Supertest for Python

The distribution is schematic-supertest; import it as schematic. Install from this repository until a registry release exists:

python3 -m pip install -e .

from schematic import *

@supertest
def integer_division_is_bounded(value: int, divisor: int):
    assume(value >= 0 and divisor > 0)
    assert value // divisor <= value

Explicit imports (from schematic import assume, supertest) and qualified names (@schematic.supertest, schematic.assume(...)) also work at runtime. Current Pup documentation requires the star import for discovery of the bare decorator.

The decorator preserves the original callable and attaches immutable metadata. is_supertest and metadata_of expose it; duplicate markers and marker arguments are rejected. Marked functions are excluded from pytest collection.

A true assumption continues. assume(False, message=None) raises AssumptionNotMet, a SystemExit subclass with code == 0. Uncaught on the main thread, it exits successfully without a traceback. The optional message is available as .message. Run one input per process; this does not generate inputs or prove properties.

Python caveat: SystemExit can be intercepted, cleanup in finally blocks still runs, and raising it on a worker thread only terminates that thread. Use the main thread without catching SystemExit/BaseException, or explicitly catch AssumptionNotMet in a runner that classifies rejected inputs. Ordinary except Exception handlers do not catch it. This changes the base class from the old library, which treated rejection as an ordinary exception.

python3 -m unittest discover -s tests -v

Requires Python 3.11+. Adapted from schematic-tech/schematic-supertests at b6ccba3b5e0d42ea2846d0476eeb9496afb4a7f4.

Distribution

The public package name is singular: schematic-supertest on PyPI.

python -m pip install schematic-supertest

It supplies the same schematic import package. Releases contain a platform-neutral wheel and a source distribution, each carrying both license texts and typing metadata. See release automation and PyPI setup.

License

This interoperability library is available under either MIT or Apache-2.0, at your option.