from multiprocessing import cpu_count
from multiprocessing import  Pool
from dask import dataframe as dd
from dask.multiprocessing import get



def parallelize_df_transform(df, transform_f, n_partitions=cpu_count()):
    df_split = np.array_split(df, n_partitions)
    pool = Pool(n_partitions)
    df = pd.concat(pool.map(transform_f, df_split))
    pool.close()
    pool.join()
    return df

def featurise(df):
    df['username_email_sim'] = df.apply(lambda x : some_func(x), axis=1)
    return df

a_data_frame = parallelize_df_transform(a_data_frame, featurise)