Repository navigation
Improve ndarray return types - #2052
Merged
bmcfee merged 13 commits intoJun 16, 2026
Merged
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## 1.0.0dev #2052 +/- ##
=========================================
Coverage 98.89% 98.89%
=========================================
Files 34 34
Lines 5236 5238 +2
=========================================
+ Hits 5178 5180 +2
Misses 58 58
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
bmcfee
reviewed
Jun 16, 2026
Contributor
There was a problem hiding this comment.
Pull request overview
This PR continues the typing work from #2050 by tightening return type annotations for functions that return numpy.ndarray, focusing on cases where the returned array’s dtype and/or shape are deterministic.
Changes:
- Add shape-typed aliases (
_Array1D,_Array2D) and apply them to selected APIs. - Refine many return annotations from generic
np.ndarrayto dtype- (and sometimes shape-) specificNDArray[...]/_Array*D[...]. - Simplify/trim overload sets where unions can express the return types directly.
Reviewed changes
Copilot reviewed 15 out of 15 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| tests/test_display.py | Updates a test assertion to accommodate tightened typing in display multiplot helpers. |
| librosa/util/utils.py | Adds shape/dtype-specific return types for utility helpers (sorting, extrema, peak picking, interpolation, masking, uniqueness). |
| librosa/util/matching.py | Tightens match index return type to a 1D uint32 array. |
| librosa/sequence.py | Refines return types for DTW/RQA/Viterbi helpers and related internal utilities. |
| librosa/segment.py | Refines return types for segmentation helpers and introduces a generic-typed decorator signature. |
| librosa/filters.py | Refines return typing for wavelet helper output arrays. |
| librosa/effects.py | Refines return typing for silence detection and trimming helpers; trims redundant overloads. |
| librosa/display.py | Refines return typing for coordinate and multiplot setup helpers (notably object-typed arrays). |
| librosa/core/spectrum.py | Tightens scalar/array return types for dB/power/amplitude conversion utilities and removes redundant overloads. |
| librosa/core/notation.py | Refines degree-array return types and overloads for internal note parsing helpers. |
| librosa/core/intervals.py | Refines return types for interval construction helpers and overload simplification. |
| librosa/core/convert.py | Refines scalar vs array return types across conversion helpers (frames/samples/time/notes/frequencies). |
| librosa/core/audio.py | Refines return types for zero-crossing detection, click generation, tones/chirps, and μ-law expansion. |
| librosa/beat.py | Refines return typing for internal beat tracking mask output. |
| librosa/_typing.py | Adds shape-typed ndarray aliases for 1D/2D arrays to support tighter annotations elsewhere. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
bmcfee
approved these changes
Jun 16, 2026
bmcfee
left a comment
bmcfee
left a comment
Member
There was a problem hiding this comment.
LGTM - your typing efforts are so very appreciated!
If there's nothing left for this one, I'm happy to merge.
Contributor
Author
Yea this is good to merge as far as I'm concerned |
bmcfee
added a commit
that referenced
this pull request
Jul 13, 2026
* Update version.py * Update ci.yml also CI on PR to 1.0 dev branch * Update docs-ci.yml also doc CI on pr to 1.0.0dev * Test overhaul (#1925) * removed dependency on test1.wav in onset tests * removed dependency on test1.wav in beat tests * properly scoping chirp fixtures * simplifying test_icqt * simplifying onset strength tests * adding some comments on new onset tests * simplifying beat tests * simplifying griffinlim tests * cleaning up gl-cqt tests * further test optimizations * fixing a test naming issue * Reduced nnls_multiblock test * Reduced recurrence matrix test * Reduced hybrid cqt tests * reducing hpss tests * removed test1_22050 from features tests * removing unnecessary example file reference * seeing if skipping ifgram regression drops coverage * Replacing regression fixtures on stft tests * putting in skips for old regression tests * self-contained lpc correctness tests * removing old fixtures from core tests * whoops still need the fixture loader for now * removed fixture files from effects tests * removed fixture files from segment tests * removed fixture files from decompose tests * suppressing known warning in glcqt test * using a fixture for find_files tests * in-place fixture generation for multichannel tests * working on purging srand for proper fixture * purging np.random continues * more rng purging * purging randomness * purging np.random continues * srand purging * random seeding * removed all unseeded randomness in tests * transitioning core tests to internal fixtures * blacking * moving core tests to a new audio clip * migrating display tests to new test track * adding build artifacts for failed image comparison tests * revising image tests once more * progress on purging old tests * simplified tonnetz test * simplified cens tests * rewrite hz/mel tests * rewrote hz_to_octs tests * rewrote melfbnorm tests * marking more tests for removal/rewriting * rewrote centroid tests * finished with met features tests * rewrote chromafb tests * rewrote melfb tests * removing unneded code and documentation * removing unneded code and documentation * rewrote semitone filterbank tests * removed all regression tests * removed the test submodule * keeping regression tests for yin/pyin * copilot suggestions * removing another unused test * added explicit buf_to_float test * fixing coverage * blacked the tests * Peak picking by dynamic programming (#1927) * Update version.py * Update ci.yml also CI on PR to 1.0 dev branch * Update docs-ci.yml also doc CI on pr to 1.0.0dev * DP peak picking * Added tests for dp peak picker * loop optimization * linting * skipping blank lines * linting * added multichannel test for dp peak picker * a couple more extensions to peak picking * upper pinning cmake until the rest of the world gets its act together * updating documentation on peak picker to describe greedy method * spelling * mypy cleanups * removing cmake pin * squashing an unhelpful type check in multichannel * expanding docstring per copilot CR recommendations * Trying a new environment cache setup in CI (#1936) * Trying a new environment cache setup in CI * purging old ci instructions * updated cache setup for docs-ci workflow * fixing a version pin in doc CI * sphinx 8.2 doesn't like something in our multiversion builds. pin it for now * dropping sphinx pin in favor of removing historical builds on 0.9.2 * regexen lol * Variable-rate and non-iterative phase vocoding (#1931) * rewrote phase vocoder * whatever velin... * removed unnecessary unwrap * fixing missing type annotation * Finishing up rough edges on phase vocoder * coverage up on phase vocoder * coverage up on phase vocoder * final copilot recommendations * dropped python 3.8 support for the 1.0 release * bumping numba version requirement * trying to fix the min env * Include scipy-stubs in the test dependencies - v1.0.0 rebase (#1946) * Include `scipy-stubs` in the type-check dependencies https://github.com/scipy/scipy-stubs * remove unused `# type: ignore`s * restrict `scipy-stubs` to `python>=3.10` * mypy no longer ignores missing `scipy` imports * add `scipy-stubs` to the CI linting coda-forge deps * get rid of some squiggly lines * fix remaining ~mypy errors~ squiggly lines * fix `recurrence_matrix` overload for `sparse=True` * workaround a mypy type narrowing bug with `TypeIs` guards * workaround for limitations in mypy's local inference algorithm * workarounds for non-gradual `scipy.io.loadmat` return types * apply review suggestions Co-authored-by: Brian McFee <[email protected]> * post-merge fix of missing Literal annotation --------- Co-authored-by: Brian McFee <[email protected]> * bumping for scipy stubs 1.16.0.1+ * arguing with stubs * arguing with stubs * arguing with stubs * vscale option to specshow (#1938) * some very broken code in here right now * yargh * rewrote scaling code in specshow * correcting docstrings * Reverting a change in power db * fixing type annotation import * blacking and docstring fixes * updated examples to use vscale * updated examples and docstrings * linting * docstyle is strict * Fixed a typo * type hacking * specshow vscale tests * keying mpl image failure artifacts on runner metadata * trying high-precision audio for image tests * bumping error tolerance on the phase plots * adding more information to vscale image test * mpl image tests are hard * expanded docstring for specshow * expanded specshow advanced example * Switched to twilight_shifted as default phase colormap * reworking phase differential plots, updated rainbowgram example * revised plot_display * phase_unwrap_diff -> dphase * added dphase_t mode to specshow * splitting phase from db vscale tests * adding fixture * improving stability of phase display tests * bumping tolerance back up due to font layout quirks * last docstring update to specshow * more docstring fiddling * Remove default quantization for note_to_hz (#1947) * fixes #1945, leave cent deviation in place for note_to_hz * type linting * minor docstring updates * fixed docstring typo * expanded docstring added default behavior test * Freq convert nan (#1948) * fixing #1944 * added tests for nan conversion of notes * Blacking and linting * linting * suppressing bad type checks in nan converters * i don't understand type annotations at all * i can't program either * implemented colorbar helpers * added tests for colorbar helpers * updating docstring examples for new colorbar helpers * updated advanced examples for new colorbar helpers * minor tweaks to colorbar helpers * force a default color gamut on phase plots * type hinting fixes * bumping matplotlib to 3.6 (2022) * reverting to old-style formatter strings * redid image comparison test results * rearranged some code to allow more user overrides * tweaking format strings for aesthetics * fixing docstrings * Colorbar helpers (#1952) * implemented colorbar helpers * added tests for colorbar helpers * updating docstring examples for new colorbar helpers * updated advanced examples for new colorbar helpers * minor tweaks to colorbar helpers * force a default color gamut on phase plots * type hinting fixes * bumping matplotlib to 3.6 (2022) * reverting to old-style formatter strings * redid image comparison test results * rearranged some code to allow more user overrides * tweaking format strings for aesthetics * fixing docstrings * whoops missed a commit * Apply suggestions from copilot code review Mostly cosmetic changes Co-authored-by: Copilot <[email protected]> * fixing a type testing error * Two-slope norm for inferred diverging colormap (#1954) * twoslopenorm diverging colormaps * updating image comparison tests for new diverging cmap style * fixing some buglets * fixing some numerical precision issues in image tests * working on ic test for diverging colormaps * working on ic test for diverging colormaps * Apply suggestions from code review Co-authored-by: Copilot <[email protected]> * relaxing rms on divergence color tests * updating advanced example for display * updating viterbi example with new diverging cmap plot * expanded cmap to support preconstructed colormap objects * fixing type import business * Update librosa/display.py Co-authored-by: Copilot <[email protected]> * bug fixing * fixing coverage * fixing coverage * deprecation rename of cmap * deprecation rename of cmap * Update librosa/display.py More direct dtype check for boolean data in cmap inference Co-authored-by: Copilot <[email protected]> --------- Co-authored-by: Copilot <[email protected]> * Warn on perceptual weighting underflow (#1955) * added warnings for perceptual weighting curves on underflow * fixes #1941 * Warping path to step grid (#1958) * Added path-to-steps fixes #1937 * revising warping example for time synchronization * making the stretched examples more obvious in display * added inverse warping path sampling * linting * linting * quivers * using colors more smartly in stretching example * using colors more smartly in stretching example * quivering * fixing up docs * revising some plots in the advanced gallery * added some unit tests for pathtosteps * Docstring formatting * fudging test data for path conversion to avoid rounding errors * nudging scipy dependency to unstick test failures * final polish on music sync example figures * Fix subsegment example (#1959) * fixed #1950 * improved styling * SPEC07 compliant rng parameters (#1960) * transitioning random_state to rng= for spec07 * suppressing type failures * Update librosa/_typing.py Co-authored-by: Copilot <[email protected]> * fixing old type annotation code * simplified testing, consistentized errors * improved docstring styling * Update tests/test_core.py Co-authored-by: Copilot <[email protected]> * preserve support for old randomstate objects in griffinlim --------- Co-authored-by: Copilot <[email protected]> * nudging min python in setup.cfg * Upload doc build preview artifacts (#1962) * adding an artifact build for pr docs * Trying a different naming * Expire deprecations for 1.0 release (#1963) * removed audioread dependency * removed