loc gives incorrect values when given a Boolean indexer

[devin-petersohn]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): OSX
• Modin installed from (source or binary): source
• Modin version: 0.6.0.dev.0
• Python version: 3.6
• Exact command to reproduce: df.loc[df.col1 < 5]

Describe the problem

loc does not support Boolean indexers. This was originally reported over email and reproduced by me locally.

Source code / logs

import modin.pandas as pd
df = pd.DataFrame({"col1":[1,2,3,4,5,6], "col2":[1,2,3,4,5,6]})
df.loc[df.col1 < 5]

[ipacheco-uy]

Hi Devin, i check and loc does not support boolean and callable, the idea is to fallback or support it?

[devin-petersohn]

Thanks @ipacheco-uy, we won't fallback to support this because it would be really expensive.

I outlined a bit of what is needed in a post on the discourse: https://discuss.modin.org/t/indexing-update-rewrite/57

Indexing needs to be updated because some of the general implementation is too slow. Indexing should be fast and correct. We also have diverged codepaths for loc and __getitem__, which doesn't make a lot of sense because they do similar things. They should at least be the same in the backend.