection.js.map

package/dist/duplicateSelection.js.map

@@ -1 +1 @@
-{"version":3,"file":"duplicateSelection.js","names":[],"sources":["../src/duplicateSelection.ts"],"sourcesContent":["/**\n * Maximal, non-overlapping duplicate-group selection shared by the within-file and cross-file\n * detectors. Candidates are grouped by fingerprint, ranked by total coverage, kept greedily\n * without overlapping a kept region, and groups that fall below the survivor requirement are shed\n * one at a time (largest first) so their regions stop blocking smaller groups.\n */\n\nexport interface SelectableRegion {\n  fingerprint: string;\n  tokenCount: number;\n  startIndex: number;\n  endIndex: number;\n  /**\n   * Regions can only overlap within the same bucket. The within-file detector uses one bucket;\n   * the cross-file detector buckets by file index.\n   */\n  regionBucket?: number;\n}\n\n/** Caps how often the maximal-region selection reruns after shedding failed duplicate groups. */\nconst maxSelectionRerunCount = 20;\n\n/**\n * @param isSurvivingGroup whether a selected group counts (e.g. at least two occurrences, or\n *   occurrences spanning at least two files); failing groups are shed and re-selected without.\n * @param compareTies optional deterministic tie-break applied after the coverage ranking.\n */\nexport function selectMaximalGroups<T extends SelectableRegion>(\n  candidates: T[],\n  isSurvivingGroup: (group: T[]) => boolean,\n  compareTies?: (left: T, right: T) => number\n): Map<string, T[]> {\n  const byFingerprint = new Map<string, T[]>();\n  for (const candidate of candidates) {\n    const group = byFingerprint.get(candidate.fingerprint) ?? [];\n    group.push(candidate);\n    byFingerprint.set(candidate.fingerprint, group);\n  }\n\n  const groups = [...byFingerprint.values()].map(dedupeByRegion).filter(isSurvivingGroup);\n  // Greedy order ranks by total coverage (region size × copies): a 3×3-statement group must beat\n  // a 2×4-statement group overlapping two of its copies, or the third copy is silently dropped\n  // and the reported duplication shrinks as more copies are added.\n  const groupSizeByFingerprint = new Map(groups.map((group) => [group[0]?.fingerprint ?? '', group.length]));\n  const coverage = (candidate: T): number =>\n    candidate.tokenCount * (groupSizeByFingerprint.get(candidate.fingerprint) ?? 1);\n  let duplicates = groups.flat();\n  duplicates.sort((left, right) => coverage(right) - coverage(left) || (compareTies ? compareTies(left, right) : 0));\n\n  // Greedy selection can keep a candidate whose group ends up below the survivor requirement;\n  // such an uncounted region must not block smaller groups, so the largest failed group is\n  // removed and the selection reruns. One group at a time: freeing a failed group's regions can\n  // rescue another. The rerun cap bounds degenerate inputs; past it the remaining failed groups\n  // are dropped, trading a sliver of recall on such files for bounded runtime.\n  for (let rerun = 0; ; rerun += 1) {\n    const keptRegionsByBucket = new Map<number, { startIndex: number; endIndex: number }[]>();\n    const counted = new Map<string, T[]>();\n    for (const candidate of duplicates) {\n      const keptRegions = keptRegionsByBucket.get(candidate.regionBucket ?? 0) ?? [];\n      if (\n        keptRegions.some((region) => region.startIndex < candidate.endIndex && candidate.startIndex < region.endIndex)\n      ) {\n        continue;\n      }\n      keptRegions.push(candidate);\n      keptRegionsByBucket.set(candidate.regionBucket ?? 0, keptRegions);\n      const group = counted.get(candidate.fingerprint) ?? [];\n      group.push(candidate);\n      counted.set(candidate.fingerprint, group);\n    }\n\n    let failedFingerprint: string | undefined;\n    let failedTokenCount = -1;\n    for (const [fingerprint, group] of counted) {\n      const tokenCount = group[0]?.tokenCount ?? 