res_type=None support for vqt and friends * removed get/set_fftlib * removed old filterbanks for cqt * fixing flakes * removed x_axis param from waveshow * removed winlength from yin and pyin * removed audioread from version lister * clarified comment in cqt * blacking * fixing docstring formatting * fixing type annotations now that we are narrowly on scipy fft * removing dead tests * optionalizing dct norms * added generated/ to docs clean target * allow link failure on htk site * Signal mixing helpers (#1964) * progress on mixing * extended to_mono for multiple inputs * added to_multi * fixed a typo * fixed docstring formatting * fixed docstring formatting * examples and docs updates * types are silly * Update librosa/core/audio.py Co-authored-by: Copilot <[email protected]> * Update librosa/core/audio.py Co-authored-by: Copilot <[email protected]> * Update librosa/core/audio.py Co-authored-by: Copilot <[email protected]> * Added upmixing in to_stereo * fixes from copilot cr * added norm parameter to mixers * types are silly * types are silly * CR feedback * Apply suggestions from code review require at least one input for to_mono and to_stereo Co-authored-by: Copilot <[email protected]> * expanded tests for to_mono * Added pad testing for to_mono * test coverage for to_stereo * test coverage for to_multi * adding mixer helpers into more advanced examples --------- Co-authored-by: Copilot <[email protected]> * oct3 display modes (#1967) * first hack at oct3 * added oct3 display modes * added tests for oct3 modes * trying again without constrained layout * improving docstring and expanding vqt axis compatibility support * fixing image comparison test to support matplotlib 3.10.5 * Inverted waveshow (#1971) * added a proxy patch for waveshow label containment * First cut at inverted waveform display * tests for inverted waveshow * more fancy footwork with matplotlib artists * [CR] Wavebars (#1973) * first cut at wavebars * linting * linting * added basic wavebar tests * added basic wavebar tests * removing dependency on compressed layout for wavebar test * removing dependency on compressed layout for wavebar test * nudging mpl dependency * nudging mpl dependency * nudging mpl dependency entirely * docstring improvements * chore: Refine .gitignore for improved accuracy and coverage (#1969) This commit refines the `.gitignore` file by: - Removing redundant entries. - Improving comments for better readability. - Reorganizing and categorizing ignored files for better maintainability. - Consolidating testing and checkpoint related entries. * Adding plt.show() at the end of code blocks (#1979) * Fix types in beat tracking code (#1977) The development version of Numba 0.62.0 is unhappy about `range` being applied to floating point values here. * Fix #1970: Prevent AdaptiveWaveplot adaptor from being garbage collected (#1992) - Add _WAVESHOW_ADAPTORS WeakKeyDictionary to retain adaptors per-Axes - Store artists (steps, envelope) and axes as weakrefs in AdaptiveWaveplot - Add properties to preserve existing attribute API - Add GC guard in update() for safety - Add 3 regression tests * Add Metrogram Transform (#1986) * feat: add metrogram transform * allow passing no aggregate, improve documentation * feat: add interp_broadcast Reference: #1909 * show metrogram in html docs * show interp_broadcast in html docs * fix: linting issues * use keywording,reorder x1/x1_pos, check axis range * generalize axis for metrogram transform also fix a logic bug with checking index bounds * overload interp_broadcast when func = None * fix interp_broadcast not accepting negative axis * formatting * set np.expand_dims to -3 instead of using a param * add metrogram example * add interp_broadcast example * add tests for metrogram * rename func -> op in interp_broadcast * simplify interp_broadcast example choosing to keep the example minimal like the other examples in utils an exhaustive use of all parameters is better suited for the tests * add tests for interp_broadcast * metrogram example: keep the zero in funt_freqs * remove duplicate imports added unintentionally * remove blank line after doc strings * add tests for interp_broadcast axis * update metrogram docs * Fix typo Co-authored-by: Copilot <[email protected]> * Update librosa/feature/rhythm.py Co-authored-by: Copilot <[email protected]> * Update librosa/util/utils.py Co-authored-by: Copilot <[email protected]> * Update librosa/util/utils.py Co-authored-by: Copilot <[email protected]> * fix: input shape validation based on broadcasting rules Co-authored-by: Brian McFee <[email protected]> * docs: type change kind: str -> kind: _InterpKind --------- Co-authored-by: Copilot <[email protected]> Co-authored-by: Brian McFee <[email protected]> * Update type annotations (#1980) * nudging matplotlib min version for spec0 * needs mpl in lint environment now * type hacking * type hacking * type annotations are the worst * type hacking continues * type hacking continues * types types types * bumping numpy for env compatibility * fixing scipy sparse indexing type annotations * type hacking * mypy is stupid * mypy is stupid * mypy is stupid * strengthening tests stupidly for mypy * Fixing type annotations in display and tests * upper pinning sphinx dependency until sphinx-multiversion is fixed * trying to address copilot reviews * more type annotation nonsense * Trying to fix numba/numpy compatibility for test environments * type annotations * Migrate from SciPy spmatrix to sparray (#1994) * [ENH] Migrate from scipy.sparse matrix to array interface (#1902) * spell check * spell check * Remove redundant empty-index check in sparsify_rows; add test for 1D sparse shear rejection * Fix 1D sparse shear test for scipy < 1.14 compatibility * Address review: accept sparse matrices in shear, vectorize sparsify_rows * Suppress scipy-stubs attr-defined errors in sparse matrix shear test * Python 3.14 test support (#1999) * adding 3.14 to test matrix * is it a name mismatch problem? * forcing python versions back into place * Trying to force a source build on the samplerate package * reverting source build for samplerate, it doesn't work * nudging python for docs build to 3.11 * nudging python for docs build to 3.11 * bypassing zenodo linkchecks * updated libsndfile link in docs * feat: implement hybrid_tempogram (#1991) (#2004) * feat: implement hybrid_tempogram per Peeters 2005 * refactor: address CR feedback, explicit kwargs, finite interpolation * Merge branch '1.0.0dev' into feature/hybrid-tempogram and resolve conflicts * Merge upstream 1.0.0dev and resolve conflicts * Merge upstream 1.0.0dev and fix typo in metrogram * Resolve rhythm imports and add to __all__ * style: fix linting and restore metrogram * test: add unit tests for hybrid_tempogram coverage * fix: use fourier_tempo_frequencies for hybrid_tempogram freq grid * fix: import fourier_tempo_frequencies * fix: remove non-english text from docstring to pass codespell * refactor: simplify hybrid_tempogram array handling and fix kwargs * style: remove trailing whitespace in docstring to pass velin linter * fix: resolve mypy type hints for interp1d and remove duplicate decorator * refactor: add type hints, docstring examples, and optimize interp kwargs per review * refactor: apply robust type hinting, clean docstrings, and array return per strict mypy checks * fix: finalize hybrid_tempogram with strict typing, optimized interpolation, and test coverage * fix: remove assume_sorted=True to fix interpolation on descending lag domain * refactor: apply reviewer feedback for optimizations and test fixtures * Update librosa/feature/rhythm.py Co-authored-by: Copilot <[email protected]> --------- Co-authored-by: Brian McFee <[email protected]> Co-authored-by: Copilot <[email protected]> * Document specshow frequency axis scaling (#2006) Co-authored-by: Terry Chen <[email protected]> * Surface contributing guide in docs site, link build instructions (#2009) * Surface contributing guide in docs site, link build instructions Adds myst-parser to render CONTRIBUTING.md inside the docs sidebar, so first-time contributors can find the contributor guide without leaving the docs site. Also updates CONTRIBUTING.md to reference docs/BUILD.md for documentation build instructions, replacing a stale reference to a doc/README that no longer exists. Fixes #1800. * Add myst-parser to docs env * fixed broken link in contributing guidelines --------- Co-authored-by: Brian McFee <[email protected]> * Multiplot (#2012) * first cut at multiplot * multiplot cleanups * type annotating * example fixups * added legend aggregator, fixing sphinx configs * fixing type hints * support variadic multiplot inputs * polishing multiplot * velinizing * fixing singleton multiplots * Added legend_for_axes tests * multiplot helper tests * updated multilegend plots to be more informative * coverage up * coverage up * fixing type checks in display tests * test blacking * tying up loose ends * covering legend loc inference * multiplot coverage * adding fixtures for multiplot * Added titles= support to multiplot * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * improving on copilot suggested test * improved docstrings * velinizing * added example for legend_for_axes * simplified docstring * fixing #2013 * renaming function aliases in multiplot * expanded unit tests to cover multiplot titles * expanded legend_for_axes docstring to clarify loc quirkiness * fixing type annotations for optional labels and titles * working on stefanv's comments * Accept user-supplied property cycles in multiplot * fixing positional argument in display tests * simplified docstring examples * updated docstring on multiplot orient * finished docstring fixups --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Spec0 2026.06 (#2018) * attempting spec0 compliance * attempting spec0 compliance * revised min doc build version in build.md * added dead batteries to doc build matrix * bumped numba version * trying with strict in conda env * bumping packaging to 23.2 for python 3.12 support * bumped soxr to 1.0.0 * fixing environment naming while i'm at it * Bumping msgpack-python for environment consistency * bumping lazy and typing_extensions * bumping min versions again for py3.12 environment compatibility * bumping versions to ensure python 3.12 compatibility * removed version constraint on scipy stubs * removed resource management shim dead branch * cleanups around python version handling * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * display.wavef0 (was: melodyplot) (#1976) * added mask to waveshow * fixing stuff * melody plot first cut * added missing docstring update * Added test coverage * nudging to mpl 3.8 * Temporarily nudging ci minimal to mpl 3.10 to see if that works * Temporarily nudging ci minimal to mpl 3.10 to see if that works * Temporarily nudging ci minimal to mpl 3.8.4 to see if that works * should fix wavef0 inverted mode * i can't program in python apparently * nudging back to mpl 3.8 for datalim hacking * trying mpl 3.9 * trying mpl 3.9 * trying mpl 3.9 * working on making mypy happy * more mypy business * fixed typing in wavef0 * added round trip test for transformf0 * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <[email protected]> * revised docstring * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * velinizing * protect transformf0 from nonpositive f0 values * streamlined wavef0 norm calculation * copilot cr comments * copilot comments on docstrings * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * rearranging error checks * fixing type checking * fixed display.