Python, Pandas: Helpful function for creating dummy vars on a df using pd.get_dummies()

df_get_dummies.py

def df_get_dummies(df, drops, cats, pkey=None, binary=1):
    
    ### Import packages
    import pandas as pd
    import numpy as np
    ###
    
    # The following columns aren't used in modelling so drop
    df = df.drop(drops, 1)
    # Ensure BPE and LE are categories before getting dummies
    df[cats] = df[cats].apply(lambda x: int(x)).astype('category')
    if pkey=None:
    	pkey = df.columns[0]
    else:
    	pass
 	pkey_temp = df[pkey]
    print 'Total number of columns before dummifying is %d.' % len(df.columns)
    df = df.drop(pkey,1)
    df = pd.get_dummies(df)
    df = df.applymap(lambda x: int(x)).astype(np.int8)
    df[pkey] = pkey_temp
    # Choose to calculate IF predictor has occured for customer
    if binary == 1:
        df = df.groupby([pkey]).max().reset_index()
    # Choose to calculate HOW MANY predictor occurences for customer
    elif binary == 0:
        df = df.groupby([pkey]).sum().reset_index()
    # Choose to output unaggregated dataset
    elif binary == 2:
        pass
    
    print 'Total number of columns after dummifying is %d.' % len(df.columns)

    return df