0;\n      if (!isSurvivingGroup(group) && tokenCount > failedTokenCount) {\n        failedFingerprint = fingerprint;\n        failedTokenCount = tokenCount;\n      }\n    }\n    // No failed fingerprint means every counted group met the survivor requirement.\n    if (failedFingerprint === undefined) {\n      return counted;\n    }\n    if (rerun >= maxSelectionRerunCount) {\n      for (const [fingerprint, group] of counted) {\n        if (!isSurvivingGroup(group)) {\n          counted.delete(fingerprint);\n        }\n      }\n      return counted;\n    }\n\n    duplicates = duplicates.filter((candidate) => candidate.fingerprint !== failedFingerprint);\n  }\n}\n\n/** Drops candidates covering the same source region (a block and the statement run spanning it). */\nexport function dedupeByRegion<T extends SelectableRegion>(group: T[]): T[] {\n  const byRegion = new Map<string, T>();\n  for (const candidate of group) {\n    const key = `${candidate.regionBucket ?? 0}:${candidate.startIndex}:${candidate.endIndex}`;\n    const existing = byRegion.get(key);\n    if (!existing || candidate.tokenCount > existing.tokenCount) {\n      byRegion.set(key, candidate);\n    }\n  }\n  return [...byRegion.values()];\n}\n"],"mappings":"AA2BA,SAAgB,EACd,EACA,EACA,EACkB,CAClB,IAAM,EAAgB,IAAI,IAC1B,IAAK,IAAM,KAAa,EAAY,CAClC,IAAM,EAAQ,EAAc,IAAI,EAAU,WAAW,GAAK,CAAC,EAC3D,EAAM,KAAK,CAAS,EACpB,EAAc,IAAI,EAAU,YAAa,CAAK,CAChD,CAEA,IAAM,EAAS,CAAC,GAAG,EAAc,OAAO,CAAC,CAAC,CAAC,IAAI,CAAc,CAAC,CAAC,OAAO,CAAgB,EAIhF,EAAyB,IAAI,IAAI,EAAO,IAAK,GAAU,CAAC,EAAM,EAAE,EAAE,aAAe,GAAI,EAAM,MAAM,CAAC,CAAC,EACnG,EAAY,GAChB,EAAU,YAAc,EAAuB,IAAI,EAAU,WAAW,GAAK,GAC3E,EAAa,EAAO,KAAK,EAC7B,EAAW,MAAM,EAAM,IAAU,EAAS,CAAK,EAAI,EAAS,CAAI,IAAM,EAAc,EAAY,EAAM,CAAK,EAAI,EAAE,EAOjH,IAAK,IAAI,EAAQ,GAAK,GAAS,EAAG,CAChC,IAAM,EAAsB,IAAI,IAC1B,EAAU,IAAI,IACpB,IAAK,IAAM,KAAa,EAAY,CAClC,IAAM,EAAc,EAAoB,IAAI,EAAU,cAAgB,CAAC,GAAK,CAAC,EAC7E,GACE,EAAY,KAAM,GAAW,EAAO,WAAa,EAAU,UAAY,EAAU,WAAa,EAAO,QAAQ,EAE7G,SAEF,EAAY,KAAK,CAAS,EAC1B,EAAoB,IAAI,EAAU,cAAgB,EAAG,CAAW,EAChE,IAAM,EAAQ,EAAQ,IAAI,EAAU,WAAW,GAAK,CAAC,EACrD,EAAM,KAAK,CAAS,EACpB,EAAQ,IAAI,EAAU,YAAa,CAAK,CAC1C,CAEA,IAAI,EACA,EAAmB,GACvB,IAAK,GAAM,CAAC,EAAa,KAAU,EAAS,CAC1C,IAAM,EAAa,EAAM,EAAE,EAAE,YAAc,EACvC,CAAC,EAAiB,CAAK,GAAK,EAAa,IAC3C,EAAoB,EACpB,EAAmB,EAEvB,CAEA,GAAI,IAAsB,IAAA,GACxB,OAAO,EAET,GAAI,GAAS,GAAwB,CACnC,IAAK,GAAM,CAAC,EAAa,KAAU,EAC5B,EAAiB,CAAK,GACzB,EAAQ,OAAO,CAAW,EAG9B,OAAO,CACT,CAEA,EAAa,EAAW,OAAQ,GAAc,EAAU,cAAgB,CAAiB,CAC3F,CACF,CAGA,SAAgB,EAA2C,EAAiB,CAC1E,IAAM,EAAW,IAAI,IACrB,IAAK,IAAM,KAAa,EAAO,CAC7B,IAAM,EAAM,GAAG,EAAU,cAAgB,EAAE,GAAG,EAAU,WAAW,GAAG,EAAU,WAC1E,EAAW,EAAS,IAAI,CAAG,GAC7B,CAAC,GAAY,EAAU,WAAa,EAAS,aAC/C,EAAS,IAAI,EAAK,CAAS,CAE/B,CACA,MAAO,CAAC,GAAG,EAAS,OAAO,CAAC,CAC9B"}
+{"version":3,"file":"duplicateSelection.js","names":[],"sources":["../src/duplicateSelection.ts"],"sourcesContent":["/**\n * Maximal, non-overlapping duplicate-group selection for the cross-file detector (the native\n * within-file detector mirrors its greedy ranking and shedding, but not the nested-copy retention\n * below, which is cross-file only). Candidates are grouped by fingerprint, ranked by total\n * coverage, kept greedily without overlapping a kept region, and groups that fall below the\n * survivor requirement are shed one at a time (largest first) so their regions stop blocking\n * smaller groups. A copy lying entirely inside a larger group's region stays with its group as a\n * nested copy (whichever group the greedy order kept first), so a standalone copy elsewhere is\n * still reported as duplicating it.\n */\n\nexport interface SelectableRegion {\n  fingerprint: string;\n  tokenCount: number;\n  startIndex: number;\n  endIndex: number;\n  /**\n   * Regions can only overlap within the same bucket. The within-file detector uses one bucket;\n   * the cross-file detector buckets by file index.\n   */\n  regionBucket?: number;\n  /**\n   * Set by selectMaximalGroups on a copy nested inside a larger group's region: it is reported with\n   * its group, but its span is already counted by that larger group.