__all__ * fixing thinko in f0 bounds checking * Added docstrings to transformf0 methods * minor docstring updates * linting * contorting around edge cases in wavef0 * fixing numerical problems in wavef0 edge cases * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * added cqt overlay example in wavef0 docstring --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Yet another Viterbi acceleration PR (#2007) * removed over alloation in viterbi core * temporary hack faster viterbi * pyin speedups * added safety barrier on transition thresholding * added tests for viterbi acceleration * one more test case coverage * Streamlining changes * polishing viterbi optimization * Updates for mypy 2.1 (#2025) * making mypy happy in 2.1 * making mypy happy in 2.1 * mypy making us jump through hoops again * mypy making us jump through hoops again * making mypy happy in 2.1 * mypy is happy for now * maybe removing numpy typing plugins was premature * making mypy happy on minimal numpy * fidgeting some type annotations to work across multiple numpy versions * fixing some import silliness in tests * undoing some bad optimizations * fixes #2014 (#2026) * Modern linting (#2027) * migrating to ruff * migrating to ruff * down into line length constraints * removed unused imports * whitespace fixing * import ordering via ruff * pyi linting * setting output format * fixing quote styles * working on bugbear * bugbear tests * simplify linting * ruffing * ruff lint rules * Dynamic vectorize (#2028) * dynamic vectorization in beat * dynamic vectorization in core * dynamic vectorization in core * dynamic vectorization in utils * removed all static vectorized signatures * fixing some type business * rewrote abs2 for more direct dtype handling * cleaning up some typing code * more mypy tweaking * Even faster import (#2029) * being better about enforcing type checking blocks * localizing heavy imports to accelerate package import * more typechecking blocks * fixed missing import * purging typing extensions * stricter typecheck blocks * lingering type block * Negative offset load (#2030) * first cut at negative offset loading * added flac example and fixture for negative load offset test * marking ogg seek tests as optional pass * added negative offset load in stream * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <[email protected]> --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Loadx (#2031) * added loadx * updated examples to use loadx where appropriate * updated examples to use loadx where appropriate * Clarified docstring for loadx * Docstring formatting * typing and docstring improvements from copilot * more helpful failure mode in loadx (#2032) * Modernize types in `librosa._typing` (#2035) * Modernize types in `librosa._typing` * remove sudden unused `# type: ignore[misc]` * forcing upgrade to codecov action * modernizing type annotations for the future (#2036) * mdoernizing type annotations for the future * loosened util.tiny type annotation just a wee bit * more type annotation modernization (#2038) * Fill in missing type annotations (#2037) * Fill in missing type annotations in `display.py` * Fill in missing type annotations in `core/intervals.py` * Fill in missing type annotations in `beat.py` * remove expired `librosa.beat.tempo` deprecation * Display Highlighter (#2034) * first cut at display.highlight * updated examples to use highlighter * more example updates * added tests for highlighter * highlighter tests * highlighter test images * spelling * velinizing * path coverage * copilot review comments * copilot review comments * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <[email protected]> * one more copilot suggestion * modernized annotations * type modernizing --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * fixed #2040 (#2041) * avoid turning scalars into 0-d arrays in `core.{convert,spectrum}` and `util.utils` (#2039) * `core.convert`: avoid turning scalars into 0-d arrays * `core.spectrum`: avoid turning scalars into 0-d arrays * `util.utils`: avoid turning scalars into 0-d arrays * fix irreproducible mypy error * Waveplot styling (#2044) * style waveplot proxy after steps, not fill * fixed #2042 * updated tests * cleanups in the new waveplot legend code * styling and ruffing * Final polishing on waveplot legend code * improved `dtype_r2c` and `dtype_c2r` typing in `util.utils` (#2043) * Doc check modernizing (#2047) * switched velin to numpydoc lint * configuring numpydoc lint * revising numpydoc configs to avoid hobgoblins * starting to get somewhere with numpydoc linting * fixing exceptions to numpydoc linting * [WIP] Fix failing GitHub Actions job for lint and code analysis (#2048) * Initial plan * Apply remaining changes --------- Co-authored-by: copilot-swe-agent[bot] <[email protected]> * working through numpydoc corrections * working through numpydoc corrections * working through numpydoc corrections * fixing numpydoc lint * fixing exceptions to numpydoc linting * numpydoc is all passing --------- Co-authored-by: Copilot <[email protected]> * Contributing and agents update (#2017) * updating contributing guidelines with ai policy * Added agents.md * Fix CONTRIBUTING.md markdown formatting * revising contributing guidelines * integrating comments * formatting * updated contributing guidelines for numpydoc linting * reformat `**/__init__.pyi` re-exports (#2049) * Include str dtype in `ndarray` return type annotations (#2050) * Multichannel db (#2051) * implemented axis targeting for db conversion * db calculation now works properly with target axes * fixed scalar processing for db * added 'auto' mode to db conversion axes inference * fixed regression in amplitude_to_db * fixed regression in amplitude_to_db * copilot reviews * fixing some type annotations * copilot comments * cleaning up tests * pytest nonsense * Fix decorator callable types in `librosa.util.decorators` (#2053) * Improve `ndarray` return types (#2052) * Generic shaped `ndarray` type aliases * Improve `ndarray` return types in `librosa.core` * Improve `ndarray` return types in `librosa.util` * Improve `ndarray` return types in `librosa.beat` * Improve `ndarray` return types in `librosa.display` * Improve `ndarray` return types in `librosa.effect` * Improve `ndarray` return types in `librosa.filters` * Improve `ndarray` return types in `librosa.segment` * Improve `ndarray` return types in `librosa.sequence` * Fix mypy errors with `numpy<2.4` * Ignore false positive mypy error in tests related to `np.object_` array iteration * Address review comments for the typing changes in `util.utils` * Work around a mypy bug where it fails to expand `bool` into `Literal[True] | Literal[False]`. * Support matplotlib 3.11 in tests (#2054) * redoing tests for matplotlib 3.11, tightening tolerance where possible * updated failure artifact upload recipe * loosening tests more on mpl 3.10 * bypassing image comparison tests on 3.10 completely * fixed types for legend in matplotlib 3.11 * Stream resampling (#2055) * initial cut of resampling stream passes previous tests * progress on stream resampling * clarified docstring, reverted changes * updated tests * feature complete stream resampling * refactored tests * better coverage for edge cases * working in some copilot review comments * more copilot suggestions * copilot and gemini comments * cleaning up stream tests * tightening memory allocations in stream * added a future default change decorator and bundled onto stream * forgot to kw-only a parameter name * ratcheting down default stream parameter setting deprecation to 1.1 * last round of copilot reviews * massive simplification of legend_for_axes, fixes #2057 (#2059) * Refactor frequency formatters (#2060) * fixed minor ticking for oct3 modes * numpydoc linting * copilot review comments * copilot review comments * Updating actions (#2065) * updating actions * fixing an out= allocation issue * more warnings cleanup * libsamplerate comes via conda now * upload-artifact update * disable auto-activate * Standardize n_fft parameter docstrings (#2068) * Standardize n_fft parameter docstrings Rebuilt against 1.0.0dev after retargeting. Standardizes the analysis n_fft descriptions on 'length of the FFT window', fixes the bins/window conflations (piptrack, estimate_tuning, filters.mel, filters.chroma, feature.inverse), clarifies istft's frame-length inference, corrects mel_to_stft's documented return shape, and fixes frequency-axis references that assumed 2D input (istft win_length, phase_vocoder hop_length). Incorporates review feedback: 'frame' terminology in get_duration; audio-output functions reference the intermediate STFT without a bins clause. * Use 'frame' terminology for n_fft descriptions Per review: n_fft refers to frame length, not window length, since a window can be shorter than the frame. Swept all standardized n_fft descriptions accordingly. --------- Co-authored-by: Cameron Brooks <[email protected]> * precise dtype annotations in `librosa.beat` (#2069) Co-authored-by: Brian McFee <[email protected]> * Infer n_bins automatically in cqt/vqt (#2066) * added n_bins=None to cqt/vqt * added n_bins=None to cqt/vqt tests * linting * revisions * revisions * one more test * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Added nbins inference to hybrid and pseudo cqt * linting * copypasta * fixed #2071, tightened icqt tests * revised downsampling logic in vqt * copilot review comments --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Add out= parameter to to_mono for in-place downmixing (#2058) * Add out= buffer parameter to to_mono, to_stereo, to_multi Lets callers pass a pre-allocated output buffer to the mixing functions, avoiding a fresh allocation per call (e.g. per-block downmix in stream()). A shared _init_output helper validates the buffer's shape and dtype against the computed output, zero-fills it, and returns it; out=None preserves the previous allocate-fresh behavior. Implements #2056 against the redesigned 1.0 mixing API. * Remove joblib caching from to_mono/to_stereo/to_multi Per review (#2058): the @cache decorator is incompatible with out=, since a cache hit returns the memoized array and skips writing into the caller's buffer, and hashing an uninitialized out= buffer is costly/non-deterministic. Dropping caching keeps the out= contract simple and correct. * Reuse output buffers in to_stereo/to_multi and stream mono downmix Addresses review feedback on #2058: - to_stereo: downmix directly into the output rows via out=, removing a second allocation. Cast left/right to the shared dtype first so the pass-through buffer always matches to_mono's computed dtype. - to_multi: pass the output row into to_mono when no length adjustment is needed; fall back to allocate-and-fix_length only when lengths differ. - stream: reuse a single mono downmix buffer across blocks, writing the final short block into a truncated view. - Add a mixed-dtype regression test for the downmix pass-through. * Relax out= dtype check to accept higher-precision buffers Per review (#2058): rather than requiring an exact dtype match, accept any out= buffer whose dtype can hold the result without loss of precision (promote_types(out.dtype, dtype) == out.dtype). A caller who explicitly allocates a higher-precision buffer keeps that precision; too-narrow buffers are still rejected. Updated docstrings and added a wider-dtype test. --------- Co-authored-by: Brian McFee <[email protected]> * Fix RQA single-axis inputs (#2063) * Fix RQA single-axis inputs * Update tests/test_sequence.py Work around mypy for test type checks --------- Co-authored-by: Brian McFee <[email protected]> * BoundedNorm for specshow on boolean data (#2072) * added boundednorm for boolean data in specshow * Added tests for boundarynorm boolean specshow --------- Co-authored-by: Joren Hammudoglu <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Ale Lloveras <[email protected]> Co-authored-by: Haydn Lam <[email protected]> Co-authored-by: Emily <[email protected]> Co-authored-by: dhaasb <[email protected]> Co-authored-by: Daniel Fernandes <[email protected]> Co-authored-by: Gopesh Pandey <[email protected]> Co-authored-by: T <[email protected]> Co-authored-by: Terry Chen <[email protected]> Co-authored-by: Petra Kühnle <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Cameron Brooks <[email protected]> Co-authored-by: Cameron Brooks <[email protected]> Co-authored-by: ejwong24 <[email protected]> Co-authored-by: Suhas Holla Karkada Chandrashekar <[email protected]>
bmcfee
added a commit
that referenced
this pull request
Jul 13, 2026
* Update version.py * Update ci.yml also CI on PR to 1.0 dev branch * Update docs-ci.yml also doc CI on pr to 1.0.0dev * Test overhaul (#1925) * removed dependency on test1.wav in onset tests * removed dependency on test1.wav in beat tests * properly scoping chirp fixtures * simplifying test_icqt * simplifying onset strength tests * adding some comments on new onset tests * simplifying beat tests * simplifying griffinlim tests * cleaning up gl-cqt tests * further test optimizations * fixing a test naming issue * Reduced nnls_multiblock test * Reduced recurrence matrix test * Reduced hybrid cqt tests * reducing hpss tests * removed test1_22050 from features tests * removing unnecessary example file reference * seeing if skipping ifgram regression drops coverage * Replacing regression fixtures on stft tests * putting in skips for old regression tests * self-contained lpc correctness tests * removing old fixtures from core tests * whoops still need the fixture loader for now * removed fixture files from effects tests * removed fixture files from segment tests * removed fixture files from decompose tests * suppressing known warning in glcqt test * using a fixture for find_files tests * in-place fixture generation for multichannel tests * working on purging srand for proper fixture * purging np.random continues * more rng purging * purging randomness * purging np.random continues * srand purging * random seeding * removed all unseeded randomness in tests * transitioning core tests to internal fixtures * blacking * moving core tests to a new audio clip * migrating display tests to new test track * adding build artifacts for failed image comparison tests * revising image tests once more * progress on purging old tests * simplified tonnetz test * simplified cens tests * rewrite hz/mel tests * rewrote hz_to_octs tests * rewrote melfbnorm tests * marking more tests for removal/rewriting * rewrote centroid tests * finished with met features tests * rewrote chromafb tests * rewrote melfb tests * removing unneded code and documentation * removing unneded code and documentation * rewrote semitone filterbank tests * removed all regression tests * removed the test submodule * keeping regression tests for yin/pyin * copilot suggestions * removing another unused test * added explicit buf_to_float test * fixing coverage * blacked the tests * Peak picking by dynamic programming (#1927) * Update version.py * Update ci.yml also CI on PR to 1.0 dev branch * Update docs-ci.yml also doc CI on pr to 1.0.0dev * DP peak picking * Added tests for dp peak picker * loop optimization * linting * skipping blank lines * linting * added multichannel test for dp peak picker * a couple more extensions to peak picking * upper pinning cmake until the rest of the world gets its act together * updating documentation on peak picker to describe greedy method * spelling * mypy cleanups * removing cmake pin * squashing an unhelpful type check in multichannel * expanding docstring per copilot CR recommendations * Trying a new environment cache setup in CI (#1936) * Trying a new environment cache setup in CI * purging old ci instructions * updated cache setup for docs-ci workflow * fixing a version pin in doc CI * sphinx 8.2 doesn't like something in our multiversion builds. pin it for now * dropping sphinx pin in favor of removing historical builds on 0.9.2 * regexen lol * Variable-rate and non-iterative phase vocoding (#1931) * rewrote phase vocoder * whatever velin... * removed unnecessary unwrap * fixing missing type annotation * Finishing up rough edges on phase vocoder * coverage up on phase vocoder * coverage up on phase vocoder * final copilot recommendations * dropped python 3.8 support for the 1.0 release * bumping numba version requirement * trying to fix the min env * Include scipy-stubs in the test dependencies - v1.0.0 rebase (#1946) * Include `scipy-stubs` in the type-check dependencies https://github.com/scipy/scipy-stubs * remove unused `# type: ignore`s * restrict `scipy-stubs` to `python>=3.10` * mypy no longer ignores missing `scipy` imports * add `scipy-stubs` to the CI linting coda-forge deps * get rid of some squiggly lines * fix remaining ~mypy errors~ squiggly lines * fix `recurrence_matrix` overload for `sparse=True` * workaround a mypy type narrowing bug with `TypeIs` guards * workaround for limitations in mypy's local inference algorithm * workarounds for non-gradual `scipy.io.loadmat` return types * apply review suggestions Co-authored-by: Brian McFee <[email protected]> * post-merge fix of missing Literal annotation --------- Co-authored-by: Brian McFee <[email protected]> * bumping for scipy stubs 1.16.0.1+ * arguing with stubs * arguing with stubs * arguing with stubs * vscale option to specshow (#1938) * some very broken code in here right now * yargh * rewrote scaling code in specshow * correcting docstrings * Reverting a change in power db * fixing type annotation import * blacking and docstring fixes * updated examples to use vscale * updated examples and docstrings * linting * docstyle is strict * Fixed a typo * type hacking * specshow vscale tests * keying mpl image failure artifacts on runner metadata * trying high-precision audio for image tests * bumping error tolerance on the phase plots * adding more information to vscale image test * mpl image tests are hard * expanded docstring for specshow * expanded specshow advanced example * Switched to twilight_shifted as default phase colormap * reworking phase differential plots, updated rainbowgram example * revised plot_display * phase_unwrap_diff -> dphase * added dphase_t mode to specshow * splitting phase from db vscale tests * adding fixture * improving stability of phase display tests * bumping tolerance back up due to font layout quirks * last docstring update to specshow * more docstring fiddling * Remove default quantization for note_to_hz (#1947) * fixes #1945, leave cent deviation in place for note_to_hz * type linting * minor docstring updates * fixed docstring typo * expanded docstring added default behavior test * Freq convert nan (#1948) * fixing #1944 * added tests for nan conversion of notes * Blacking and linting * linting * suppressing bad type checks in nan converters * i don't understand type annotations at all * i can't program either * implemented colorbar helpers * added tests for colorbar helpers * updating docstring examples for new colorbar helpers * updated advanced examples for new colorbar helpers * minor tweaks to colorbar helpers * force a default color gamut on phase plots * type hinting fixes * bumping matplotlib to 3.6 (2022) * reverting to old-style formatter strings * redid image comparison test results * rearranged some code to allow more user overrides * tweaking format strings for aesthetics * fixing docstrings * Colorbar helpers (#1952) * implemented colorbar helpers * added tests for colorbar helpers * updating docstring examples for new colorbar helpers * updated advanced examples for new colorbar helpers * minor tweaks to colorbar helpers * force a default color gamut on phase plots * type hinting fixes * bumping matplotlib to 3.6 (2022) * reverting to old-style formatter strings * redid image comparison test results * rearranged some code to allow more user overrides * tweaking format strings for aesthetics * fixing docstrings * whoops missed a commit * Apply suggestions from copilot code review Mostly cosmetic changes Co-authored-by: Copilot <[email protected]> * fixing a type testing error * Two-slope norm for inferred diverging colormap (#1954) * twoslopenorm diverging colormaps * updating image comparison tests for new diverging cmap style * fixing some buglets * fixing some numerical precision issues in image tests * working on ic test for diverging colormaps * working on ic test for diverging colormaps * Apply suggestions from code review Co-authored-by: Copilot <[email protected]> * relaxing rms on divergence color tests * updating advanced example for display * updating viterbi example with new diverging cmap plot * expanded cmap to support preconstructed colormap objects * fixing type import business * Update librosa/display.py Co-authored-by: Copilot <[email protected]> * bug fixing * fixing coverage * fixing coverage * deprecation rename of cmap * deprecation rename of cmap * Update librosa/display.py More direct dtype check for boolean data in cmap inference Co-authored-by: Copilot <[email protected]> --------- Co-authored-by: Copilot <[email protected]> * Warn on perceptual weighting underflow (#1955) * added warnings for perceptual weighting curves on underflow * fixes #1941 * Warping path to step grid (#1958) * Added path-to-steps fixes #1937 * revising warping example for time synchronization * making the stretched examples more obvious in display * added inverse warping path sampling * linting * linting * quivers * using colors more smartly in stretching example * using colors more smartly in stretching example * quivering * fixing up docs * revising some plots in the advanced gallery * added some unit tests for pathtosteps * Docstring formatting * fudging test data for path conversion to avoid rounding errors * nudging scipy dependency to unstick test failures * final polish on music sync example figures * Fix subsegment example (#1959) * fixed #1950 * improved styling * SPEC07 compliant rng parameters (#1960) * transitioning random_state to rng= for spec07 * suppressing type failures * Update librosa/_typing.py Co-authored-by: Copilot <[email protected]> * fixing old type annotation code * simplified testing, consistentized errors * improved docstring styling * Update tests/test_core.py Co-authored-by: Copilot <[email protected]> * preserve support for old randomstate objects in griffinlim --------- Co-authored-by: Copilot <[email protected]> * nudging min python in setup.cfg * Upload doc build preview artifacts (#1962) * adding an artifact build for pr docs * Trying a different naming * Expire deprecations for 1.0 release (#1963) * removed audioread dependency * removed res_type=None support for vqt and friends * removed get/set_fftlib * removed old filterbanks for cqt * fixing flakes * removed x_axis param from waveshow * removed winlength from yin and pyin * removed audioread from version lister * clarified comment in cqt * blacking * fixing docstring formatting * fixing type annotations now that we are narrowly on scipy fft * removing dead tests * optionalizing dct norms * added generated/ to docs clean target * allow link failure on htk site * Signal mixing helpers (#1964) * progress on mixing * extended to_mono for multiple inputs * added to_multi * fixed a typo * fixed docstring formatting * fixed docstring formatting * examples and docs updates * types are silly * Update librosa/core/audio.py Co-authored-by: Copilot <[email protected]> * Update librosa/core/audio.py Co-authored-by: Copilot <[email protected]> * Update librosa/core/audio.py Co-authored-by: Copilot <[email protected]> * Added upmixing in to_stereo * fixes from copilot cr * added norm parameter to mixers * types are silly * types are silly * CR feedback * Apply suggestions from code review require at least one input for to_mono and to_stereo Co-authored-by: Copilot <[email protected]> * expanded tests for to_mono * Added pad testing for to_mono * test coverage for to_stereo * test coverage for to_multi * adding mixer helpers into more advanced examples --------- Co-authored-by: Copilot <[email protected]> * oct3 display modes (#1967) * first hack at oct3 * added oct3 display modes * added tests for oct3 modes * trying again without constrained layout * improving docstring and expanding vqt axis compatibility support * fixing image comparison test to support matplotlib 3.10.5 * Inverted waveshow (#1971) * added a proxy patch for waveshow label containment * First cut at inverted waveform display * tests for inverted waveshow * more fancy footwork with matplotlib artists * [CR] Wavebars (#1973) * first cut at wavebars * linting * linting * added basic wavebar tests * added basic wavebar tests * removing dependency on compressed layout for wavebar test * removing dependency on compressed layout for wavebar test * nudging mpl dependency * nudging mpl dependency * nudging mpl dependency entirely * docstring improvements * chore: Refine .gitignore for improved accuracy and coverage (#1969) This commit refines the `.gitignore` file by: - Removing redundant entries. - Improving comments for better readability. - Reorganizing and categorizing ignored files for better maintainability. - Consolidating testing and checkpoint related entries. * Adding plt.show() at the end of code blocks (#1979) * Fix types in beat tracking code (#1977) The development version of Numba 0.62.0 is unhappy about `range` being applied to floating point values here. * Fix #1970: Prevent AdaptiveWaveplot adaptor from being garbage collected (#1992) - Add _WAVESHOW_ADAPTORS WeakKeyDictionary to retain adaptors per-Axes - Store artists (steps, envelope) and axes as weakrefs in AdaptiveWaveplot - Add properties to preserve existing attribute API - Add GC guard in update() for safety - Add 3 regression tests * Add Metrogram Transform (#1986) * feat: add metrogram transform * allow passing no aggregate, improve documentation * feat: add interp_broadcast Reference: #1909 * show metrogram in html docs * show interp_broadcast in html docs * fix: linting issues * use keywording,reorder x1/x1_pos, check axis range * generalize axis for metrogram transform also fix a logic bug with checking index bounds * overload interp_broadcast when func = None * fix interp_broadcast not accepting negative axis * formatting * set np.expand_dims to -3 instead of using a param * add metrogram example * add interp_broadcast example * add tests for metrogram * rename func -> op in interp_broadcast * simplify interp_broadcast example choosing to keep the example minimal like the other examples in utils an exhaustive use of all parameters is better suited for the tests * add tests for interp_broadcast * metrogram example: keep the zero in funt_freqs * remove duplicate imports added unintentionally * remove blank line after doc strings * add tests for interp_broadcast axis * update metrogram docs * Fix typo Co-authored-by: Copilot <[email protected]> * Update librosa/feature/rhythm.py Co-authored-by: Copilot <[email protected]> * Update librosa/util/utils.py Co-authored-by: Copilot <[email protected]> * Update librosa/util/utils.py Co-authored-by: Copilot <[email protected]> * fix: input shape validation based on broadcasting rules Co-authored-by: Brian McFee <[email protected]> * docs: type change kind: str -> kind: _InterpKind --------- Co-authored-by: Copilot <[email protected]> Co-authored-by: Brian McFee <[email protected]> * Update type annotations (#1980) * nudging matplotlib min version for spec0 * needs mpl in lint environment now * type hacking * type hacking * type annotations are the worst * type hacking continues * type hacking continues * types types types * bumping numpy for env compatibility * fixing scipy sparse indexing type annotations * type hacking * mypy is stupid * mypy is stupid * mypy is stupid * strengthening tests stupidly for mypy * Fixing type annotations in display and tests * upper pinning sphinx dependency until sphinx-multiversion is fixed * trying to address copilot reviews * more type annotation nonsense * Trying to fix numba/numpy compatibility for test environments * type annotations * Migrate from SciPy spmatrix to sparray (#1994) * [ENH] Migrate from scipy.sparse matrix to array interface (#1902) * spell check * spell check * Remove redundant empty-index check in sparsify_rows; add test for 1D sparse shear rejection * Fix 1D sparse shear test for scipy < 1.14 compatibility * Address review: accept sparse matrices in shear, vectorize sparsify_rows * Suppress scipy-stubs attr-defined errors in sparse matrix shear test * Python 3.14 test support (#1999) * adding 3.14 to test matrix * is it a name mismatch problem? * forcing python versions back into place * Trying to force a source build on the samplerate package * reverting source build for samplerate, it doesn't work * nudging python for docs build to 3.11 * nudging python for docs build to 3.11 * bypassing zenodo linkchecks * updated libsndfile link in docs * feat: implement hybrid_tempogram (#1991) (#2004) * feat: implement hybrid_tempogram per Peeters 2005 * refactor: address CR feedback, explicit kwargs, finite interpolation * Merge branch '1.0.0dev' into feature/hybrid-tempogram and resolve conflicts * Merge upstream 1.0.0dev and resolve conflicts * Merge upstream 1.0.0dev and fix typo in metrogram * Resolve rhythm imports and add to __all__ * style: fix linting and restore metrogram * test: add unit tests for hybrid_tempogram coverage * fix: use fourier_tempo_frequencies for hybrid_tempogram freq grid * fix: import fourier_tempo_frequencies * fix: remove non-english text from docstring to pass codespell * refactor: simplify hybrid_tempogram array handling and fix kwargs * style: remove trailing whitespace in docstring to pass velin linter * fix: resolve mypy type hints for interp1d and remove duplicate decorator * refactor: add type hints, docstring examples, and optimize interp kwargs per review * refactor: apply robust type hinting, clean docstrings, and array return per strict mypy checks * fix: finalize hybrid_tempogram with strict typing, optimized interpolation, and test coverage * fix: remove assume_sorted=True to fix interpolation on descending lag domain * refactor: apply reviewer feedback for optimizations and test fixtures * Update librosa/feature/rhythm.py Co-authored-by: Copilot <[email protected]> --------- Co-authored-by: Brian McFee <[email protected]> Co-authored-by: Copilot <[email protected]> * Document specshow frequency axis scaling (#2006) Co-authored-by: Terry Chen <[email protected]> * Surface contributing guide in docs site, link build instructions (#2009) * Surface contributing guide in docs site, link build instructions Adds myst-parser to render CONTRIBUTING.md inside the docs sidebar, so first-time contributors can find the contributor guide without leaving the docs site. Also updates CONTRIBUTING.md to reference docs/BUILD.md for documentation build instructions, replacing a stale reference to a doc/README that no longer exists. Fixes #1800. * Add myst-parser to docs env * fixed broken link in contributing guidelines --------- Co-authored-by: Brian McFee <[email protected]> * Multiplot (#2012) * first cut at multiplot * multiplot cleanups * type annotating * example fixups * added legend aggregator, fixing sphinx configs * fixing type hints * support variadic multiplot inputs * polishing multiplot * velinizing * fixing singleton multiplots * Added legend_for_axes tests * multiplot helper tests * updated multilegend plots to be more informative * coverage up * coverage up * fixing type checks in display tests * test blacking * tying up loose ends * covering legend loc inference * multiplot coverage * adding fixtures for multiplot * Added titles= support to multiplot * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * improving on copilot suggested test * improved docstrings * velinizing * added example for legend_for_axes * simplified docstring * fixing #2013 * renaming function aliases in multiplot * expanded unit tests to cover multiplot titles * expanded legend_for_axes docstring to clarify loc quirkiness * fixing type annotations for optional labels and titles * working on stefanv's comments * Accept user-supplied property cycles in multiplot * fixing positional argument in display tests * simplified docstring examples * updated docstring on multiplot orient * finished docstring fixups --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Spec0 2026.06 (#2018) * attempting spec0 compliance * attempting spec0 compliance * revised min doc build version in build.md * added dead batteries to doc build matrix * bumped numba version * trying with strict in conda env * bumping packaging to 23.2 for python 3.12 support * bumped soxr to 1.0.0 * fixing environment naming while i'm at it * Bumping msgpack-python for environment consistency * bumping lazy and typing_extensions * bumping min versions again for py3.12 environment compatibility * bumping versions to ensure python 3.12 compatibility * removed version constraint on scipy stubs * removed resource management shim dead branch * cleanups around python version handling * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * display.wavef0 (was: melodyplot) (#1976) * added mask to waveshow * fixing stuff * melody plot first cut * added missing docstring update * Added test coverage * nudging to mpl 3.8 * Temporarily nudging ci minimal to mpl 3.10 to see if that works * Temporarily nudging ci minimal to mpl 3.10 to see if that works * Temporarily nudging ci minimal to mpl 3.8.4 to see if that works * should fix wavef0 inverted mode * i can't program in python apparently * nudging back to mpl 3.8 for datalim hacking * trying mpl 3.9 * trying mpl 3.9 * trying mpl 3.9 * working on making mypy happy * more mypy business * fixed typing in wavef0 * added round trip test for transformf0 * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <[email protected]> * revised docstring * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * velinizing * protect transformf0 from nonpositive f0 values * streamlined wavef0 norm calculation * copilot cr comments * copilot comments on docstrings * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * rearranging error checks * fixing type checking * fixed display.__all__ * fixing thinko in f0 bounds checking * Added docstrings to transformf0 methods * minor docstring updates * linting * contorting around edge cases in wavef0 * fixing numerical problems in wavef0 edge cases * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * added cqt overlay example in wavef0 docstring --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Yet another Viterbi acceleration PR (#2007) * removed over alloation in viterbi core * temporary hack faster viterbi * pyin speedups * added safety barrier on transition thresholding * added tests for viterbi acceleration * one more test case coverage * Streamlining changes * polishing viterbi optimization * Updates for mypy 2.1 (#2025) * making mypy happy in 2.1 * making mypy happy in 2.1 * mypy making us jump through hoops again * mypy making us jump through hoops again * making mypy happy in 2.1 * mypy is happy for now * maybe removing numpy typing plugins was premature * making mypy happy on minimal numpy * fidgeting some type annotations to work across multiple numpy versions * fixing some import silliness in tests * undoing some bad optimizations * fixes #2014 (#2026) * Modern linting (#2027) * migrating to ruff * migrating to ruff * down into line length constraints * removed unused imports * whitespace fixing * import ordering via ruff * pyi linting * setting output format * fixing quote styles * working on bugbear * bugbear tests * simplify linting * ruffing * ruff lint rules * Dynamic vectorize (#2028) * dynamic vectorization in beat * dynamic vectorization in core * dynamic vectorization in core * dynamic vectorization in utils * removed all static vectorized signatures * fixing some type business * rewrote abs2 for more direct dtype handling * cleaning up some typing code * more mypy tweaking * Even faster import (#2029) * being better about enforcing type checking blocks * localizing heavy imports to accelerate package import * more typechecking blocks * fixed missing import * purging typing extensions * stricter typecheck blocks * lingering type block * Negative offset load (#2030) * first cut at negative offset loading * added flac example and fixture for negative load offset test * marking ogg seek tests as optional pass * added negative offset load in stream * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <[email protected]> --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Loadx (#2031) * added loadx * updated examples to use loadx where appropriate * updated examples to use loadx where appropriate * Clarified docstring for loadx * Docstring formatting * typing and docstring improvements from copilot * more helpful failure mode in loadx (#2032) * Modernize types in `librosa._typing` (#2035) * Modernize types in `librosa._typing` * remove sudden unused `# type: ignore[misc]` * forcing upgrade to codecov action * modernizing type annotations for the future (#2036) * mdoernizing type annotations for the future * loosened util.tiny type annotation just a wee bit * more type annotation modernization (#2038) * Fill in missing type annotations (#2037) * Fill in missing type annotations in `display.py` * Fill in missing type annotations in `core/intervals.py` * Fill in missing type annotations in `beat.py` * remove expired `librosa.beat.tempo` deprecation * Display Highlighter (#2034) * first cut at display.highlight * updated examples to use highlighter * more example updates * added tests for highlighter * highlighter tests * highlighter test images * spelling * velinizing * path coverage * copilot review comments * copilot review comments * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <[email protected]> * one more copilot suggestion * modernized annotations * type modernizing --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * fixed #2040 (#2041) * avoid turning scalars into 0-d arrays in `core.{convert,spectrum}` and `util.utils` (#2039) * `core.convert`: avoid turning scalars into 0-d arrays * `core.spectrum`: avoid turning scalars into 0-d arrays * `util.utils`: avoid turning scalars into 0-d arrays * fix irreproducible mypy error * Waveplot styling (#2044) * style waveplot proxy after steps, not fill * fixed #2042 * updated tests * cleanups in the new waveplot legend code * styling and ruffing * Final polishing on waveplot legend code * improved `dtype_r2c` and `dtype_c2r` typing in `util.utils` (#2043) * Doc check modernizing (#2047) * switched velin to numpydoc lint * configuring numpydoc lint * revising numpydoc configs to avoid hobgoblins * starting to get somewhere with numpydoc linting * fixing exceptions to numpydoc linting * [WIP] Fix failing GitHub Actions job for lint and code analysis (#2048) * Initial plan * Apply remaining changes --------- Co-authored-by: copilot-swe-agent[bot] <[email protected]> * working through numpydoc corrections * working through numpydoc corrections * working through numpydoc corrections * fixing numpydoc lint * fixing exceptions to numpydoc linting * numpydoc is all passing --------- Co-authored-by: Copilot <[email protected]> * Contributing and agents update (#2017) * updating contributing guidelines with ai policy * Added agents.md * Fix CONTRIBUTING.md markdown formatting * revising contributing guidelines * integrating comments * formatting * updated contributing guidelines for numpydoc linting * reformat `**/__init__.pyi` re-exports (#2049) * Include str dtype in `ndarray` return type annotations (#2050) * Multichannel db (#2051) * implemented axis targeting for db conversion * db calculation now works properly with target axes * fixed scalar processing for db * added 'auto' mode to db conversion axes inference * fixed regression in amplitude_to_db * fixed regression in amplitude_to_db * copilot reviews * fixing some type annotations * copilot comments * cleaning up tests * pytest nonsense * Fix decorator callable types in `librosa.util.decorators` (#2053) * Improve `ndarray` return types (#2052) * Generic shaped `ndarray` type aliases * Improve `ndarray` return types in `librosa.core` * Improve `ndarray` return types in `librosa.util` * Improve `ndarray` return types in `librosa.beat` * Improve `ndarray` return types in `librosa.display` * Improve `ndarray` return types in `librosa.effect` * Improve `ndarray` return types in `librosa.filters` * Improve `ndarray` return types in `librosa.segment` * Improve `ndarray` return types in `librosa.sequence` * Fix mypy errors with `numpy<2.4` * Ignore false positive mypy error in tests related to `np.object_` array iteration * Address review comments for the typing changes in `util.utils` * Work around a mypy bug where it fails to expand `bool` into `Literal[True] | Literal[False]`. * Support matplotlib 3.11 in tests (#2054) * redoing tests for matplotlib 3.11, tightening tolerance where possible * updated failure artifact upload recipe * loosening tests more on mpl 3.10 * bypassing image comparison tests on 3.10 completely * fixed types for legend in matplotlib 3.11 * Stream resampling (#2055) * initial cut of resampling stream passes previous tests * progress on stream resampling * clarified docstring, reverted changes * updated tests * feature complete stream resampling * refactored tests * better coverage for edge cases * working in some copilot review comments * more copilot suggestions * copilot and gemini comments * cleaning up stream tests * tightening memory allocations in stream * added a future default change decorator and bundled onto stream * forgot to kw-only a parameter name * ratcheting down default stream parameter setting deprecation to 1.1 * last round of copilot reviews * massive simplification of legend_for_axes, fixes #2057 (#2059) * Refactor frequency formatters (#2060) * fixed minor ticking for oct3 modes * numpydoc linting * copilot review comments * copilot review comments * Updating actions (#2065) * updating actions * fixing an out= allocation issue * more warnings cleanup * libsamplerate comes via conda now * upload-artifact update * disable auto-activate * Standardize n_fft parameter docstrings (#2068) * Standardize n_fft parameter docstrings Rebuilt against 1.0.0dev after retargeting. Standardizes the analysis n_fft descriptions on 'length of the FFT window', fixes the bins/window conflations (piptrack, estimate_tuning, filters.mel, filters.chroma, feature.inverse), clarifies istft's frame-length inference, corrects mel_to_stft's documented return shape, and fixes frequency-axis references that assumed 2D input (istft win_length, phase_vocoder hop_length). Incorporates review feedback: 'frame' terminology in get_duration; audio-output functions reference the intermediate STFT without a bins clause. * Use 'frame' terminology for n_fft descriptions Per review: n_fft refers to frame length, not window length, since a window can be shorter than the frame. Swept all standardized n_fft descriptions accordingly. --------- Co-authored-by: Cameron Brooks <[email protected]> * precise dtype annotations in `librosa.beat` (#2069) Co-authored-by: Brian McFee <[email protected]> * Infer n_bins automatically in cqt/vqt (#2066) * added n_bins=None to cqt/vqt * added n_bins=None to cqt/vqt tests * linting * revisions * revisions * one more test * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Added nbins inference to hybrid and pseudo cqt * linting * copypasta * fixed #2071, tightened icqt tests * revised downsampling logic in vqt * copilot review comments --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Add out= parameter to to_mono for in-place downmixing (#2058) * Add out= buffer parameter to to_mono, to_stereo, to_multi Lets callers pass a pre-allocated output buffer to the mixing functions, avoiding a fresh allocation per call (e.g. per-block downmix in stream()). A shared _init_output helper validates the buffer's shape and dtype against the computed output, zero-fills it, and returns it; out=None preserves the previous allocate-fresh behavior. Implements #2056 against the redesigned 1.0 mixing API. * Remove joblib caching from to_mono/to_stereo/to_multi Per review (#2058): the @cache decorator is incompatible with out=, since a cache hit returns the memoized array and skips writing into the caller's buffer, and hashing an uninitialized out= buffer is costly/non-deterministic. Dropping caching keeps the out= contract simple and correct. * Reuse output buffers in to_stereo/to_multi and stream mono downmix Addresses review feedback on #2058: - to_stereo: downmix directly into the output rows via out=, removing a second allocation. Cast left/right to the shared dtype first so the pass-through buffer always matches to_mono's computed dtype. - to_multi: pass the output row into to_mono when no length adjustment is needed; fall back to allocate-and-fix_length only when lengths differ. - stream: reuse a single mono downmix buffer across blocks, writing the final short block into a truncated view. - Add a mixed-dtype regression test for the downmix pass-through. * Relax out= dtype check to accept higher-precision buffers Per review (#2058): rather than requiring an exact dtype match, accept any out= buffer whose dtype can hold the result without loss of precision (promote_types(out.dtype, dtype) == out.dtype). A caller who explicitly allocates a higher-precision buffer keeps that precision; too-narrow buffers are still rejected. Updated docstrings and added a wider-dtype test. --------- Co-authored-by: Brian McFee <[email protected]> * Fix RQA single-axis inputs (#2063) * Fix RQA single-axis inputs * Update tests/test_sequence.py Work around mypy for test type checks --------- Co-authored-by: Brian McFee <[email protected]> * BoundedNorm for specshow on boolean data (#2072) * added boundednorm for boolean data in specshow * Added tests for boundarynorm boolean specshow --------- Co-authored-by: Joren Hammudoglu <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Ale Lloveras <[email protected]> Co-authored-by: Haydn Lam <[email protected]> Co-authored-by: Emily <[email protected]> Co-authored-by: dhaasb <[email protected]> Co-authored-by: Daniel Fernandes <[email protected]> Co-authored-by: Gopesh Pandey <[email protected]> Co-authored-by: T <[email protected]> Co-authored-by: Terry Chen <[email protected]> Co-authored-by: Petra Kühnle <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Cameron Brooks <[email protected]> Co-authored-by: Cameron Brooks <[email protected]> Co-authored-by: ejwong24 <[email protected]> Co-authored-by: Suhas Holla Karkada Chandrashekar <[email protected]>
bmcfee
added a commit
that referenced
this pull request
Jul 21, 2026
* Update version.py * Update ci.yml also CI on PR to 1.0 dev branch * Update docs-ci.yml also doc CI on pr to 1.0.0dev * Test overhaul (#1925) * removed dependency on test1.wav in onset tests * removed dependency on test1.wav in beat tests * properly scoping chirp fixtures * simplifying test_icqt * simplifying onset strength tests * adding some comments on new onset tests * simplifying beat tests * simplifying griffinlim tests * cleaning up gl-cqt tests * further test optimizations * fixing a test naming issue * Reduced nnls_multiblock test * Reduced recurrence matrix test * Reduced hybrid cqt tests * reducing hpss tests * removed test1_22050 from features tests * removing unnecessary example file reference * seeing if skipping ifgram regression drops coverage * Replacing regression fixtures on stft tests * putting in skips for old regression tests * self-contained lpc correctness tests * removing old fixtures from core tests * whoops still need the fixture loader for now * removed fixture files from effects tests * removed fixture files from segment tests * removed fixture files from decompose tests * suppressing known warning in glcqt test * using a fixture for find_files tests * in-place fixture generation for multichannel tests * working on purging srand for proper fixture * purging np.random continues * more rng purging * purging randomness * purging np.random continues * srand purging * random seeding * removed all unseeded randomness in tests * transitioning core tests to internal fixtures * blacking * moving core tests to a new audio clip * migrating display tests to new test track * adding build artifacts for failed image comparison tests * revising image tests once more * progress on purging old tests * simplified tonnetz test * simplified cens tests * rewrite hz/mel tests * rewrote hz_to_octs tests * rewrote melfbnorm tests * marking more tests for removal/rewriting * rewrote centroid tests * finished with met features tests * rewrote chromafb tests * rewrote melfb tests * removing unneded code and documentation * removing unneded code and documentation * rewrote semitone filterbank tests * removed all regression tests * removed the test submodule * keeping regression tests for yin/pyin * copilot suggestions * removing another unused test * added explicit buf_to_float test * fixing coverage * blacked the tests * Peak picking by dynamic programming (#1927) * Update version.py * Update ci.yml also CI on PR to 1.0 dev branch * Update docs-ci.yml also doc CI on pr to 1.0.0dev * DP peak picking * Added tests for dp peak picker * loop optimization * linting * skipping blank lines * linting * added multichannel test for dp peak picker * a couple more extensions to peak picking * upper pinning cmake until the rest of the world gets its act together * updating documentation on peak picker to describe greedy method * spelling * mypy cleanups * removing cmake pin * squashing an unhelpful type check in multichannel * expanding docstring per copilot CR recommendations * Trying a new environment cache setup in CI (#1936) * Trying a new environment cache setup in CI * purging old ci instructions * updated cache setup for docs-ci workflow * fixing a version pin in doc CI * sphinx 8.2 doesn't like something in our multiversion builds. pin it for now * dropping sphinx pin in favor of removing historical builds on 0.9.2 * regexen lol * Variable-rate and non-iterative phase vocoding (#1931) * rewrote phase vocoder * whatever velin... * removed unnecessary unwrap * fixing missing type annotation * Finishing up rough edges on phase vocoder * coverage up on phase vocoder * coverage up on phase vocoder * final copilot recommendations * dropped python 3.8 support for the 1.0 release * bumping numba version requirement * trying to fix the min env * Include scipy-stubs in the test dependencies - v1.0.0 rebase (#1946) * Include `scipy-stubs` in the type-check dependencies https://github.com/scipy/scipy-stubs * remove unused `# type: ignore`s * restrict `scipy-stubs` to `python>=3.10` * mypy no longer ignores missing `scipy` imports * add `scipy-stubs` to the CI linting coda-forge deps * get rid of some squiggly lines * fix remaining ~mypy errors~ squiggly lines * fix `recurrence_matrix` overload for `sparse=True` * workaround a mypy type narrowing bug with `TypeIs` guards * workaround for limitations in mypy's local inference algorithm * workarounds for non-gradual `scipy.io.loadmat` return types * apply review suggestions Co-authored-by: Brian McFee <[email protected]> * post-merge fix of missing Literal annotation --------- Co-authored-by: Brian McFee <[email protected]> * bumping for scipy stubs 1.16.0.1+ * arguing with stubs * arguing with stubs * arguing with stubs * vscale option to specshow (#1938) * some very broken code in here right now * yargh * rewrote scaling code in specshow * correcting docstrings * Reverting a change in power db * fixing type annotation import * blacking and docstring fixes * updated examples to use vscale * updated examples and docstrings * linting * docstyle is strict * Fixed a typo * type hacking * specshow vscale tests * keying mpl image failure artifacts on runner metadata * trying high-precision audio for image tests * bumping error tolerance on the phase plots * adding more information to vscale image test * mpl image tests are hard * expanded docstring for specshow * expanded specshow advanced example * Switched to twilight_shifted as default phase colormap * reworking phase differential plots, updated rainbowgram example * revised plot_display * phase_unwrap_diff -> dphase * added dphase_t mode to specshow * splitting phase from db vscale tests * adding fixture * improving stability of phase display tests * bumping tolerance back up due to font layout quirks * last docstring update to specshow * more docstring fiddling * Remove default quantization for note_to_hz (#1947) * fixes #1945, leave cent deviation in place for note_to_hz * type linting * minor docstring updates * fixed docstring typo * expanded docstring added default behavior test * Freq convert nan (#1948) * fixing #1944 * added tests for nan conversion of notes * Blacking and linting * linting * suppressing bad type checks in nan converters * i don't understand type annotations at all * i can't program either * implemented colorbar helpers * added tests for colorbar helpers * updating docstring examples for new colorbar helpers * updated advanced examples for new colorbar helpers * minor tweaks to colorbar helpers * force a default color gamut on phase plots * type hinting fixes * bumping matplotlib to 3.6 (2022) * reverting to old-style formatter strings * redid image comparison test results * rearranged some code to allow more user overrides * tweaking format strings for aesthetics * fixing docstrings * Colorbar helpers (#1952) * implemented colorbar helpers * added tests for colorbar helpers * updating docstring examples for new colorbar helpers * updated advanced examples for new colorbar helpers * minor tweaks to colorbar helpers * force a default color gamut on phase plots * type hinting fixes * bumping matplotlib to 3.6 (2022) * reverting to old-style formatter strings * redid image comparison test results * rearranged some code to allow more user overrides * tweaking format strings for aesthetics * fixing docstrings * whoops missed a commit * Apply suggestions from copilot code review Mostly cosmetic changes Co-authored-by: Copilot <[email protected]> * fixing a type testing error * Two-slope norm for inferred diverging colormap (#1954) * twoslopenorm diverging colormaps * updating image comparison tests for new diverging cmap style * fixing some buglets * fixing some numerical precision issues in image tests * working on ic test for diverging colormaps * working on ic test for diverging colormaps * Apply suggestions from code review Co-authored-by: Copilot <[email protected]> * relaxing rms on divergence color tests * updating advanced example for display * updating viterbi example with new diverging cmap plot * expanded cmap to support preconstructed colormap objects * fixing type import business * Update librosa/display.py Co-authored-by: Copilot <[email protected]> * bug fixing * fixing coverage * fixing coverage * deprecation rename of cmap * deprecation rename of cmap * Update librosa/display.py More direct dtype check for boolean data in cmap inference Co-authored-by: Copilot <[email protected]> --------- Co-authored-by: Copilot <[email protected]> * Warn on