cuthbertLab / music21 / 1593 / 2
93%
master: 93%

DEFAULT BRANCH: master
Ran
Files 197
Run time 11s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 90.067% (+0.002%) from 90.065%
1593.2

push

travis-ci

mscuthbert 
Corrected melodic arcs feature extractors; other minor changes (#281)

* typo in comment

* Fixed/rewrote the Note Density feature. Note that the test result was incorrect; current result matches the original jSymbolic implementation. Partitioning by seconds seems like an inelegant solution for a non-discrete problem.

* Implemented feature extractor: standard deviation of note durations

* Implemented Duration feature.

* Fixed NoteDensityFeature (second commit… the test needed to be updated as well)

* Removed print statement meant for debugging/checking work

* Altered tests for NoteDensityFeature. This should match the jSymbolic implementation. Note that density in the jSymbolic library refers simply to NoteOn events and *not* to the number of simultaneosly sounding voices at each moment.

* Assign NoteDensity value to f.vector[0] and not f.vector.

* Expanded documentation and test suite

* Added new feature extractors to the relevant lists. Recategorized DurationFeature as R36 as per jSymbolic. Removed references in the comments to how many were implemented, as that comment could easily become out of date.

* Completion stats and feature indexes need to be updated after adding two new features.

* Added feature to completion stats

* "Normalize" in mathematical/statistical terms does *not* mean to divide by the largest absolute value to place everything between -1 and 1; it means that everything has a *sum* of 1. I've renamed the old "normalize" function to divide_by_max.

The functions that used divide_by_max actually gave a results different from the original jSymbolic implementation, which use a true normalization (ie dividing by the sum). Thus I have used divide_by_sum for these functions instead. The test suite has been updated.

The "Most Common Pitch" feature extractor divided its result by the bin label of the most common pitch - something the jSymbolic version does not do. This has been corrected to match.
... (continued)

67293 of 74714 relevant lines covered (90.07%)

0.9 hits per line

Source Files on job 1593.2