Backends (C extension vs pure Python)

pyhotpants supports two execution backends. Backward compatibility is a hard requirement: existing scripts that omit backend flags continue to use the C extension path.

Default: C extension

from hotpants import Hotpants, HotpantsConfig

hp = Hotpants(template, science, config=HotpantsConfig(...))
# equivalent to use_c_extension=True
results = hp.run_pipeline()

This is the stable path shipped on main (0.1.x). Prefer it for production workflows unless you need a pure-Python-only feature.

Opt-in: pure Python

hp = Hotpants(
    template,
    science,
    config=HotpantsConfig(...),
    use_c_extension=False,
)

The pure backend reimplements the Alard & Lupton pipeline in hotpants.pure (NumPy / Numba). For classic grid stamps with oversample=1, it is intended to recover the same scientific answer as the C path (see C vs pure Python parity policy).

Pure-Python-only features

These require use_c_extension=False:

• oversample > 1 — high-resolution template / low-resolution science

• stamp_mode="connected_regions" — irregular connected stamps from a star catalog (see Connected-region stamps)

Attempting oversampled templates or connected-region stamps with the C backend raises an error.