perceptual weighting underflow (#1955) * added warnings for perceptual weighting curves on underflow * fixes #1941 * Warping path to step grid (#1958) * Added path-to-steps fixes #1937 * revising warping example for time synchronization * making the stretched examples more obvious in display * added inverse warping path sampling * linting * linting * quivers * using colors more smartly in stretching example * using colors more smartly in stretching example * quivering * fixing up docs * revising some plots in the advanced gallery * added some unit tests for pathtosteps * Docstring formatting * fudging test data for path conversion to avoid rounding errors * nudging scipy dependency to unstick test failures * final polish on music sync example figures * Fix subsegment example (#1959) * fixed #1950 * improved styling * SPEC07 compliant rng parameters (#1960) * transitioning random_state to rng= for spec07 * suppressing type failures * Update librosa/_typing.py Co-authored-by: Copilot <[email protected]> * fixing old type annotation code * simplified testing, consistentized errors * improved docstring styling * Update tests/test_core.py Co-authored-by: Copilot <[email protected]> * preserve support for old randomstate objects in griffinlim --------- Co-authored-by: Copilot <[email protected]> * nudging min python in setup.cfg * Upload doc build preview artifacts (#1962) * adding an artifact build for pr docs * Trying a different naming * Expire deprecations for 1.0 release (#1963) * removed audioread dependency * removed res_type=None support for vqt and friends * removed get/set_fftlib * removed old filterbanks for cqt * fixing flakes * removed x_axis param from waveshow * removed winlength from yin and pyin * removed audioread from version lister * clarified comment in cqt * blacking * fixing docstring formatting * fixing type annotations now that we are narrowly on scipy fft * removing dead tests * optionalizing dct norms * added generated/ to docs clean target * allow link failure on htk site * Signal mixing helpers (#1964) * progress on mixing * extended to_mono for multiple inputs * added to_multi * fixed a typo * fixed docstring formatting * fixed docstring formatting * examples and docs updates * types are silly * Update librosa/core/audio.py Co-authored-by: Copilot <[email protected]> * Update librosa/core/audio.py Co-authored-by: Copilot <[email protected]> * Update librosa/core/audio.py Co-authored-by: Copilot <[email protected]> * Added upmixing in to_stereo * fixes from copilot cr * added norm parameter to mixers * types are silly * types are silly * CR feedback * Apply suggestions from code review require at least one input for to_mono and to_stereo Co-authored-by: Copilot <[email protected]> * expanded tests for to_mono * Added pad testing for to_mono * test coverage for to_stereo * test coverage for to_multi * adding mixer helpers into more advanced examples --------- Co-authored-by: Copilot <[email protected]> * oct3 display modes (#1967) * first hack at oct3 * added oct3 display modes * added tests for oct3 modes * trying again without constrained layout * improving docstring and expanding vqt axis compatibility support * fixing image comparison test to support matplotlib 3.10.5 * Inverted waveshow (#1971) * added a proxy patch for waveshow label containment * First cut at inverted waveform display * tests for inverted waveshow * more fancy footwork with matplotlib artists * [CR] Wavebars (#1973) * first cut at wavebars * linting * linting * added basic wavebar tests * added basic wavebar tests * removing dependency on compressed layout for wavebar test * removing dependency on compressed layout for wavebar test * nudging mpl dependency * nudging mpl dependency * nudging mpl dependency entirely * docstring improvements * chore: Refine .gitignore for improved accuracy and coverage (#1969) This commit refines the `.gitignore` file by: - Removing redundant entries. - Improving comments for better readability. - Reorganizing and categorizing ignored files for better maintainability. - Consolidating testing and checkpoint related entries. * Adding plt.show() at the end of code blocks (#1979) * Fix types in beat tracking code (#1977) The development version of Numba 0.62.0 is unhappy about `range` being applied to floating point values here. * Fix #1970: Prevent AdaptiveWaveplot adaptor from being garbage collected (#1992) - Add _WAVESHOW_ADAPTORS WeakKeyDictionary to retain adaptors per-Axes - Store artists (steps, envelope) and axes as weakrefs in AdaptiveWaveplot - Add properties to preserve existing attribute API - Add GC guard in update() for safety - Add 3 regression tests * Add Metrogram Transform (#1986) * feat: add metrogram transform * allow passing no aggregate, improve documentation * feat: add interp_broadcast Reference: #1909 * show metrogram in html docs * show interp_broadcast in html docs * fix: linting issues * use keywording,reorder x1/x1_pos, check axis range * generalize axis for metrogram transform also fix a logic bug with checking index bounds * overload interp_broadcast when func = None * fix interp_broadcast not accepting negative axis * formatting * set np.expand_dims to -3 instead of using a param * add metrogram example * add interp_broadcast example * add tests for metrogram * rename func -> op in interp_broadcast * simplify interp_broadcast example choosing to keep the example minimal like the other examples in utils an exhaustive use of all parameters is better suited for the tests * add tests for interp_broadcast * metrogram example: keep the zero in funt_freqs * remove duplicate imports added unintentionally * remove blank line after doc strings * add tests for interp_broadcast axis * update metrogram docs * Fix typo Co-authored-by: Copilot <[email protected]> * Update librosa/feature/rhythm.py Co-authored-by: Copilot <[email protected]> * Update librosa/util/utils.py Co-authored-by: Copilot <[email protected]> * Update librosa/util/utils.py Co-authored-by: Copilot <[email protected]> * fix: input shape validation based on broadcasting rules Co-authored-by: Brian McFee <[email protected]> * docs: type change kind: str -> kind: _InterpKind --------- Co-authored-by: Copilot <[email protected]> Co-authored-by: Brian McFee <[email protected]> * Update type annotations (#1980) * nudging matplotlib min version for spec0 * needs mpl in lint environment now * type hacking * type hacking * type annotations are the worst * type hacking continues * type hacking continues * types types types * bumping numpy for env compatibility * fixing scipy sparse indexing type annotations * type hacking * mypy is stupid * mypy is stupid * mypy is stupid * strengthening tests stupidly for mypy * Fixing type annotations in display and tests * upper pinning sphinx dependency until sphinx-multiversion is fixed * trying to address copilot reviews * more type annotation nonsense * Trying to fix numba/numpy compatibility for test environments * type annotations * Migrate from SciPy spmatrix to sparray (#1994) * [ENH] Migrate from scipy.sparse matrix to array interface (#1902) * spell check * spell check * Remove redundant empty-index check in sparsify_rows; add test for 1D sparse shear rejection * Fix 1D sparse shear test for scipy < 1.14 compatibility * Address review: accept sparse matrices in shear, vectorize sparsify_rows * Suppress scipy-stubs attr-defined errors in sparse matrix shear test * Python 3.14 test support (#1999) * adding 3.14 to test matrix * is it a name mismatch problem? * forcing python versions back into place * Trying to force a source build on the samplerate package * reverting source build for samplerate, it doesn't work * nudging python for docs build to 3.11 * nudging python for docs build to 3.11 * bypassing zenodo linkchecks * updated libsndfile link in docs * feat: implement hybrid_tempogram (#1991) (#2004) * feat: implement hybrid_tempogram per Peeters 2005 * refactor: address CR feedback, explicit kwargs, finite interpolation * Merge branch '1.0.0dev' into feature/hybrid-tempogram and resolve conflicts * Merge upstream 1.0.0dev and resolve conflicts * Merge upstream 1.0.0dev and fix typo in metrogram * Resolve rhythm imports and add to __all__ * style: fix linting and restore metrogram * test: add unit tests for hybrid_tempogram coverage * fix: use fourier_tempo_frequencies for hybrid_tempogram freq grid * fix: import fourier_tempo_frequencies * fix: remove non-english text from docstring to pass codespell * refactor: simplify hybrid_tempogram array handling and fix kwargs * style: remove trailing whitespace in docstring to pass velin linter * fix: resolve mypy type hints for interp1d and remove duplicate decorator * refactor: add type hints, docstring examples, and optimize interp kwargs per review * refactor: apply robust type hinting, clean docstrings, and array return per strict mypy checks * fix: finalize hybrid_tempogram with strict typing, optimized interpolation, and test coverage * fix: remove assume_sorted=True to fix interpolation on descending lag domain * refactor: apply reviewer feedback for optimizations and test fixtures * Update librosa/feature/rhythm.py Co-authored-by: Copilot <[email protected]> --------- Co-authored-by: Brian McFee <[email protected]> Co-authored-by: Copilot <[email protected]> * Document specshow frequency axis scaling (#2006) Co-authored-by: Terry Chen <[email protected]> * Surface contributing guide in docs site, link build instructions (#2009) * Surface contributing guide in docs site, link build instructions Adds myst-parser to render CONTRIBUTING.md inside the docs sidebar, so first-time contributors can find the contributor guide without leaving the docs site. Also updates CONTRIBUTING.md to reference docs/BUILD.md for documentation build instructions, replacing a stale reference to a doc/README that no longer exists. Fixes #1800. * Add myst-parser to docs env * fixed broken link in contributing guidelines --------- Co-authored-by: Brian McFee <[email protected]> * Multiplot (#2012) * first cut at multiplot * multiplot cleanups * type annotating * example fixups * added legend aggregator, fixing sphinx configs * fixing type hints * support variadic multiplot inputs * polishing multiplot * velinizing * fixing singleton multiplots * Added legend_for_axes tests * multiplot helper tests * updated multilegend plots to be more informative * coverage up * coverage up * fixing type checks in display tests * test blacking * tying up loose ends * covering legend loc inference * multiplot coverage * adding fixtures for multiplot * Added titles= support to multiplot * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * improving on copilot suggested test * improved docstrings * velinizing * added example for legend_for_axes * simplified docstring * fixing #2013 * renaming function aliases in multiplot * expanded unit tests to cover multiplot titles * expanded legend_for_axes docstring to clarify loc quirkiness * fixing type annotations for optional labels and titles * working on stefanv's comments * Accept user-supplied property cycles in multiplot * fixing positional argument in display tests * simplified docstring examples * updated docstring on multiplot orient * finished docstring fixups --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Spec0 2026.06 (#2018) * attempting spec0 compliance * attempting spec0 compliance * revised min doc build version in build.md * added dead batteries to doc build matrix * bumped numba version * trying with strict in conda env * bumping packaging to 23.2 for python 3.12 support * bumped soxr to 1.0.0 * fixing environment naming while i'm at it * Bumping msgpack-python for environment consistency * bumping lazy and typing_extensions * bumping min versions again for py3.12 environment compatibility * bumping versions to ensure python 3.12 compatibility * removed version constraint on scipy stubs * removed resource management shim dead branch * cleanups around python version handling * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * display.wavef0 (was: melodyplot) (#1976) * added mask to waveshow * fixing stuff * melody plot first cut * added missing docstring update * Added test coverage * nudging to mpl 3.8 * Temporarily nudging ci minimal to mpl 3.10 to see if that works * Temporarily nudging ci minimal to mpl 3.10 to see if that works * Temporarily nudging ci minimal to mpl 3.8.4 to see if that works * should fix wavef0 inverted mode * i can't program in python apparently * nudging back to mpl 3.8 for datalim hacking * trying mpl 3.9 * trying mpl 3.9 * trying mpl 3.9 * working on making mypy happy * more mypy business * fixed typing in wavef0 * added round trip test for transformf0 * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <[email protected]> * revised docstring * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * velinizing * protect transformf0 from nonpositive f0 values * streamlined wavef0 norm calculation * copilot cr comments * copilot comments on docstrings * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * rearranging error checks * fixing type checking * fixed display.