Filtering groups by condition

[bkamins]

In the past the following DataFrames.jl syntax was also supported in DataFramesMeta.jl:

filter(groupby(df, :id), x -> mean(x.col) > 10)

In general - maybe we should have @filter (instead of @rsubset) that is rowwise/pergroup, to match planned @rselect, @rcombine?

[pdeffebach]

I did not see this! I think gmail is filtering github notifications from this Repo.

No, I don't think we should have @filter instead of @rsubset. Since we can do @subset df @byrow ... we should also have @rsubset for symmetry and consistency with @rselect etc.

I think a @filter is a good idea, letting us filter by groups. If we allow @filter for a DataFrame it would definitely be redundant with @rsubset, but that isn't the end of the world. We have some redundancy with filter and subset(df, ... ByRow(...)...) as well.

[bkamins]

I am OK with having both @filter and @rsubset

[pdeffebach]

Do we need filter for 1.0?

[bkamins]

No, @filter is essential for filtering GroupedDataFrame groups only which is a quite rare use case I would say.
Also it can be added at any point later without breaking anything.

[pdeffebach]

This not a bad idea. Maybe @groupsubset is a better name, though.