7

I have a situation where I need to index a dask dataframe by location. I see that there is not an .iloc method available. Is there an alternative? Or am I required to use label-based indexing?

For example, I would like to 

import dask.dataframe as dd
import numpy as np
import pandas as pd
df = dd.from_pandas(pd.DataFrame({k:np.random.random(10) for k in ['a', 'b']}), npartitions=2)
inds = [1, 4, 6, 8]
df.iloc[inds]

Is this not possible with dask? (e.g., Perhaps a positional location is not well-defined?) In this case, what can I do if I only know the positional indices (not labels) of the values I need to access?

Tim Morton
  • 240
  • 1
  • 3
  • 11

1 Answers1

6

Positional indexing is not available for dask dataframe, nor is it likely to be available in the near future.

MRocklin
  • 55,641
  • 23
  • 163
  • 235