- Update to 0.9.2:
* Fixed librosa.show_versions to match package dependencies.
* Fixed a bug in librosa.effects.split when applied to multichannel data.
* Removed test data from main repository and reduced the git attic.
* Accept pre-constructed audioread objects in librosa.load.
* Accelerate librosa.magphase.
* Nearly full support for multi-channel processing.
* Option to disable unicode characters in display functions.
* Significantly expanded the library of example audio clips.
* The default padding mode for most functions (including STFT) is now
zero-padding.
* librosa.load and librosa.stream can now operate directly on open
soundfile objects.
* librosa.display.specshow now uses centered coordinate grids.
* librosa.iirt now exposes API control over resampling modes.
* Maximum frequency is now correctly inferred as Nyquist in onset strength
calculation.
* librosa.effects.deemphasis no longer modifies the input signal in-place.
* librosa.util.frame now correctly works for arbitrary memory layouts and
numbers of axes.
* Corrected a normalization error in inverse CQT.
* ibrosa.cqt now supports arbitrary hop lengths.
* Added a run-time check for minimally supported matplotlib versions.
* Enhanced continuous integration testing for oldest and newest environments.
* librosa.effects.deemphasis, inverse operation of
librosa.effects.preemphasis.
* librosa.display.waveshow, adaptively visualize waveforms by amplitude
envelope when zoomed out, or raw sample values when zoomed in.
- Add patch remove-contextlib2.patch:
* No longer require contextlib2.