Output & Report

中文

dup-code-check supports both text output and JSON output. Text is for humans; JSON is for post-processing and CI integration.

1) Duplicate files (default mode)

Text

You’ll see:

• duplicate groups: <N>
• for each group:
  • hash=<...> normalized_len=<...> files=<...>
  • - [repoLabel] path

JSON (--json)

JSON output is an array, each element:

interface DuplicateGroup {
  hash: string;          // 16 hex chars (FNV-1a 64)
  normalizedLen: number; // byte length after ASCII whitespace removal
  files: { repoId: number; repoLabel: string; path: string }[];
}

2) Suspected duplicate code spans (--code-spans)

Text

• duplicate code span groups: <N>
• per group:
  • hash=<...> normalized_len=<...> occurrences=<...>
  • preview=<...>
  • - [repoLabel] path:startLine-endLine

JSON (--json)

JSON output is an array, each element:

interface DuplicateSpanGroup {
  hash: string;
  normalizedLen: number;
  preview: string;
  occurrences: {
    repoId: number;
    repoLabel: string;
    path: string;
    startLine: number;
    endLine: number;
  }[];
}

3) Scan stats (--stats)

JSON mode

With --json --stats:

• default / --code-spans: { groups, scanStats }
• --report: { report, scanStats }

scanStats fields include:

• candidateFiles, scannedFiles, scannedBytes
• gitFastPathFallbacks: non-zero when the scan attempted the Git fast path and had to fall back to the filesystem walker
• skippedNotFound, skippedPermissionDenied, skippedTooLarge, skippedBinary, skippedOutsideRoot, skippedRelativizeFailed, skippedWalkErrors
• skippedOutsideRoot: paths outside roots or unsafe paths (e.g. symlink targets outside roots, or unsafe paths emitted by the Git fast path)
• skippedBudgetMaxFiles: non-zero when the scan stopped early due to the maxFiles budget
• skippedBudgetMaxTotalBytes: skipped due to maxTotalBytes (reading would exceed the total bytes budget)
• skippedBudgetMaxNormalizedChars: non-zero when the scan stopped early due to the maxNormalizedChars budget
• skippedBudgetMaxTokens: non-zero when the scan stopped early due to the maxTokens budget (report mode)
• skippedBucketTruncated: detector guardrail; fingerprint buckets were truncated to cap worst-case cost (results may miss some matches)

Text mode

In text mode, --stats prints stats to stderr while keeping results on stdout:

dup-code-check --stats . >result.txt 2>stats.txt

4) Strict mode (--strict)

--strict is intended for CI and answers “was the scan complete?”:

• exits 1 on PermissionDenied, outside_root, relativize_failed, traversal errors, bucket truncation, or budget limits (maxFiles / maxTotalBytes / maxNormalizedChars / maxTokens)
• does not fail on NotFound, TooLarge, or Binary

When --json is enabled and --stats is not, --strict still prints stats to stderr on failure (so you can see why).

5) Report mode (--report)

Text output contains multiple sections (in this order):

1. file duplicates
2. code span duplicates
3. line span duplicates
4. token span duplicates
5. block duplicates
6. AST subtree duplicates
7. similar blocks (minhash)
8. similar blocks (simhash)

JSON output:

interface DuplicationReport {
  fileDuplicates: DuplicateGroup[];
  codeSpanDuplicates: DuplicateSpanGroup[];
  lineSpanDuplicates: DuplicateSpanGroup[];
  tokenSpanDuplicates: DuplicateSpanGroup[];
  blockDuplicates: DuplicateSpanGroup[];
  astSubtreeDuplicates: DuplicateSpanGroup[];
  similarBlocksMinhash: SimilarityPair[];
  similarBlocksSimhash: SimilarityPair[];
}

For the meaning/implementation ideas of each section, see Detectors & Algorithms.