Efficient sequence alignment path data structure

[mataton]

This PR will implement the sequence alignment path data structure as outlined in issue #1974.

Please complete the following checklist:

• I have read the contribution guidelines.

• I have documented all public-facing changes in the changelog.

• This pull request includes code, documentation, or other content derived from external source(s). If this is the case, ensure the external source's license is compatible with scikit-bio's license. Include the license in the licenses directory and add a comment in the code giving proper attribution. Ensure any other requirements set forth by the license and/or author are satisfied.

  • It is your responsibility to disclose code, documentation, or other content derived from external source(s). If you have questions about whether something can be included in the project or how to give proper attribution, include those questions in your pull request and a reviewer will assist you.

• This pull request does not include code, documentation, or other content derived from external source(s).

Note: This document may also be helpful to see some of the things code reviewers will be verifying when reviewing your pull request.

[codecov]

Codecov Report

Attention: Patch coverage is 97.75281% with 10 lines in your changes are missing coverage. Please review.

Project coverage is 98.47%. Comparing base (541a930) to head (cdcf229).
Report is 1 commits behind head on main.

Files	Patch %	Lines
skbio/alignment/_path.py	95.47%	8 Missing and 1 partial ⚠️
skbio/alignment/_tabular_msa.py	90.90%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2011      +/-   ##
==========================================
- Coverage   98.48%   98.47%   -0.02%     
==========================================
  Files         182      184       +2     
  Lines       31057    31502     +445     
  Branches     7563     7673     +110     
==========================================
+ Hits        30586    31021     +435     
- Misses        455      464       +9     
- Partials       16       17       +1     

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

[qiyunzhu]

There is a shorter may that can replace the code between lines 332 to 354:

if state == 0:
    match_arr = ...
else:
    cigar.append(str(length) + codes[state])
flipped = state ^ 3
idx1 += flipped & 1 and length
idx2 += flipped & 2 and length

This method uses indexing and bitwise operations to avoid rewriting cigar.append and idx += for each condition. It is not necessarily faster than the current solution, nor is it more elegant (bitwise operation is hard to understand!). Therefore, it is only for reference.

[mortonjt]

My comments have been addressed

[qiyunzhu]

Here, indices is compared against gap twice. You can further optimize by doing it only once.

bits = indices == gap
starts = indices[np.arange(indices.shape[0]), np.argmin(bits, axis=1)]
return cls.from_bits(bits, starts)

[qiyunzhu]

Should gaps be starts?

[qiyunzhu]

@mataton Thanks! Let's merge.