Simpler interface for computing distances

[bobot]

The correlation function in polars has a simple interface. It takes two polars expressions as arguments. It allows to compute those column dynamically.

On the other hand polars-distance seems to really only compute the distance on arrays. On list for example it is turning it into a set before computing something. It makes impossible to filter the datas, or grouping them before computing the distance (since arrays have a fixed length and can't be computed dynamically).

For example, computing the correlation matrix of the time taken by different algorithms is easy:

df = cross_results.group_by("algo1","algo2").agg(corr=pl.corr("cpuTime1","cpuTime2", method="pearson"))

I don't see how to do it using distances defined by polars-distance.

Thanks for this library it can be a great addition to polars.