Fixed the problem that caused installation to fail.

[decfrr]

What

Fix #126

Why

The problem was due to an unset version limit for numpy. With the release of version 2.0.0 on the 16th of this month, which contained some destructive changes, the build of this package was failing. By setting the upper version limit, the build will succeed.

Tasks

• [x] Passed all tests
• [x] Successful build

Successfully built nnmnkwii pysptk fastdtw
Installing collected packages: tqdm, threadpoolctl, scipy, joblib, fastdtw, decorator, cython, scikit-learn, pysptk, nnmnkwii
Successfully installed cython-3.0.10 decorator-5.1.1 fastdtw-0.3.4 joblib-1.4.2 nnmnkwii-0.1.2+67ff5ff pysptk-0.2.2 scikit-learn-1.5.0 scipy-1.13.1 threadpoolctl-3.5.0 tqdm-4.66.4

[r9y9]

Looks good to me. Can you apply the changes in #128 to fix CI failures?

Run actions/setup-python@v1
Error: Version 3.9 with arch x64 not found

[decfrr]

Thanks for the confirmation. I have merged and imported the contents of the patch. Please confirm.

[r9y9]

umm, looks like numpy 2.0.0 is installed on CI 🤔

Using cached numpy-2.0.0-cp39-cp39-macosx_14_0_arm64.whl (5.2 MB)

[decfrr]

I guess when we try to build this package using non version 2.0 numpy, we need to remove the cache in advance.

[decfrr]

Also, rather than specifying less than a specific version, I thought I should specify a version less than 2.0. I don't think this is a point to be particular about since it is a primary response, but I think it is better this way, so I changed it.

[decfrr]

I guess when we try to build this package using non version 2.0 numpy, we need to remove the cache in advance.

I too had a problem with ci downloading numpy 2.0. I will investigate.

It seems not a package requirements problem, because if we try to install nnmnkwii from the git, like pip install git+https://github.com... pip use numpy under version 2.0 and build succsessfully. Umm...

[r9y9]

I guess I found the reason: setup_requires is deprecated

https://setuptools.pypa.io/en/latest/userguide/dependency_management.html

In previous versions of setuptools, this used to be accomplished with the setup_requires keyword but is now considered deprecated in favor of the PEP 517 style described above. To peek into how this legacy keyword is used, consult our guide on deprecated practice (WIP).

so, what about removing setup_requires and instead add the numpy requirement in install_requires?

[decfrr]

By using install_requires instead of setup_requires, the problem of numpy version had been solved.

Also, some test problems had occured, I fixed them.

1. np.int type has been deprecated https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
2. change error type RuntimeError -> ValueError of wrong number test
3. Clarified librosa.effects.time_stretch argument

And I found a error of deprecated field np.object, I fixed it.

[r9y9]

Thank you for the fixes! CI still fails on python 3.7 but I think it's OK to drop python 3.7 from CI. Shall we check all tests are passing except for python 3.7?

[decfrr]

I see that the test fails in 3.7. I'd like you to run the test on other versions just to be sure.

[r9y9]

it works fine with me locally with python 3.8, numpy 1.24.3, OSX. I needed https://github.com/r9y9/nnmnkwii/pull/128/commits/907162ed95aeaafeab2422d0cc9bd290c1ea9505 for tests on OSX but everything else was OK with this PR.

[r9y9]

Testing other versions/platforms with github actions (https://github.com/r9y9/nnmnkwii/pull/128) and it is working OK too. I'll merge this PR after all CI gets green by either

1. dropping python 3.7 (easiest; I'm OK. It's already EOL https://devguide.python.org/versions/)
2. another idea if you prefer

[decfrr]

I have merged the changes of including dropping python 3.7 from ci. It seems that all CI will complete jobs successfully.

[decfrr]

Thank you for valuable opportunity. I appereciate your ongoing maintenance.