Binary/Text Diff & Similarity Ratio

Inspectable records

Ledger state

The current ledger marks this instrument verified clean within the source-bound verification corpus.

Area

Computing & Information

Tranche

tranche-5

Evidence ID

t5-089

Ledger rows

1

No ledger note is attached.

Review findings

The Levenshtein formula is standard only after defining the distance metric, length unit, percentage conversion, and empty-input behavior. The specification is NEEDS_REVIEW because the selectable LCS variant and the diff-generation rules are materially ambiguous.

Formula and scope findings

• The edit_distance metric is not explicitly defined; the formula requires standard Levenshtein distance if that is intended.
• The formula returns a ratio from 0 to 1, while the stated output is similarity percent; the implementation must multiply by 100 for percent output.
• The LCS-based ratio variant has no formula. Common definitions include LCS/max(len1,len2) and 2*LCS/(len1+len2), which produce different results.
• The length unit is unspecified: character count, Unicode code-point count, UTF-16 code-unit count, byte count, and grapheme-cluster count can differ.
• The formula does not define how inline or line-level diff highlighting is generated, including edit-script tie-breaking and newline handling.
• The formula is not defined for two empty inputs because max(len1,len2) is zero.

Unit assumptions

• Test vectors assume standard character-level Levenshtein distance with insertion, deletion, and substitution each costing 1.
• len1 and len2 are assumed to be counts of Unicode characters/code points, not bytes or UTF-16 code units.
• The similarity ratio is assumed to be converted to a percentage by multiplying by 100.
• The displayed inline diff uses a canonical minimal edit script; exact highlighting may vary unless tie-breaking is specified.
• For the LCS variant, no unit assumption can resolve the missing ratio definition.

Boundary hardening

BREAK_CLAIMED · SELF_FLAGGED

The specified formula is undefined for two empty text blocks because max(len1,len2)=0, causing division by zero. The result is also convention-sensitive because case handling is unspecified, and the available LCS-based ratio can produce a different score from the Levenshtein formula.

Errata and corrections

No applied erratum is attached to this unique-name record.

Inspectable example

4 review vectors are recorded. One source-bound sample is shown below.

{
  "inputs": {
    "text1": "cat",
    "text2": "cat"
  },
  "expected_outputs": {
    "edit_distance": 0,
    "similarity_percent": 100,
    "inline_diff": [
      {
        "op": "equal",
        "text": "cat"
      }
    ]
  },
  "working": "len1=3, len2=3; max(3,3)=3. Levenshtein distance=0 because the strings are identical. similarity=1-0/3=1. percent=1*100=100%. Diff: all 3 characters are equal."
}

Related workflows

No source-defined multi-instrument workflow membership is recorded.

Sources and evidence

These pointers bind the page to evidence. They do not grant authority or collapse verification into deployment.

• VERIFIED_LEDGER.json · VERIFIED_CLEAN
• evidence-tranche5/tranche5-results.json
• BOUNDARY_HAZARD_CATALOG.json