librosa.load returns wrongly-typed int object

[tarepan]

Describe the bug
librosa.load returns int object instead of float.

To Reproduce

import librosa

import librosa
wave, sr = librosa.load("test.wav", sr=None, mono=True)
print(type(sr))
# <class 'int'>

Expected behavior
Typing of .load says that it returns float.
https://github.com/librosa/librosa/blob/09e4a622456bbad314e2e6d5c75879f56d728b9b/librosa/core/audio.py#L66C1-L66C31
So, the function should return float object.

Software versions

Linux-5.15.109+-x86_64-with-glibc2.35
Python 3.10.12 (main, Jun 11 2023, 05:26:28) [GCC 11.4.0]
NumPy 1.23.5
SciPy 1.10.1
librosa 0.10.1
INSTALLED VERSIONS
------------------
python: 3.10.12 (main, Jun 11 2023, 05:26:28) [GCC 11.4.0]

librosa: 0.10.1

audioread: 3.0.0
numpy: 1.23.5
scipy: 1.10.1
sklearn: 1.2.2
joblib: 1.3.2
decorator: 4.4.2
numba: 0.56.4
soundfile: 0.12.1
pooch: v1.7.0
soxr: 0.3.6
typing_extensions: installed, no version number available
lazy_loader: installed, no version number available
msgpack: 1.0.5

numpydoc: None
sphinx: 5.0.2
sphinx_rtd_theme: None
matplotlib: 3.7.1
sphinx_multiversion: None
sphinx_gallery: None
mir_eval: None
ipython: None
sphinxcontrib.rsvgconverter: None
pytest: 7.4.0
pytest_mpl: None
pytest_cov: None
samplerate: None
resampy: None
presets: None
packaging: 23.1```

[bmcfee]

Thanks for catching this! It's a rather subtle thing, as the value and type of sr will depend on which codec path is taken (audioread vs soundfile), which resampling backend is used, etc.

However, there are a couple of things that should be fixed here:

1. the default sr has value 22050 (int) and type Optional[float]. We should make this 22050.0.
2. We should explicitly cast sr to float on return.

These are pretty easy fixes, and I don't anticipate any downstream effects.

[tarepan]

I don't anticipate any downstream effects.

I have some concern.
I notice this bug when I use float specific method (float.is_integer()).
Integer also have int-specific methods (method list).
Semantically this is subtle thing, but practically this could have some downstream effects (of course, very minor case).

I totally agree with your bug-fix plan, but we should emphasis backward-incompatibility notice in release note / change log.

[bmcfee]

Ah, to clarify, I meant unintended consequences for the rest of the library, not for downstream code.

[tarepan]

Oh, I misunderstanded it, now I got it.

Can I make a pull request with your bug-fix plan?

[bmcfee]

Sure, thanks!