pandas: slice Multindex with many indices

Question

I have a dataframe d with about 100,000,000 rows and 3 columns. It looks something like this:

import pandas as pd 

In [17]: d = pd.DataFrame({'id': ['a', 'b', 'c', 'd', 'e'], 'val': [1, 2, 3, 4, 5], 'n': [34, 22, 95, 86, 44]}) 

In [18]: d.set_index(['id', 'val'], inplace = True)

I have another dataframe with values of id and val that I want to keep in d. There are around 600,000 combinations of id and val that I want to keep:

In [20]: keep = pd.DataFrame({'id':['a', 'b'], 'val' : [1, 2]})

I have tried this in the following way:

In [21]: keep.set_index(['id', 'val'], inplace = True)

In [22]: d.loc[d.index.isin(keep.index), :] 
Out [22]:         
                   n
         id val    
          a  1    34
          b  2    22

This works but seems clunky and is very slow. Is there a better approach here? What is the fastest way to slice on Multindex in pandas?

Answer 1

Using reindex

d.reindex(pd.MultiIndex.from_frame(keep))
Out[151]: 
         n
id val    
a  1    34
b  2    22