__all__ * fixing thinko in f0 bounds checking * Added docstrings to transformf0 methods * minor docstring updates * linting * contorting around edge cases in wavef0 * fixing numerical problems in wavef0 edge cases * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * added cqt overlay example in wavef0 docstring --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Yet another Viterbi acceleration PR (#2007) * removed over alloation in viterbi core * temporary hack faster viterbi * pyin speedups * added safety barrier on transition thresholding * added tests for viterbi acceleration * one more test case coverage * Streamlining changes * polishing viterbi optimization * Updates for mypy 2.1 (#2025) * making mypy happy in 2.1 * making mypy happy in 2.1 * mypy making us jump through hoops again * mypy making us jump through hoops again * making mypy happy in 2.1 * mypy is happy for now * maybe removing numpy typing plugins was premature * making mypy happy on minimal numpy * fidgeting some type annotations to work across multiple numpy versions * fixing some import silliness in tests * undoing some bad optimizations * fixes #2014 (#2026) * Modern linting (#2027) * migrating to ruff * migrating to ruff * down into line length constraints * removed unused imports * whitespace fixing * import ordering via ruff * pyi linting * setting output format * fixing quote styles * working on bugbear * bugbear tests * simplify linting * ruffing * ruff lint rules * Dynamic vectorize (#2028) * dynamic vectorization in beat * dynamic vectorization in core * dynamic vectorization in core * dynamic vectorization in utils * removed all static vectorized signatures * fixing some type business * rewrote abs2 for more direct dtype handling * cleaning up some typing code * more mypy tweaking * Even faster import (#2029) * being better about enforcing type checking blocks * localizing heavy imports to accelerate package import * more typechecking blocks * fixed missing import * purging typing extensions * stricter typecheck blocks * lingering type block * Negative offset load (#2030) * first cut at negative offset loading * added flac example and fixture for negative load offset test * marking ogg seek tests as optional pass * added negative offset load in stream * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <[email protected]> --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Loadx (#2031) * added loadx * updated examples to use loadx where appropriate * updated examples to use loadx where appropriate * Clarified docstring for loadx * Docstring formatting * typing and docstring improvements from copilot * more helpful failure mode in loadx (#2032) * Modernize types in `librosa._typing` (#2035) * Modernize types in `librosa._typing` * remove sudden unused `# type: ignore[misc]` * forcing upgrade to codecov action * modernizing type annotations for the future (#2036) * mdoernizing type annotations for the future * loosened util.tiny type annotation just a wee bit * more type annotation modernization (#2038) * Fill in missing type annotations (#2037) * Fill in missing type annotations in `display.py` * Fill in missing type annotations in `core/intervals.py` * Fill in missing type annotations in `beat.py` * remove expired `librosa.beat.tempo` deprecation * Display Highlighter (#2034) * first cut at display.highlight * updated examples to use highlighter * more example updates * added tests for highlighter * highlighter tests * highlighter test images * spelling * velinizing * path coverage * copilot review comments * copilot review comments * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <[email protected]> * one more copilot suggestion * modernized annotations * type modernizing --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * fixed #2040 (#2041) * avoid turning scalars into 0-d arrays in `core.{convert,spectrum}` and `util.utils` (#2039) * `core.convert`: avoid turning scalars into 0-d arrays * `core.spectrum`: avoid turning scalars into 0-d arrays * `util.utils`: avoid turning scalars into 0-d arrays * fix irreproducible mypy error * Waveplot styling (#2044) * style waveplot proxy after steps, not fill * fixed #2042 * updated tests * cleanups in the new waveplot legend code * styling and ruffing * Final polishing on waveplot legend code * improved `dtype_r2c` and `dtype_c2r` typing in `util.utils` (#2043) * Doc check modernizing (#2047) * switched velin to numpydoc lint * configuring numpydoc lint * revising numpydoc configs to avoid hobgoblins * starting to get somewhere with numpydoc linting * fixing exceptions to numpydoc linting * [WIP] Fix failing GitHub Actions job for lint and code analysis (#2048) * Initial plan * Apply remaining changes --------- Co-authored-by: copilot-swe-agent[bot] <[email protected]> * working through numpydoc corrections * working through numpydoc corrections * working through numpydoc corrections * fixing numpydoc lint * fixing exceptions to numpydoc linting * numpydoc is all passing --------- Co-authored-by: Copilot <[email protected]> * Contributing and agents update (#2017) * updating contributing guidelines with ai policy * Added agents.md * Fix CONTRIBUTING.md markdown formatting * revising contributing guidelines * integrating comments * formatting * updated contributing guidelines for numpydoc linting * reformat `**/__init__.pyi` re-exports (#2049) * Include str dtype in `ndarray` return type annotations (#2050) * Multichannel db (#2051) * implemented axis targeting for db conversion * db calculation now works properly with target axes * fixed scalar processing for db * added 'auto' mode to db conversion axes inference * fixed regression in amplitude_to_db * fixed regression in amplitude_to_db * copilot reviews * fixing some type annotations * copilot comments * cleaning up tests * pytest nonsense * Fix decorator callable types in `librosa.util.decorators` (#2053) * Improve `ndarray` return types (#2052) * Generic shaped `ndarray` type aliases * Improve `ndarray` return types in `librosa.core` * Improve `ndarray` return types in `librosa.util` * Improve `ndarray` return types in `librosa.beat` * Improve `ndarray` return types in `librosa.display` * Improve `ndarray` return types in `librosa.effect` * Improve `ndarray` return types in `librosa.filters` * Improve `ndarray` return types in `librosa.segment` * Improve `ndarray` return types in `librosa.sequence` * Fix mypy errors with `numpy<2.4` * Ignore false positive mypy error in tests related to `np.object_` array iteration * Address review comments for the typing changes in `util.utils` * Work around a mypy bug where it fails to expand `bool` into `Literal[True] | Literal[False]`. * Support matplotlib 3.11 in tests (#2054) * redoing tests for matplotlib 3.11, tightening tolerance where possible * updated failure artifact upload recipe * loosening tests more on mpl 3.10 * bypassing image comparison tests on 3.10 completely * fixed types for legend in matplotlib 3.11 * Stream resampling (#2055) * initial cut of resampling stream passes previous tests * progress on stream resampling * clarified docstring, reverted changes * updated tests * feature complete stream resampling * refactored tests * better coverage for edge cases * working in some copilot review comments * more copilot suggestions * copilot and gemini comments * cleaning up stream tests * tightening memory allocations in stream * added a future default change decorator and bundled onto stream * forgot to kw-only a parameter name * ratcheting down default stream parameter setting deprecation to 1.1 * last round of copilot reviews * massive simplification of legend_for_axes, fixes #2057 (#2059) * Refactor frequency formatters (#2060) * fixed minor ticking for oct3 modes * numpydoc linting * copilot review comments * copilot review comments * Updating actions (#2065) * updating actions * fixing an out= allocation issue * more warnings cleanup * libsamplerate comes via conda now * upload-artifact update * disable auto-activate * Standardize n_fft parameter docstrings (#2068) * Standardize n_fft parameter docstrings Rebuilt against 1.0.0dev after retargeting. Standardizes the analysis n_fft descriptions on 'length of the FFT window', fixes the bins/window conflations (piptrack, estimate_tuning, filters.mel, filters.chroma, feature.inverse), clarifies istft's frame-length inference, corrects mel_to_stft's documented return shape, and fixes frequency-axis references that assumed 2D input (istft win_length, phase_vocoder hop_length). Incorporates review feedback: 'frame' terminology in get_duration; audio-output functions reference the intermediate STFT without a bins clause. * Use 'frame' terminology for n_fft descriptions Per review: n_fft refers to frame length, not window length, since a window can be shorter than the frame. Swept all standardized n_fft descriptions accordingly. --------- Co-authored-by: Cameron Brooks <[email protected]> * precise dtype annotations in `librosa.beat` (#2069) Co-authored-by: Brian McFee <[email protected]> * Infer n_bins automatically in cqt/vqt (#2066) * added n_bins=None to cqt/vqt * added n_bins=None to cqt/vqt tests * linting * revisions * revisions * one more test * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Added nbins inference to hybrid and pseudo cqt * linting * copypasta * fixed #2071, tightened icqt tests * revised downsampling logic in vqt * copilot review comments --------- Co-authored-by: Copilot Autofix powered by AI <[email protected]> * Add out= parameter to to_mono for in-place downmixing (#2058) * Add out= buffer parameter to to_mono, to_stereo, to_multi Lets callers pass a pre-allocated output buffer to the mixing functions, avoiding a fresh allocation per call (e.g. per-block downmix in stream()). A shared _init_output helper validates the buffer's shape and dtype against the computed output, zero-fills it, and returns it; out=None preserves the previous allocate-fresh behavior. Implements #2056 against the redesigned 1.0 mixing API. * Remove joblib caching from to_mono/to_stereo/to_multi Per review (#2058): the @cache decorator is incompatible with out=, since a cache hit returns the memoized array and skips writing into the caller's buffer, and hashing an uninitialized out= buffer is costly/non-deterministic. Dropping caching keeps the out= contract simple and correct. * Reuse output buffers in to_stereo/to_multi and stream mono downmix Addresses review feedback on #2058: - to_stereo: downmix directly into the output rows via out=, removing a second allocation. Cast left/right to the shared dtype first so the pass-through buffer always matches to_mono's computed dtype. - to_multi: pass the output row into to_mono when no length adjustment is needed; fall back to allocate-and-fix_length only when lengths differ. - stream: reuse a single mono downmix buffer across blocks, writing the final short block into a truncated view. - Add a mixed-dtype regression test for the downmix pass-through. * Relax out= dtype check to accept higher-precision buffers Per review (#2058): rather than requiring an exact dtype match, accept any out= buffer whose dtype can hold the result without loss of precision (promote_types(out.dtype, dtype) == out.dtype). A caller who explicitly allocates a higher-precision buffer keeps that precision; too-narrow buffers are still rejected. Updated docstrings and added a wider-dtype test. --------- Co-authored-by: Brian McFee <[email protected]> * Fix RQA single-axis inputs (#2063) * Fix RQA single-axis inputs * Update tests/test_sequence.py Work around mypy for test type checks --------- Co-authored-by: Brian McFee <[email protected]> * BoundedNorm for specshow on boolean data (#2072) * added boundednorm for boolean data in specshow * Added tests for boundarynorm boolean specshow --------- Co-authored-by: Joren Hammudoglu <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Ale Lloveras <[email protected]> Co-authored-by: Haydn Lam <[email protected]> Co-authored-by: Emily <[email protected]> Co-authored-by: dhaasb <[email protected]> Co-authored-by: Daniel Fernandes <[email protected]> Co-authored-by: Gopesh Pandey <[email protected]> Co-authored-by: T <[email protected]> Co-authored-by: Terry Chen <[email protected]> Co-authored-by: Petra Kühnle <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Cameron Brooks <[email protected]> Co-authored-by: Cameron Brooks <[email protected]> Co-authored-by: ejwong24 <[email protected]> Co-authored-by: Suhas Holla Karkada Chandrashekar <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
As promised in #2050, here are a bunch of other typing improvements related to
ndarrayreturn types.There are still (many) other
ndarrayreturn types that could still be improved, but those are all input-dependent in either their shape-type or dtype. So this PR only covers the "absolute" cases where the dtype and/or shape-type of the returnedndarrayis always the same (with a couple of exceptions). I'll leave those input-dependent return types (not to be confused with the type-theoretical"dependent types", which are types that depend on values, and here we're talking about output types depending on input types), as they'll require a bit more effort (generic functions, a.k.a. parametric polymorphism) to tackle, and this PR is already larger than I'm usualy comfortable with haha.
Oh and in case you were wondering why I removed those
[Any]type arguments fornp.floatingand other abstract generic numpy scalar types; those are no longer needed in recent numpy versions (deprecated, even). Without thoseAnys it's also a biteasierless difficult too read IMO.