MultiprocessPandas

MultiprocessPandas package extends functionality of Pandas to easily run operations on multiple cores i.e. parallelize the operations. The current version of the package provides capability to parallelize apply() methods on DataFrames, Series and DataFrameGroupBy .

Importing the applyparallel module will add apply_parallel() method to DataFrame, Series and DataFrameGroupBy, which will allow you to run operation on multiple cores.

Installation

The source code is currently hosted on GitHub at: https://github.com/akhtarshahnawaz/multiprocesspandas. The package can be pulled from GitHub or can be installed from PyPi directly.

To install using pip

pip install multiprocesspandas

Setting up the Library

To use the library, you have to import applyparallel module. Import will attach required methods to pandas, and you can call them directly on Pandas data objects.

from multiprocesspandas import applyparallel

Usage

Once imported, the library adds functionality to call apply_parallel() method on your DataFrame, Series or DataFrameGroupBy . The methods accepts a function that has to be applied, and two named arguments:

static_data (External Data required by passed function, defaults to None)
num_processes (Defaults to maximum available cores on your CPU)
axis (Only for DataFrames, defaults to 0 i.e. rows. For columns, set axis=1.

Note: Any extra module required by the passed function must be re-imported again inside the function.

Usage with DataFrameGroupBy

def func(x):
    import pandas as pd
    return pd.Series([x['C'].mean()])

df.groupby(["A","B"]).apply_parallel(func, num_processes=30)

If you need some external data inside func(), it has to be passed and received as named argument static_data. If there is more that one external data that is required, then static_data can be a list of all required data, and can be accessed inside func by indexing.

data1 = pd.Series([1,2,3])
data2 = 20

def func(x, static_data):
    import pandas as pd
    output = static_data[0] - x['C'].mean()
    return output * static_data[1]

df.groupby(["A","B"]).apply_parallel(func, num_processes=30, static_data=[data1, data2])

Usage with DataFrame

Usage with DataFrames is very similar to the one with DataFrameGroupBy, however you have to pass an extra argument 'axis' which tells whether to apply function on the rows or the columns.

def func(x):
    return x.mean()

df.apply_parallel(func, num_processes=30, axis=1)

External data can be passed in same way as we did in DataFrameGroupBy

data = pd.Series([1,2,3])

def func(x, static_data):
    return static_data.sum() + x.mean()

df.apply_parallel(func, num_processes=30, static_data=data)

Usage with Series

Usage with Series is very similar to the usage with DataFrames and DataFrameGroupBy.

data = pd.Series([1,2,3])

def func(x, static_data):
    return static_data-x

series.apply_parallel(func, num_processes=30, static_data=data)