griblet 0.3.0

Carve computation trees from a burl of dependencies.

griblet

The griblet knows the easiest path through your data.

• It thrives where there are multiple, looping paths.
• It can account for cache and disk state.
• It loves scientific post-processing.

Installation

Install griblet in the usual way:

pip install griblet

The griblet has no runtime dependencies.

Example

Start by describing common computational paths in your data. For example, below one can compute volume via two paths: via area, and directly from length, width, and height.

from griblet import Graph

graph = Graph()

graph.add("length", lambda: 5.0)
graph.add("width", lambda: 4.0)
graph.add("height", lambda: 3.0)

graph.add(
    "area",
    lambda length, width: length * width,
    needs=("length", "width"),
    cost=2.0,
)

graph.add(
    "volume",
    lambda area, height: area * height,
    needs=("area", "height"),
    cost=2.0,
)
graph.add(
    "volume",
    lambda length, width, height: length * width * height,
    needs=("length", "width", "height"),
    cost=3.0,
)

print(graph.compute("volume"))

The griblet finds the easiest path.

The griblet is okay with disconnected graphs; when there is no path to the requested field, griblet raises NoPathError. The user can then consider adding further computational paths with graph.add.