\n   */\n  nestedInLargerGroup?: boolean;\n}\n\n/** Caps how often the maximal-region selection reruns after shedding failed duplicate groups. */\nconst maxSelectionRerunCount = 20;\n\n/**\n * @param isSurvivingGroup whether a selected group counts (e.g. at least two occurrences, or\n *   occurrences spanning at least two files); failing groups are shed and re-selected without.\n * @param compareTies optional deterministic tie-break applied after the coverage ranking.\n */\nexport function selectMaximalGroups<T extends SelectableRegion>(\n  candidates: T[],\n  isSurvivingGroup: (group: T[]) => boolean,\n  compareTies?: (left: T, right: T) => number\n): Map<string, T[]> {\n  const byFingerprint = new Map<string, T[]>();\n  for (const candidate of candidates) {\n    const group = byFingerprint.get(candidate.fingerprint) ?? [];\n    group.push(candidate);\n    byFingerprint.set(candidate.fingerprint, group);\n  }\n\n  const groups = [...byFingerprint.values()].map(dedupeByRegion).filter(isSurvivingGroup);\n  // Greedy order ranks by total coverage (region size × copies): a 3×3-statement group must beat\n  // a 2×4-statement group overlapping two of its copies, or the third copy is silently dropped\n  // and the reported duplication shrinks as more copies are added.\n  const groupSizeByFingerprint = new Map(groups.map((group) => [group[0]?.fingerprint ?? '', group.length]));\n  const coverage = (candidate: T): number =>\n    candidate.tokenCount * (groupSizeByFingerprint.get(candidate.fingerprint) ?? 1);\n  let duplicates = groups.flat();\n  duplicates.sort((left, right) => coverage(right) - coverage(left) || (compareTies ? compareTies(left, right) : 0));\n\n  // Greedy selection can keep a candidate whose group ends up below the survivor requirement;\n  // such an uncounted region must not block smaller groups, so the largest failed group is\n  // removed and the selection reruns. One group at a time: freeing a failed group's regions can\n  // rescue another. The rerun cap bounds degenerate inputs; past it the remaining failed groups\n  // are dropped, trading a sliver of recall on such files for bounded runtime.\n  for (let rerun = 0; ; rerun += 1) {\n    const keptRegionsByBucket = new Map<number, T[]>();\n    const counted = new Map<string, T[]>();\n    const nestedByFingerprint = new Map<string, T[]>();\n    for (const candidate of duplicates) {\n      const keptRegions = keptRegionsByBucket.get(candidate.regionBucket ?? 0) ?? [];\n      // A plain loop: this runs once per candidate over every kept region of the bucket, so\n      // allocating a filtered array per candidate would dominate project-scale runs. Kept regions\n      // never overlap each other, so a candidate inside one cannot partially overlap another.\n      let containedInKept = false;\n      let partiallyOverlaps = false;\n      let enclosedKept: T[] | undefined;\n      for (const region of keptRegions) {\n        if (region.startIndex >= candidate.endIndex || candidate.startIndex >= region.endIndex) {\n          continue;\n        }\n        if (region.startIndex <= candidate.startIndex && candidate.endIndex <= region.endIndex) {\n          containedInKept = true;\n          break;\n        }\n        if (candidate.startIndex <= region.startIndex && region.endIndex <= candidate.endIndex) {\n          (enclosedKept ??= []).push(region);\n        } else {\n          partiallyOverlaps = true;\n          break;\n        }\n      }\n      if (containedInKept) {\n        const nested = nestedByFingerprint.get(candidate.fingerprint) ?? [];\n        nested.push({ ...candidate, nestedInLargerGroup: true });\n        nestedByFingerprint.set(candidate.fingerprint, nested);\n        continue;\n      }\n      if (partiallyOverlaps) {\n        continue;\n      }\n      // Containment must not depend on greedy order: a candidate enclosing kept copies of smaller\n      // groups occupies its region, and those copies become nested copies of their groups.