Anchored Ratio

A 3‑way similarity metric anchored_ratio(O, R, R_hat) producing a score in [0,1] comparing two edited versions (R, R̂) relative to a base O.

• Denominator normalises over the union of changed base intervals plus insertion slots.
• Numerator rewards aligned equal lines (and optionally per-line Levenshtein similarity) within each union block and insertion slot.

from conflict_collection.metrics.anchored_ratio import anchored_ratio
score = anchored_ratio(O, R, R_hat, use_line_levenshtein=True)

If neither side changes anything (no base changes, no insertions) the score is defined as 1.0.

When to Use

Useful for measuring convergence of independent resolution attempts, or similarity between automated and manual merges.

Rationale

Traditional pairwise diffs ignore the role of the base. Anchoring both transformations to O allows fairer comparison of expansion/compression asymmetries and insertion slots.

API

See function reference.