Refactor for clarity, add comments.

[matthew-brett]

We were finding it hard to follow the algorithm, and this seemed generally undesirable, so we have attempted a refactor for greater clarity.

We've also used the more standard RNG instance pattern for the random number generation.

The tests are failing, but this seems also to be true of main.

[matthew-brett]

For the tests - I wasn't sure what to do. I think the original tests are assuming that something like the following will always give the same output value i:

# In the test function.
np.random.seed(2025)
# In `select_proposals`
np.random.seed(None)
i = np.random.choice(1000)  # e.g.

but in my hands, that is not true. Accordingly the tests on main are failing with random actual output differing from expected. On this branch the tests fail with specific and repeatable actual output. I'm happy to check the expected output by hand if that's helpful.

[lucascolley]

LGTM, thanks Matthew!

[matthew-brett]

@lucascolley - what say you for the tests?

[lucascolley]

#57 (comment) sounds good to me. What is the diff to make the tests pass? https://github.com/15r10nk/inline-snapshot could make that easier to find.

[matthew-brett]

OK - I've added some tests to assert the proportions match the weights, and updated the test output to the current test output.

I did a little more refactoring of the project random selection for clarity.

I've also added a test.yml file as .github/tests.pending, because I can't push with the workflow enabled:

 ! [remote rejected] refactor-algorithm -> refactor-algorithm (refusing to allow a Personal Access Token to create or update workflow `.github/workflows/test.yml` without `workflow` scope)
error: failed to push some refs to 'https://github.com/matthew-brett/small-development-grant-proposals'

[matthew-brett]

I've reverted the algorithm to the stepwise selection procedure; ordering and then confirming is neater to my eye, but I thought it would require extra review work to confirm it gives equivalent results. I did use inline_snapshot for the tests.