\n      const enclosed = enclosedKept;\n      for (const inner of enclosed ?? []) {\n        const group = counted.get(inner.fingerprint) ?? [];\n        const index = group.indexOf(inner);\n        if (index !== -1) {\n          group[index] = { ...inner, nestedInLargerGroup: true };\n        }\n      }\n      // The enclosed regions give way to the enclosing one, keeping kept regions mutually\n      // non-overlapping: a later candidate inside this region must see it, not a region it\n      // swallowed (which the candidate could straddle instead).\n      const occupied = enclosed ? keptRegions.filter((region) => !enclosed.includes(region)) : keptRegions;\n      occupied.push(candidate);\n      keptRegionsByBucket.set(candidate.regionBucket ?? 0, occupied);\n      const group = counted.get(candidate.fingerprint) ?? [];\n      group.push(candidate);\n      counted.set(candidate.fingerprint, group);\n    }\n    // Nested copies join only a group that kept a standalone copy; on their own they would merely\n    // restate the larger group.\n    for (const [fingerprint, nested] of nestedByFingerprint) {\n      counted.get(fingerprint)?.push(...nested);\n    }\n    for (const [fingerprint, group] of counted) {\n      if (group.every((candidate) => candidate.nestedInLargerGroup)) {\n        counted.delete(fingerprint);\n      }\n    }\n\n    let failedFingerprint: string | undefined;\n    let failedTokenCount = -1;\n    for (const [fingerprint, group] of counted) {\n      const tokenCount = group[0]?.tokenCount ?? 0;\n      if (!isSurvivingGroup(group) && tokenCount > failedTokenCount) {\n        failedFingerprint = fingerprint;\n        failedTokenCount = tokenCount;\n      }\n    }\n    // No failed fingerprint means every counted group met the survivor requirement.\n    if (failedFingerprint === undefined) {\n      return counted;\n    }\n    if (rerun >= maxSelectionRerunCount) {\n      dropFailedGroups(counted, isSurvivingGroup);\n      return counted;\n    }\n\n    duplicates = duplicates.filter((candidate) => candidate.fingerprint !== failedFingerprint);\n  }\n}\n\n/**\n * Past the rerun cap, still-failing groups are dropped without another selection pass. A dropped\n * group's regions may have been what nested copies of surviving groups lay inside, and such a copy\n * would then be counted by no group at all, so those copies are dropped too and the shrunk groups\n * are re-checked until nothing changes.\n */\nfunction dropFailedGroups<T extends SelectableRegion>(\n  counted: Map<string, T[]>,\n  isSurvivingGroup: (group: T[]) => boolean\n): void {\n  for (let changed = true; changed;) {\n    changed = false;\n    for (const [fingerprint, group] of counted) {\n      if (!isSurvivingGroup(group) || group.every((candidate) => candidate.nestedInLargerGroup)) {\n        counted.delete(fingerprint);\n        changed = true;\n      }\n    }\n    const standalone = [...counted.values()].flat().filter((candidate) => !candidate.nestedInLargerGroup);\n    for (const [fingerprint, group] of counted) {\n      const kept = group.filter(\n        (candidate) =>\n          !candidate.nestedInLargerGroup ||\n          standalone.some(\n            (region) =>\n              (region.regionBucket ?? 0) === (candidate.regionBucket ?? 0) &&\n              region.startIndex <= candidate.startIndex &&\n              candidate.endIndex <= region.endIndex\n          )\n      );\n      if (kept.length !== group.length) {\n        counted.set(fingerprint, kept);\n        changed = true;\n      }\n    }\n  }\n}\n\n/** Drops candidates covering the same source region (a block and the statement run spanning it). */\nexport function dedupeByRegion<T extends SelectableRegion>(group: T[]): T[] {\n  const byRegion = new Map<string, T>();\n  for (const candidate of group) {\n    const key = `${candidate.regionBucket ?? 