Faster and more Robust Kern Import

[manoskary]

Faster and more Robust Kern Import

Updates

• Using numpy.readtxt to ensure fast parsing and correct typesetting.
• Two versions of import:
  1. Same number of voices from beginning to end
  2. Dynamic voices (spline splitting). This version prunes each voice starting from the leftmost side and removes them gradually from each line. Finally it replaces everything with a square matrix.
• First creating partitura timed objects in a numpy array and deciding position on the timeline afterwards (See our paper on duration algebras)
• Added some exceptions and corner cases for renaissance durations, a.o.

STATISTICS

• Up to 3x faster for large corpora loading such as string quartets
• Loads 98% of polish music corpus (~10k) scores.

NOT implemented yet

• Articulation
• Dynamics
• Tempo markings
• Page comments
• Automatic Instrument grouping

[codecov-commenter]

Codecov Report

Attention: 98 lines in your changes are missing coverage. Please review.

Comparison is base (018f264) 84.96% compared to head (ef751af) 84.12%.
Report is 1 commits behind head on develop.

Files	Patch %	Lines
partitura/io/importkern_v2.py	90.84%	37 Missing ⚠️
partitura/score.py	42.85%	36 Missing ⚠️
partitura/io/exportkern.py	84.93%	25 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #344      +/-   ##
===========================================
- Coverage    84.96%   84.12%   -0.85%     
===========================================
  Files           82       83       +1     
  Lines        14236    14569     +333     
===========================================
+ Hits         12096    12256     +160     
- Misses        2140     2313     +173     

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

[fosfrancesco]

Ideally, it would be nice to have a lot of tests for the functions to fill rests since they can be a bit tricky. But since they are not the main topic of this branch, I would open another branch with only these tests, and eventual modifications to the functions.

[manoskary]

Ideally, it would be nice to have a lot of tests for the functions to fill rests since they can be a bit tricky. But since they are not the main topic of this branch, I would open another branch with only these tests, and eventual modifications to the functions.

Thank you for the review. I agree this is particularly important. The kern import uses only measure wise rest infilling, meaning it doesn't happen for the entire piece but only for certain measures where the number of voices changes. I noticed that there are some lines that do not have coverage right now and I will address this in the tests. For the rest infilling, I agree that a separate PR with tests and checks would be probably better and clearer. Once the current PRs are closed I will update the tests.