wannesm / dtaidistance

Time series distances: Dynamic Time Warping (fast DTW implementation in C)
Other
1.09k stars 184 forks source link

Add badge and notes for conda-forge #143

Closed m-rossi closed 2 years ago

m-rossi commented 2 years ago

Thanks a lot for your package. One of my students started using it and as our infrastructure relies heavily on conda-forge I added a recipe for your package. This has the advantage as you do not need to have a compiler on your machine to use the fast C-version of your package.

I also have a question - Is cython really needed during runtime: https://github.com/wannesm/dtaidistance/blob/master/setup.py#L340? In the recipe I remove that dependency as I thought this may be a mistake and cython is just needed during build.

If you like I can also add you as a recipe maintainers if you like. If not, don't mind, I will happily maintain the recipe.

wannesm commented 2 years ago

Hi @m-rossi , we're not using Conda ourselves so your conda recipe is much appreciated and I have merged the links. I would appreciate being added as a recipe maintainer, mainly to learn about this system.

You are right that the cython dependency in the setup.py should not be required, only in the toml file, and I've removed it from setup.py. There are some suboptimal parts in the setup.py as a result of efforts to try to get it to work on multiple platforms. Especially the link to OpenMP. Are the precompiled conda versions with OpenMP the parallelization?

m-rossi commented 2 years ago

I created https://github.com/conda-forge/dtaidistance-feedstock/issues/2, you should be added to the maintainers list sometime today.

In the initial version of the recipe I have not configured anything regarding OpenMP. There is a section in the conda-forge docs: https://conda-forge.org/docs/maintainer/knowledge_base.html#openmp I started adding explicit support in https://github.com/conda-forge/dtaidistance-feedstock/pull/3