0}:${candidate.startIndex}:${candidate.endIndex}`;\n    const existing = byRegion.get(key);\n    if (!existing || candidate.tokenCount > existing.tokenCount) {\n      byRegion.set(key, candidate);\n    }\n  }\n  return [...byRegion.values()];\n}\n"],"mappings":"AAoCA,SAAgB,EACd,EACA,EACA,EACkB,CAClB,IAAM,EAAgB,IAAI,IAC1B,IAAK,IAAM,KAAa,EAAY,CAClC,IAAM,EAAQ,EAAc,IAAI,EAAU,WAAW,GAAK,CAAC,EAC3D,EAAM,KAAK,CAAS,EACpB,EAAc,IAAI,EAAU,YAAa,CAAK,CAChD,CAEA,IAAM,EAAS,CAAC,GAAG,EAAc,OAAO,CAAC,CAAC,CAAC,IAAI,CAAc,CAAC,CAAC,OAAO,CAAgB,EAIhF,EAAyB,IAAI,IAAI,EAAO,IAAK,GAAU,CAAC,EAAM,EAAE,EAAE,aAAe,GAAI,EAAM,MAAM,CAAC,CAAC,EACnG,EAAY,GAChB,EAAU,YAAc,EAAuB,IAAI,EAAU,WAAW,GAAK,GAC3E,EAAa,EAAO,KAAK,EAC7B,EAAW,MAAM,EAAM,IAAU,EAAS,CAAK,EAAI,EAAS,CAAI,IAAM,EAAc,EAAY,EAAM,CAAK,EAAI,EAAE,EAOjH,IAAK,IAAI,EAAQ,GAAK,GAAS,EAAG,CAChC,IAAM,EAAsB,IAAI,IAC1B,EAAU,IAAI,IACd,EAAsB,IAAI,IAChC,IAAK,IAAM,KAAa,EAAY,CAClC,IAAM,EAAc,EAAoB,IAAI,EAAU,cAAgB,CAAC,GAAK,CAAC,EAIzE,EAAkB,GAClB,EAAoB,GACpB,EACJ,IAAK,IAAM,KAAU,EACf,OAAO,YAAc,EAAU,UAAY,EAAU,YAAc,EAAO,UAG9E,IAAI,EAAO,YAAc,EAAU,YAAc,EAAU,UAAY,EAAO,SAAU,CACtF,EAAkB,GAClB,KACF,CACA,GAAI,EAAU,YAAc,EAAO,YAAc,EAAO,UAAY,EAAU,UAC3E,IAAiB,CAAC,EAAA,CAAG,KAAK,CAAM,MAC5B,CACL,EAAoB,GACpB,KACF,CANA,CAQF,GAAI,EAAiB,CACnB,IAAM,EAAS,EAAoB,IAAI,EAAU,WAAW,GAAK,CAAC,EAClE,EAAO,KAAK,CAAE,GAAG,EAAW,oBAAqB,EAAK,CAAC,EACvD,EAAoB,IAAI,EAAU,YAAa,CAAM,EACrD,QACF,CACA,GAAI,EACF,SAIF,IAAM,EAAW,EACjB,IAAK,IAAM,KAAS,GAAY,CAAC,EAAG,CAClC,IAAM,EAAQ,EAAQ,IAAI,EAAM,WAAW,GAAK,CAAC,EAC3C,EAAQ,EAAM,QAAQ,CAAK,EAC7B,IAAU,KACZ,EAAM,GAAS,CAAE,GAAG,EAAO,oBAAqB,EAAK,EAEzD,CAIA,IAAM,EAAW,EAAW,EAAY,OAAQ,GAAW,CAAC,EAAS,SAAS,CAAM,CAAC,EAAI,EACzF,EAAS,KAAK,CAAS,EACvB,EAAoB,IAAI,EAAU,cAAgB,EAAG,CAAQ,EAC7D,IAAM,EAAQ,EAAQ,IAAI,EAAU,WAAW,GAAK,CAAC,EACrD,EAAM,KAAK,CAAS,EACpB,EAAQ,IAAI,EAAU,YAAa,CAAK,CAC1C,CAGA,IAAK,GAAM,CAAC,EAAa,KAAW,EAClC,EAAQ,IAAI,CAAW,CAAC,EAAE,KAAK,GAAG,CAAM,EAE1C,IAAK,GAAM,CAAC,EAAa,KAAU,EAC7B,EAAM,MAAO,GAAc,EAAU,mBAAmB,GAC1D,EAAQ,OAAO,CAAW,EAI9B,IAAI,EACA,EAAmB,GACvB,IAAK,GAAM,CAAC,EAAa,KAAU,EAAS,CAC1C,IAAM,EAAa,EAAM,EAAE,EAAE,YAAc,EACvC,CAAC,EAAiB,CAAK,GAAK,EAAa,IAC3C,EAAoB,EACpB,EAAmB,EAEvB,CAEA,GAAI,IAAsB,IAAA,GACxB,OAAO,EAET,GAAI,GAAS,GAEX,OADA,EAAiB,EAAS,CAAgB,EACnC,EAGT,EAAa,EAAW,OAAQ,GAAc,EAAU,cAAgB,CAAiB,CAC3F,CACF,CAQA,SAAS,EACP,EACA,EACM,CACN,IAAK,IAAI,EAAU,GAAM,GAAU,CACjC,EAAU,GACV,IAAK,GAAM,CAAC,EAAa,KAAU,GAC7B,CAAC,EAAiB,CAAK,GAAK,EAAM,MAAO,GAAc,EAAU,mBAAmB,KACtF,EAAQ,OAAO,CAAW,EAC1B,EAAU,IAGd,IAAM,EAAa,CAAC,GAAG,EAAQ,OAAO,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,OAAQ,GAAc,CAAC,EAAU,mBAAmB,EACpG,IAAK,GAAM,CAAC,EAAa,KAAU,EAAS,CAC1C,IAAM,EAAO,EAAM,OAChB,GACC,CAAC,EAAU,qBACX,EAAW,KACR,IACE,EAAO,cAAgB,MAAQ,EAAU,cAAgB,IAC1D,EAAO,YAAc,EAAU,YAC/B,EAAU,UAAY,EAAO,QACjC,CACJ,EACI,EAAK,SAAW,EAAM,SACxB,EAAQ,IAAI,EAAa,CAAI,EAC7B,EAAU,GAEd,CACF,CACF,CAGA,SAAgB,EAA2C,EAAiB,CAC1E,IAAM,EAAW,IAAI,IACrB,IAAK,IAAM,KAAa,EAAO,CAC7B,IAAM,EAAM,GAAG,EAAU,cAAgB,EAAE,GAAG,EAAU,WAAW,GAAG,EAAU,WAC1E,EAAW,EAAS,IAAI,CAAG,GAC7B,CAAC,GAAY,EAAU,WAAa,EAAS,aAC/C,EAAS,IAAI,EAAK,CAAS,CAE/B,CACA,MAAO,CAAC,GAAG,EAAS,OAAO,CAAC,CAC9B"}

package/dist/duplication.cjs

@@ -1,2 +1,2 @@
-"use strict";const e={minTokens:40,maxGapTokens:30,minSimilarityPercent:70};function t(t){return{minTokens:n(t?.minTokens,e.minTokens),maxGapTokens:n(t?.maxGapTokens,e.maxGapTokens),minSimilarityPercent:n(t?.minSimilarityPercent,e.minSimilarityPercent)}}function n(e,t){return e===void 0||Number.isNaN(e)?t:Math.min(Math.max(Math.trunc(e),0),4294967295)}function r(e,t){return e*5>=t}function i(e){let t=new Int32Array(e.length+1);for(let[n,r]of e.entries())t[n+1]=(t[n]??0)+(r.literalHash===void 0?0:1);return t}function a(e,t,n){let r=[],i=[],a=[];for(let[t,n]of e.entries())for(let e of n.containers)i.push(t),a.push(e);let c=t=>e[i[t]??0],l=new Map,d=a.map((e,n)=>s(c(n)?.tokens??[],e,t));for(let[e,t]of d.entries()){let n=i[e]??0;for(let[r,i]of t.windowKeysByStart.entries())for(let t of i){if(t===void 0)continue;let i=l.get(t);i?(i.count+=1,i.containerIndex!