Peter Steinbach

added a small flag that allows to filter per row (not yet ported to geojson), this PR is now ready for review and potential merge

@ionelmc can you elaborate on setuptools_scm ? I was not so happy with bumpversion automagically setup with this cookiecutter. Especially when SIGTERMing bumpversion all kinds of weird stuff happened. that...

See #103 for another thread on this. I feel `versioneer` is a bit more flexible than setuptools_scm as it doesn't require carrying the relative path to the top level folder...

maybe use a [dictionary based choice](https://cookiecutter.readthedocs.io/en/latest/advanced/dict_variables.html)? With this one could collect all the coverage tools and add a short help text.

Ah, that didn't occur to me that `sklearn` offers such functionality. But triggered by your comment @maumueller , I found that [`sklearn.neighbors.NearestNeighbors`](https://scikit-learn.org/stable/modules/generated/sklearn.neighbors.NearestNeighbors.html#sklearn.neighbors.NearestNeighbors) offers the value `brute` for the `algorithm` parameter....

I had a look at this [see my benchmark code](https://gist.github.com/psteinb/90a8728df5b7d86cb5fafb9bd5e94d9f) with 64k samples of 12dims each (128 queries only). ``` ------------------------------------------------------------------------------------------- benchmark: 4 tests ------------------------------------------------------------------------------------------ Name (time in ms) Min...

Thanks for moving this over. I'll mark this with help wanted so that potential newcomers know where to potentially start.

Yes, yes, yes! You are right on the spot. In [HPC-in-a-day](https://psteinb.github.io/hpc-in-a-day), I did use a narrative which needs some improvement perhaps. I am planning on sending PRs based on that,...

+1

Let's go with this approach. I like it.