[PKG-3035] umap-learn 0.5.4

[Arishamays1]

PKG-3035 umap-learn 0.5.4

upstream: https://github.com/lmcinnes/umap/tree/0.5.4
dependencies: https://github.com/lmcinnes/umap/blob/0.5.4/setup.py

destination channel: defaults

conda-forge: https://github.com/conda-forge/umap-learn-feedstock

actions:

• clean the meta.yaml, remove cross compilation
• use LICENSE.txt file from the upstream
• patch setup.py to remove tbb dependency, which was causing pip check to fail and would leave the user's environment not clean.
• add tests (source from GitHub archive)
• do not install tensor-flow for some tests (jax is missing for some tensor-flow python versions)
• set numpy<1.24.0 (see recipe)
• skip 1 test for ppc64le which gives seg fault
• dependency for Update 0.15.0 bertopic-feedstock#2

other information about the tbb problem:

• umap 0.5.4 is causing a broken package error in Conda feedstock builds lmcinnes/umap#1054
• tbb optional requirement should be configurable lmcinnes/umap#1060 (comment)
• Using TBB with numba fails with pip install, works with apt install numba/numba#7148

[anaconda-pkg-build]

Linter check found the following problems:

The following problems have been found:

WARNING: clone/recipe/meta.yaml:60: documentation_specifies_language: Use the generic link, not a language specific one
Warnings were found

[anaconda-pkg-build]

Linter check found the following problems:

The following problems have been found:

WARNING: clone/recipe/meta.yaml:60: documentation_specifies_language: Use the generic link, not a language specific one
Warnings were found

[anaconda-pkg-build]

Linter check found the following problems:

The following problems have been found:

WARNING: clone/recipe/meta.yaml:60: documentation_specifies_language: Use the generic link, not a language specific one
Warnings were found

[anaconda-pkg-build]

Linter check found the following problems:

The following problems have been found:

WARNING: clone/recipe/meta.yaml:61: documentation_specifies_language: Use the generic link, not a language specific one
Warnings were found

[anaconda-pkg-build]

Linter check found the following problems:

The following problems have been found:

WARNING: clone/recipe/meta.yaml:62: documentation_specifies_language: Use the generic link, not a language specific one
Warnings were found

[varlackc]

There is currently a pull request for this package that is open and needs to be closed

it is version 0.5.3 and the pull request number is 1
here is a link to the pull request:
#1

[anaconda-pkg-build]

Linter check found the following problems:

The following problems have been found:

WARNING: clone/recipe/meta.yaml:62: documentation_specifies_language: Use the generic link, not a language specific one
Warnings were found

[skupr-anaconda]

@Arishamays1 @lorepirri You can disable pip check as tbb is not a required dependency. The source code hasn't mentioned any usage of it. It's required by numba, see lmcinnes/umap#1060 (comment).
And leave a comment about why we disable it

[lorepirri]

@Arishamays1 @lorepirri You can disable pip check as tbb is not a required dependency. The source code hasn't mentioned any usage of it. It's required by numba, see lmcinnes/umap#1060 (comment). And leave a comment about why we disable it

Thank you @skupr-anaconda . Since the conclusion is that tbb is optional, I did patch the setup.py removing the tbb dependency, so now pip check runs without problems (in case the user wants to have a clean installation).

[skupr-anaconda]

@Arishamays1 @lorepirri You can disable pip check as tbb is not a required dependency. The source code hasn't mentioned any usage of it. It's required by numba, see lmcinnes/umap#1060 (comment). And leave a comment about why we disable it

Thank you @skupr-anaconda . Since the conclusion is that tbb is optional, I did patch the setup.py removing the tbb dependency, so now pip check runs without problems (in case the user wants to have a clean installation).

But we should add tbb to run for x86_64

[skupr-anaconda]

LGTM

[skupr-anaconda]

@lorepirri Did you forget to add abs.yaml with:

upload_channels:
  - sfe1ed40 

Because of:
conda search umap-learn -c sfe1ed40

Loading channels: done
# Name                       Version           Build  Channel
umap-learn                     0.5.3 py310hca03da5_0  pkgs/main
umap-learn                     0.5.3 py311hca03da5_0  pkgs/main
umap-learn                     0.5.3  py38hca03da5_0  pkgs/main
umap-learn                     0.5.3  py39hca03da5_0  pkgs/main

[lorepirri]

@lorepirri Did you forget to add abs.yaml with:

upload_channels:
  - sfe1ed40 

Because of: conda search umap-learn -c sfe1ed40

Loading channels: done
# Name                       Version           Build  Channel
umap-learn                     0.5.3 py310hca03da5_0  pkgs/main
umap-learn                     0.5.3 py311hca03da5_0  pkgs/main
umap-learn                     0.5.3  py38hca03da5_0  pkgs/main
umap-learn                     0.5.3  py39hca03da5_0  pkgs/main

@skupr-anaconda originally 0.5.3 was for defaults

• https://anaconda.atlassian.net/browse/PKG-2191
• [PKG-2191] umap-learn 0.5.3 #2

the command conda search umap-learn -c sfe1ed40 shows also the ones in defaults, notice the column Channel:

% conda search umap-learn
Loading channels: done
# Name                       Version           Build  Channel             
umap-learn                     0.5.3 py310hca03da5_0  pkgs/main           
umap-learn                     0.5.3 py311hca03da5_0  pkgs/main           
umap-learn                     0.5.3  py38hca03da5_0  pkgs/main           
umap-learn                     0.5.3  py39hca03da5_0  pkgs/main

For example, another package that is only on SF, json_stream, will have the SF channel:

% conda search json_stream -c sfe1ed40
Loading channels: done
# Name                       Version           Build  Channel             
json_stream                    2.3.2 py310hca03da5_0  sfe1ed40            
json_stream                    2.3.2 py311hca03da5_0  sfe1ed40            
json_stream                    2.3.2  py38hca03da5_0  sfe1ed40            
json_stream                    2.3.2  py39hca03da5_0  sfe1ed40

[Arishamays1]

LGTM

[M-Waszkiewicz-Anaconda]

LGTM

[lorepirri]

@Arishamays1 I will merge this one and upload it.