DataFrame.loc does not work with list of indices

If i = [3, 5, 1] and df is a pandas data frame this returns the numbered rows in the defined order: df.loc[i].

This is the expected behavior for dask.dataframe.DataFrame but if df is such an object then df.loc[i] does not work. Note that you can give it a slice: df.loc[3:6] for example.

On a side note, although using numbers in the example above, loc is for labels and as such we could expect partitions made of indices (or multindices) of strings, datetimes, and other objects…

[abast]

What about this one for larger lists?

ddf.map_partitions(lambda x: x[x.index.isin(L)])

[jdthorpe]

I’m a little late to the party, but as of now, this is is an effective solution:

ddf.loc[ddf.index.isin(L)].compute()