==e&&(i.containerIndex=-1),i.contextIndex!==n&&(i.contextIndex=-1),i.minStart=Math.min(i.minStart,r),i.maxStart=Math.max(i.maxStart,r)):l.set(t,{count:1,containerIndex:e,contextIndex:n,minStart:r,maxStart:r})}}let f=(e,t)=>{if(e===void 0)return!1;let r=l.get(e);return r===void 0||r.count<2?!1:n?r.contextIndex===-1:r.containerIndex===-1||r.maxStart-r.minStart>=t},p=e=>{let t=d[e.containerIndex]?.statementHashes??[],n=t[e.start];for(let r=e.start+1;r<e.start+e.length;r+=1)if(t[r]!==n)return!0;return!1},m=[];for(let[e,t]of d.entries())for(let[n,r]of t.windowKeysByStart.entries())for(let[i,a]of r.entries()){if(!f(a,i)||!p({containerIndex:e,start:n,length:i}))continue;let r=t.windowKeysByStart[n]?.[i+1],o=t.windowKeysByStart[n-1]?.[i+1];f(r,i+1)||f(o,i+1)||m.push({containerIndex:e,start:n,length:i})}let g=new Set(m.map(o)),_=m;for(;_.length>0;){let e=[];for(let t of _){let n=a[t.containerIndex],o=n?.[t.start],s=n?.[t.start+t.length-1],l=c(t.containerIndex);if(!o||!s||!l)continue;let d=`s:${h(l.tokens,l.literalCountPrefix,o.startTokenIndex,s.endTokenIndex)}`;r.push({candidate:u(d,o.startTokenIndex,s.endTokenIndex,o,s),contextIndex:i[t.containerIndex]??0}),e.push(t)}_=[];for(let t of e)for(let e of[t.start,t.start+1]){let n={containerIndex:t.containerIndex,start:e,length:t.length-1},r=d[t.containerIndex]?.windowKeysByStart[e]?.[n.length];g.has(o(n))||!f(r,n.length)||!p(n)||(g.add(o(n)),_.push(n))}}return r}function o(e){return`${e.containerIndex}:${e.start}:${e.length}`}function s(e,t,n){let r=t.map(t=>g(e,t.startTokenIndex,t.endTokenIndex)),i=[];for(let e=0;e<t.length;e+=1){let a=[],o=5381,s=0,c=Math.min(t.length,e+100);for(let i=e;i<c;i+=1){let c=t[i],l=r[i];if(!c||l===void 0)break;o=b(o,l),s+=c.endTokenIndex-c.startTokenIndex;let u=i-e+1;a[u]=u>=2&&s>=n?b(o,u):void 0}i.push(a)}return{windowKeysByStart:i,statementHashes:r}}function c(e,t){let n=e.length+31>>>5,r=new Map;for(let[t,i]of e.entries()){let e=r.get(i);e||(e=new Uint32Array(n),r.set(i,e));let a=t>>>5;e[a]=(e[a]??0)|1<<(t&31)}let i=new Uint32Array(n);for(let e of t){let t=r.get(e),a=1,o=0;for(let e=0;e<n;e+=1){let n=i[e]??0,r=((t?.[e]??0)|n)>>>0,s=(n<<1|a)>>>0;a=n>>>31;let c=r-s-o;o=+(c<0),i[e]=r&~c}}let a=0;for(let e of i)a+=l(e);return a}function l(e){let t=e-(e>>>1&1431655765);return t=(t&858993459)+(t>>>2&858993459),Math.imul(t+(t>>>4)&252645135,16843009)>>>24&255}function u(e,t,n,r,i){return{fingerprint:e,tokenCount:n-t,startTokenIndex:t,endTokenIndex:n,startIndex:r.startIndex,endIndex:i.endIndex,startLine:r.startLine,endLine:i.endLine}}const d=[],f=[];function p(e){let t=d[e];return t===void 0&&(t=v(`$${e}`),d[e]=t),t}function m(e){let t=f[e];return t===void 0&&(t=y(`$${e}`),f[e]=t),t}function h(e,t,n,i){let[a,o]=_(e,n,i,r((t[i]??0)-(t[n]??0),i-n));return`${a}:${o}:${i-n}`}function g(e,t,n){let[r,i]=_(e,t,n,!1);return r^Math.imul(i,31)}function _(e,t,n,r){let i=new Map,a=5381,o=52711;for(let s=t;s<n;s+=1){let t=e[s];if(!t)continue;let n,c;if(t.kind===`id`){let e=i.get(t.text);e===void 0&&(e=i.size,i.set(t.text,e)),n=p(e),c=m(e)}else n=t.textHash,c=t.textHash2;a=Math.imul(a,31)+n|0,o=Math.imul(o,37)^c,r&&t.literalHash!==void 0&&t.literalHash2!==void 0&&(a=Math.imul(a,31)+t.literalHash|0,o=Math.imul(o,37)^t.literalHash2)}return[a,o]}function v(e){let t=5381;for(let n=0;n<e.length;n+=1)t=Math.imul(t,33)^e.charCodeAt(n);return t}function y(e){let t=-2128831035;for(let n=0;n<e.length;n+=1)t=Math.imul(t^e.charCodeAt(n),16777619);return t}function b(e,t){return Math.imul(e,31)+t}function x(e,t){if(t<=0||e.length<2)return e;e.sort(S);for(let n=!0;n;){n=!1;for(let r=0;r<e.length&&!n;r+=1)for(let i=r+1;i<e.length;i+=1){let a=e[r],o=e[i];if(!a||!o)continue;let s=C(a,o,t),c=s??C(o,a,t);if(!c)continue;let l=s?c.firstConsumed:c.secondConsumed,u=s?c.secondConsumed:c.firstConsumed;l&&u?