exhaust
A Python library to exhaustively enumerate a combinatorial space represented by a function.
The API is modelled after Python's random module and should feel familiar. An additional convenience function maybe() was added to cover a common case.
If you're missing a function and the corresponding space can be enumerated feel free to open an issue. Any functions that generate real-valued distributions cannot be supported.
Example
import exhaust
def generate_character(state: exhaust.State):
eyes = []
for _ in range(state.randint(1, 3)):
eyes.append({
'color': state.choice(['brown', 'blue']),
'glowing': state.maybe()
})
size = 'giant' if len(eyes) == 1 else 'normal'
accessories = []
if len(eyes) == 2:
if state.maybe():
accessories.append('hat')
if state.maybe():
accessories.append('ring')
character = {
'size': size,
'eyes': eyes,
'accessories': accessories
}
return character
# iterates over a space of 132 characters
for character in exhaust.space(generate_character):
print(character)
As you can see, navigating the space works fine within loops as well. Each time a function from the State object is called (like maybe()), you can think of it as forking the current path into multiple branches, leading to a tree that gets explored. While exploring, the user-defined function is called for each path of the tree.
See the examples/ folder for further examples that can be run on the command line.
See the API Documentation for all available State functions.
Installation
pip install exhaust
Development
Requires pip >= 21.3
Editable install:
pip install -e .
Run tests:
pip install -r requirements-test.txt
pytest
Build docs:
pip install -r requirements-docs.txt
sphinx-build -b html docs dist-docs
Build wheel:
pip install build
python -m build --wheel
Acknowledgments
This package is inspired by KerasTuner's method of defining hyperparameter search spaces.
14 Aug 2, 2021
34 Nov 28, 2022
1 Oct 27, 2021
1 Oct 31, 2021
2 Nov 28, 2021
3 Jan 6, 2022
7 Nov 20, 2022
2 Jan 19, 2022
8 Mar 23, 2022
8 Feb 27, 2022
5 May 21, 2022
100 Jan 01, 2023
1 Nov 29, 2021
2 Sep 03, 2022
32 Nov 16, 2022
0 Feb 01, 2022
24 Aug 06, 2022
7 Dec 07, 2022
3 Jul 07, 2022
10 Jun 30, 2022
14 Oct 14, 2022
1 Nov 14, 2021
11 Nov 30, 2022
13 May 26, 2022
2 Oct 24, 2021
2 Feb 18, 2022
17 Nov 24, 2022
25 Aug 20, 2022