Use exact fractional sequences per group

[victorlin]

Description of proposed changes

The previous _calculate_fractional_sequences_per_group() was an approximation of this exact value. The approximation could return a fractional value above 1, which would fail the assertion in get_probabilistic_group_sizes().

Related issue(s)

• Closes Simplify probabilistic sampling calculation #1588
• Closes Subsample has uncaught assertion in ncov #1598

Checklist

• Automated checks pass
• Check if you need to add a changelog message
• Check if you need to add tests
• Check if you need to update docs

[corneliusroemer]

Thanks! Is the linked issue correct? Shouldn't it close #1598? Rather than #1588

[victorlin]

Thanks, it should close both. I've updated the PR description.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 70.99%. Comparing base (47c83e0) to head (d73dbee).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1599      +/-   ##
==========================================
- Coverage   71.02%   70.99%   -0.04%     
==========================================
  Files          79       79              
  Lines        8256     8247       -9     
  Branches     2003     2001       -2     
==========================================
- Hits         5864     5855       -9     
  Misses       2101     2101              
  Partials      291      291              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[corneliusroemer]

There's a reason the original calculation is the way it is. I don't think that's a bug or unnecessary.

Imagine there are 90 groups with 1 sequence and 10 groups with 1000.
We want to sample 1000 sequences.

The original calculation would pick around 91 sequences per group.

Yours now picks 10 per group.

The original resulted in around 1000 sampled sequences.

Your calculation now in only 190.

[victorlin]

@corneliusroemer that scenario is not affected by the changes here. I've responded in more detail at #1588 (comment)

[victorlin]

I meant to release this as part of 25.4.0 but forgot to merge 🤦 it will come in the next release.