(e[r]=c.merged,e.splice(i,1)):u?e[i]=c.merged:e[r]=c.merged;for(let e of c.pairedRetained)e.sharedWithMergedGroup=!0;e.sort(S),n=!0;break}}return e}function S(e,t){let n=e[0],r=t[0];return(n?.startTokenIndex??0)-(r?.startTokenIndex??0)||(n?.endTokenIndex??0)-(r?.endTokenIndex??0)}function C(e,t,n){let r=e.filter(e=>!e.sharedWithMergedGroup),i=t.filter(e=>!e.sharedWithMergedGroup),a=[],o=0,s=-1;for(let e of i){for(;o<r.length;){let t=r[o];if(t&&t.endTokenIndex+n<e.startTokenIndex)o+=1;else break}let t=r[o];t&&t.endTokenIndex<=e.startTokenIndex&&t.startTokenIndex>=s&&(a.push([t,e]),s=e.endTokenIndex,o+=1)}let c=a.length===e.length,l=a.length===t.length;if(!(a.length<2||!c&&!l))return{merged:a.map(([e,t])=>({...e,sharedWithMergedGroup:void 0,segments:[...e.segments,...t.segments],tokenCount:e.tokenCount+t.tokenCount,endTokenIndex:t.endTokenIndex,endIndex:t.endIndex,endLine:t.endLine})),firstConsumed:c,secondConsumed:l,pairedRetained:c===l?[]:a.map(([e,t])=>c?t:e)}}function w(e){let t=0,n=0,r=!1;for(let i of e){if(i.sharedWithMergedGroup){r=!0;continue}t+=i.segments.length,n=Math.max(n,i.segments.length)}return r?t:t-n}function T(e,t,n,r){for(let i=e.startTokenIndex;i<e.endTokenIndex;i+=1){let e=t[i];for(let t=e?.startRow??0;t<=(e?.endRow??-1);t+=1)(!n||n.has(t+1))&&r.add(t+1)}}exports.buildLiteralCountPrefix=i,exports.collectSegmentLines=T,exports.collectSequenceWindowCandidates=a,exports.countRedundantFragments=w,exports.defaultDuplicationOptions=e,exports.lcsLength=c,exports.mergeAdjacentGroups=x,exports.resolveDuplicationOptions=t;
+"use strict";const e={minTokens:40,maxGapTokens:30,minSimilarityPercent:70};function t(t){return{minTokens:n(t?.minTokens,e.minTokens),maxGapTokens:n(t?.maxGapTokens,e.maxGapTokens),minSimilarityPercent:n(t?.minSimilarityPercent,e.minSimilarityPercent)}}function n(e,t){return e===void 0||Number.isNaN(e)?t:Math.min(Math.max(Math.trunc(e),0),4294967295)}function r(e,t){return e*5>=t}function i(e){let t=new Int32Array(e.length+1);for(let[n,r]of e.entries())t[n+1]=(t[n]??0)+(r.literalHash===void 0?0:1);return t}function a(e,t,n){let r=[],i=[],a=[];for(let[t,n]of e.entries())for(let e of n.containers)i.push(t),a.push(e);let c=t=>e[i[t]??0],l=new Map,d=a.map((e,n)=>s(c(n)?.tokens??[],e,t));for(let[e,t]of d.entries()){let n=i[e]??0;for(let[r,i]of t.windowKeysByStart.entries())for(let t of i){if(t===void 0)continue;let i=l.get(t);i?(i.count+=1,i.containerIndex!==e&&(i.containerIndex=-1),i.contextIndex!==n&&(i.contextIndex=-1),i.minStart=Math.min(i.minStart,r),i.maxStart=Math.max(i.maxStart,r)):l.set(t,{count:1,containerIndex:e,contextIndex:n,minStart:r,maxStart:r})}}let f=(e,t)=>{if(e===void 0)return!1;let r=l.get(e);return r===void 0||r.count<2?!1:n?r.contextIndex===-1:r.containerIndex===-1||r.maxStart-r.minStart>=t},p=e=>{let t=d[e.containerIndex]?.statementHashes??[],n=t[e.start];for(let r=e.start+1;r<e.start+e.length;r+=1)if(t[r]!==n)return!0;return!1},m=[];for(let[e,t]of d.entries())for(let[n,r]of t.windowKeysByStart.entries())for(let[i,a]of r.entries()){if(!f(a,i)||!p({containerIndex:e,start:n,length:i}))continue;let r=t.windowKeysByStart[n]?.[i+1],o=t.windowKeysByStart[n-1]?.[i+1];f(r,i+1)||f(o,i+1)||m.push({containerIndex:e,start:n,length:i})}let g=new Set(m.map(o)),_=m;for(;_.length>0;){let e=[];for(let t of _){let n=a[t.containerIndex],o=n?.[t.start],s=n?.[t.start+t.length-1],l=c(t.containerIndex);if(!o||!s||!l)continue;let d=`s:${h(l.tokens,l.literalCountPrefix,o.startTokenIndex,s.endTokenIndex)}`;r.push({candidate:u(d,o.startTokenIndex,s.endTokenIndex,o,s),contextIndex:i[t.containerIndex]??0}),e.push(t)}_=[];for(let t of e)for(let e of[t.start,t.start+1]){let n={containerIndex:t.containerIndex,start:e,length:t.length-1},r=d[t.containerIndex]?.windowKeysByStart[e]?.