xhochy / fletcher

Pandas ExtensionDType/Array backed by Apache Arrow

Home Page:https://fletcher.readthedocs.io/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Visualize pandas performance comparisons

xhochy opened this issue · comments

We have several benchmarks like https://github.com/xhochy/fletcher/blob/a63581d10381a41595695a9c3c89edd156375f74/benchmarks/take.py that compare the performance of a specific method of plain pandas with the implementation in fletcher. The performance difference is not covered by the standard plots that asv provides.

We should therefore:

  • Run the benchmarks and produce machine-readable output
  • Parse the output and make plots in a notebook that show the performance differences (preferably use altair for plotting).
  • Have a way to publish the run notebook as static HTML somewhere.