Python API#

Every public name lives in scikitplot.levenshtein. The generated reference pages carry the full signatures; this page says what each one is for.

Metrics#

distance(a, b, *, backend="auto", strict=False) -> int: Unit-cost Levenshtein edit distance between two strings, byte strings or sequences.
normalized_distance(a, b, ...) -> float: distance / max(len(a), len(b)) in [0, 1]; two empty inputs give 0.0.
normalized_similarity(a, b, ...) -> float: 1 - normalized_distance; two empty inputs give 1.0.
similarity(a, b, ...) -> int: max(len(a), len(b)) - distance(a, b).

rank(query, choices, *, key=None, limit=None, score_cutoff=0.0, backend="auto", strict=False) -> list[Match]: Every choice ordered by similarity, then distance, then input position.
closest(query, choices, ...) -> Match or None: The first result of rank with limit=1.
Match: One ranked result: choice, distance, similarity, index and the backend that actually computed the distance.

backend_info(name="auto") -> BackendInfo: Capability record of one backend; "auto" describes the backend that would be used now.
available_backends(*, include_gpl=True) -> tuple[BackendInfo, ...]: Records for internal, rapidfuzz, levenshtein (unless include_gpl=False) and python, in preference order.
BackendInfo: Read-only record: name, available, version, license, implementation and, when unavailable, reason. to_dict() returns it as a mapping.

This module does not import Corpus. Corpus ranks documents with it through scikitplot.corpus.make_levenshtein_ranker; see Corpus integration.