skean is a PEP 523-based hack to
add Excel-like reactivity to
python function calls. It exposes a decorator (sheath) to memoize functions and give them
push-based invalidation semantics. These semantics only occur in contexts managed by
skean.tracing. Think of it as unidirectional Trellis.
from skean import sheath, tracing
@sheath
def get_celsius() -> float:
# pretend it's a slow API that warrants memoizing
return 50.0
@sheath
def get_fahrenheit() -> float:
return (get_celsius() - 32) / 1.8
with tracing():
assert get_fahrenheit() == 10.0
# the previous call established a link between `get_celsius` and `get_fahrenheit`
# `sheath` behave like `functools.lru_cache` whose memoized results can be retrieved
# (via `__getitem__`) and, if necessary, invalidated
celsius = get_celsius[()]
assert celsius.valid is True
fahrenheit = get_fahrenheit[()]
assert celsius.callers == {fahrenheit}
# Demo invalidation (notify transitive closure of callers)
assert fahrenheit.valid is True
celsius.invalidate()
assert fahrenheit.valid is FalseLike the Dataflow programming
page explains, skean models a program as a Directed Acyclic Graph
or DAG. As such, the name dagger came to mind. Unfortunately, dagger
already exists on PyPI so we looked for synonyms.
According to wiktionary, a skean is a "double-edged, leaf-shaped, typically bronze dagger formerly used in Ireland and Scotland." This project's code quality clearly makes its usage double-edged.