How to merge two pandas dataframe in parallel (multithreading or multiprocessing)

Question

Without doing in parallel programming I can merger left and right dataframe on key column using below code, but it will be too slow since both are very large. is there any way I can do it in parallelize efficiently ?

I have 64 cores, and so practically I can use 63 of them to merge these two dataframe.

left = pd.DataFrame({'key': ['K0', 'K1', 'K2', 'K3'],
                      'A': ['A0', 'A1', 'A2', 'A3'],
                     'B': ['B0', 'B1', 'B2', 'B3']})


right = pd.DataFrame({'key': ['K0', 'K1', 'K2', 'K3'],
                      'C': ['C0', 'C1', 'C2', 'C3'],
                      'D': ['D0', 'D1', 'D2', 'D3']})


result = pd.merge(left, right, on='key')

output will be :

left:
    A   B key
0  A0  B0  K0
1  A1  B1  K1
2  A2  B2  K2
3  A3  B3  K3

right:
    C   D key
0  C0  D0  K0
1  C1  D1  K1
2  C2  D2  K2
3  C3  D3  K3

result:
    A   B key   C   D
0  A0  B0  K0  C0  D0
1  A1  B1  K1  C1  D1
2  A2  B2  K2  C2  D2
3  A3  B3  K3  C3  D3

I want to do this in parallel so I can do it at speed.

Answer 1

I believe you can use dask. and function merge.

Docs say:

What definitely works?

Cleverly parallelizable operations (also fast):

Join on index: dd.merge(df1, df2, left_index=True, right_index=True)

Or:

Operations requiring a shuffle (slow-ish, unless on index)

Set index: df.set_index(df.x)

Join not on the index: pd.merge(df1, df2, on='name')

You can also check how Create Dask DataFrames.

Example

import pandas as pd

left = pd.DataFrame({'key': ['K0', 'K1', 'K2', 'K3'],
                      'A': ['A0', 'A1', 'A2', 'A3'],
                     'B': ['B0', 'B1', 'B2', 'B3']})


right = pd.DataFrame({'key': ['K0', 'K1', 'K2', 'K3'],
                      'C': ['C0', 'C1', 'C2', 'C3'],
                      'D': ['D0', 'D1', 'D2', 'D3']})


result = pd.merge(left, right, on='key')
print result
    A   B key   C   D
0  A0  B0  K0  C0  D0
1  A1  B1  K1  C1  D1
2  A2  B2  K2  C2  D2
3  A3  B3  K3  C3  D3

import dask.dataframe as dd

#Construct a dask objects from a pandas objects
left1 = dd.from_pandas(left, npartitions=3)
right1 = dd.from_pandas(right, npartitions=3)

#merge on key
print dd.merge(left1, right1, on='key').compute()
    A   B key   C   D
0  A3  B3  K3  C3  D3
1  A1  B1  K1  C1  D1
0  A2  B2  K2  C2  D2
1  A0  B0  K0  C0  D0

#first set indexes and then merge by them
print dd.merge(left1.set_index('key').compute(), 
               right1.set_index('key').compute(), 
               left_index=True, 
               right_index=True)
      A   B   C   D
key                
K0   A0  B0  C0  D0
K1   A1  B1  C1  D1
K2   A2  B2  C2  D2
K3   A3  B3  C3  D3

Answer 2

You can improve the speed (by a factor of about 3 on the given example) of your merge by making the key column the index of your dataframes and using join instead.

left2 = left.set_index('key')
right2 = right.set_index('key')

In [46]: %timeit result2 = left2.join(right2)
1000 loops, best of 3: 361 µs per loop

In [47]: %timeit result = pd.merge(left, right, on='key')
1000 loops, best of 3: 1.01 ms per loop