Backends and fallback#

Inspect capabilities#

from scikitplot.levenshtein import available_backends

for info in available_backends():
    print(info.name, info.available, info.version, info.license)

The capability record is informational and does not import every optional accelerator eagerly when the facade itself is imported.

Automatic policy#

backend="auto" uses this preference order:

Backend	Implementation	License posture	Automatic?
`internal`	bundled `cexternals._editdistance`	MIT component in Scikit-Plots	yes
`rapidfuzz`	`rapidfuzz.distance.Levenshtein`	MIT	yes
`python`	local dynamic-programming fallback	BSD-3-Clause	yes, terminal fallback
`levenshtein`	external `Levenshtein` package	GPL-2.0-or-later	no; explicit only

Explicit selection#

from scikitplot.levenshtein import distance

distance("kitten", "sitting", backend="rapidfuzz")

If that backend is unavailable:

strict=False (default): Log a warning and fall back to the safe automatic path.
strict=True: Raise ImportError.

distance(
    "kitten",
    "sitting",
    backend="rapidfuzz",
    strict=True,
)

Runtime failure#

An accelerator can import correctly and still fail for a particular input or environment.

With strict=False the facade warns and uses pure Python. With strict=True the original runtime exception is propagated.

The pure-Python implementation is therefore the portability floor, not merely a testing backend.

Why the GPL backend is explicit-only#

Scikit-Plots can interoperate with an installed external package without making it part of the default automatic selection policy. This keeps automatic behavior license-conservative and predictable.

Use the external backend only when your environment and distribution policy have made that choice deliberately.