Template for writing Python Hive UDF Transform that supports multiprocessing

multiprocess_python_udf.py

from __future__ import print_function
import sys
from multiprocessing import Process, Pool, cpu_count

def transform(*args):
    # -- do something here -- 
    return []

def process_line(line):
    line = line.strip().split('\t')
    line = [None if x=='\N' else x for x in line]
    result = transform(*line)
    return '\t'.join(['\N' if x is None else str(x) for x in result])

pool = Pool(cpu_count())

for r in pool.map(process_line, sys.stdin):
    print(r)