[n.length];g.has(o(n))||!f(r,n.length)||!p(n)||(g.add(o(n)),_.push(n))}}return r}function o(e){return`${e.containerIndex}:${e.start}:${e.length}`}function s(e,t,n){let r=t.map(t=>g(e,t.startTokenIndex,t.endTokenIndex)),i=[];for(let e=0;e<t.length;e+=1){let a=[],o=5381,s=0,c=Math.min(t.length,e+100);for(let i=e;i<c;i+=1){let c=t[i],l=r[i];if(!c||l===void 0)break;o=b(o,l),s+=c.endTokenIndex-c.startTokenIndex;let u=i-e+1;a[u]=u>=2&&s>=n?b(o,u):void 0}i.push(a)}return{windowKeysByStart:i,statementHashes:r}}function c(e,t){let n=e.length+31>>>5,r=new Map;for(let[t,i]of e.entries()){let e=r.get(i);e||(e=new Uint32Array(n),r.set(i,e));let a=t>>>5;e[a]=(e[a]??0)|1<<(t&31)}let i=new Uint32Array(n);for(let e of t){let t=r.get(e),a=1,o=0;for(let e=0;e<n;e+=1){let n=i[e]??0,r=((t?.[e]??0)|n)>>>0,s=(n<<1|a)>>>0;a=n>>>31;let c=r-s-o;o=+(c<0),i[e]=r&~c}}let a=0;for(let e of i)a+=l(e);return a}function l(e){let t=e-(e>>>1&1431655765);return t=(t&858993459)+(t>>>2&858993459),Math.imul(t+(t>>>4)&252645135,16843009)>>>24&255}function u(e,t,n,r,i){return{fingerprint:e,tokenCount:n-t,startTokenIndex:t,endTokenIndex:n,startIndex:r.startIndex,endIndex:i.endIndex,startLine:r.startLine,endLine:i.endLine}}const d=[],f=[];function p(e){let t=d[e];return t===void 0&&(t=v(`$${e}`),d[e]=t),t}function m(e){let t=f[e];return t===void 0&&(t=y(`$${e}`),f[e]=t),t}function h(e,t,n,i){let[a,o]=_(e,n,i,r((t[i]??0)-(t[n]??0),i-n));return`${a}:${o}:${i-n}`}function g(e,t,n){let[r,i]=_(e,t,n,!1);return r^Math.imul(i,31)}function _(e,t,n,r){let i=new Map,a=5381,o=52711;for(let s=t;s<n;s+=1){let t=e[s];if(!t)continue;let n,c;if(t.kind===`id`){let e=i.get(t.text);e===void 0&&(e=i.size,i.set(t.text,e)),n=p(e),c=m(e)}else n=t.textHash,c=t.textHash2;a=Math.imul(a,31)+n|0,o=Math.imul(o,37)^c,r&&t.literalHash!==void 0&&t.literalHash2!==void 0&&(a=Math.imul(a,31)+t.literalHash|0,o=Math.imul(o,37)^t.literalHash2)}return[a,o]}function v(e){let t=5381;for(let n=0;n<e.length;n+=1)t=Math.imul(t,33)^e.charCodeAt(n);return t}function y(e){let t=-2128831035;for(let n=0;n<e.length;n+=1)t=Math.imul(t^e.charCodeAt(n),16777619);return t}function b(e,t){return Math.imul(e,31)+t}function x(e,t,n=()=>!0){if(t<=0||e.length<2)return e;e.sort(S);for(let r=!0;r;){r=!1;for(let i=0;i<e.length&&!r;i+=1)for(let a=i+1;a<e.length;a+=1){let o=e[i],s=e[a];if(!o||!s)continue;let c=C(o,s,t,n),l=c??C(s,o,t,n);if(!l)continue;let u=c?l.firstReplaced:l.secondReplaced,d=c?l.secondReplaced:l.firstReplaced;u&&d?(e[i]=l.merged,e.splice(a,1)):d?e[a]=l.merged:u?e[i]=l.merged:e.push(l.merged);for(let e of l.pairedRetained)e.spanCountedElsewhere=!0;e.sort(S),r=!0;break}}return e}function S(e,t){let n=e[0],r=t[0];return(n?.startTokenIndex??0)-(r?.startTokenIndex??0)||(n?.endTokenIndex??0)-(r?.endTokenIndex??0)}function C(e,t,n,r){let[i,a]=w(e),[o,s]=w(t),c=[],l=0,u=-1;for(let e of o){for(;l<i.length;){let t=i[l];if(t&&t.endTokenIndex+n<e.startTokenIndex)l+=1;else break}let t=i[l];t&&t.endTokenIndex<=e.startTokenIndex&&t.startTokenIndex>=u&&(c.push([t,e]),u=e.endTokenIndex,l+=1)}let d=c.length===a,f=c.length===s;if(c.length<2||!d&&!f)return;let p=d&&!e.some(e=>e.nestedInLargerGroup),m=f&&!t.some(e=>e.nestedInLargerGroup),h=c.map(([e,t])=>({...e,spanCountedElsewhere:void 0,nestedInLargerGroup:void 0,segments:[...e.segments,...t.segments],tokenCount:e.tokenCount+t.tokenCount,endTokenIndex:t.endTokenIndex,endIndex:t.endIndex,endLine:t.endLine}));if(r(h))return{merged:h,firstReplaced:p,secondReplaced:m,pairedRetained:[...p?[]:c.map(([e])=>e),...m?[]:c.map(([,e])=>e)]}}function w(e){let t=[],n=0;for(let r of e)r.nestedInLargerGroup||(n+=1),!r.spanCountedElsewhere&&!r.nestedInLargerGroup&&t.push(r);return[t,n]}function T(e){let t=0,n=0,r=!1;for(let i of e){if(i.spanCountedElsewhere){r=!0;continue}t+=i.segments.length,n=Math.max(n,i.segments.length)}return r?t:t-n}function E(e,t,n,r){for(let i=e.startTokenIndex;i<e.endTokenIndex;i+=1){let e=t[i];for(let t=e?.startRow??0;t<=(e?.endRow??-1);t+=1)(!n||n.has(t+1))&&r.add(t+1)}}exports.buildLiteralCountPrefix=i,exports.collectSegmentLines=E,exports.collectSequenceWindowCandidates=a,exports.countRedundantFragments=T,exports.defaultDuplicationOptions=e,exports.lcsLength=c,exports.mergeAdjacentGroups=x,exports.resolveDuplicationOptions=t;
 //# sourceMappingURL=duplication.cjs.map