I ran into a bug writing tests for our application. We generally convert Spark DataFrames (SDF) to H2OFrames (HDF), but in the test I was using Pandas DataFrame (PDF) conversions to HDFs.
We had two lines of code to replace values by a row mask:
If `frame` was generated from a PDF it failed, but it succeeded if it was generated from a SDF. I have attached a notebook that shows this behaviour.
The error I get is: