hash-tput.py

"""
Purpose:
This python script is meant to introduce you to parellel processing with 
threads and processes. The example uses a simple worker than randomly selects 
a letter and continually hashes this value. This demonstrates two methods for 
creating a thread, (1) by specifying a target function or (2) inheriting 
threading.Thread. 

This lesson also shows that the Thread and Process API are (for the most part)
interchangable allowing us to simply replace threading.Thread with multiprocessing.Process
to switch from threads to processes. 

"""

__author__ = "William Koch"
__email__ = "wfkoch@bu.edu"


# STEP 1) Import libraries we will be using
# This library is for multi-threading
import threading
# While this is for multi processing, the API is almost identical
import multiprocessing

#Lets log stuff
import logging

# The remaining libraries will be used to create a simple worker
# This library is for creating hashes, https://docs.python.org/2/library/hashlib.html
import hashlib
# Get times
import time
# Random sampling
import random
# Helper for accessing letters
import string

# STEP 2) Init our logger
# More info: https://docs.python.org/2/howto/logging.html#logging-basic-tutorial
logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger("disscussion-wk8")

# There are two ways to init worker (1) by function (2) inherit threading.Thread


"""
Method 2 for creating thread inheriting threading.Thread,
to use multiple processors instead just replace threading.Thread with 
multiprocessing.Process
"""
class HashPerformance(threading.Thread):
    def __init__(self, hashfn, count):
        # First call the parents constructor
        self.hashfn = hashfn
        self.count = count
        self.message = random.choice(string.ascii_uppercase)
        super(HashPerformance, self).__init__()

    def run(self):
        # Refer to threading.Thread API, teh start method will call this
        # method internally in the Thread class and is what we override. We 
        # must match the method signature which does not allow any parameters
        # therefore we must modify our code to pass them through the constructor.
        start_time = time.time()
        for i in range(self.count):
            self.message = self.message.encode("utf-8")
            self.message = hashfn(self.message).hexdigest()
        lapse_time = time.time() - start_time

        # Compute throughput
        tput = self.count / lapse_time
        logger.info(" {} Tput = {} hashes/second".format(threading.current_thread().name, tput)) 


# STEP 3) Creat a worker
"""
An example of a worker function, see below for thread initialization
"""
def compute_hash_tput(count, hashfn):
    """ Do some work, repeatidy hash a message and compute tput"""

    # Just choose a random letter to hash so output isnt deterministic 
    message = random.choice(string.ascii_uppercase)
    start_time = time.time()
    for i in range(count):
        # We has some issues on certain computers with encoding, 
        # make sure in utf-8 required by hashlib if the system is unicode by default
        message = message.encode("utf-8")
        message = hashfn(message).hexdigest()
    lapse_time = time.time() - start_time

    # Compute throughput
    tput = count / lapse_time
    logger.info(" {} Tput = {} hashes/second".format(threading.current_thread().name, tput)) 


# STEP 4) Create main and initalize variables
if __name__ == "__main__":


    # Define the hash we will use, refer to hashlib API doc for options
    hashfn = hashlib.sha256
    # number of hashes 
    num_hashes = 100000
    num_workers = 10

    logger.info("Main thread name = {}".format(threading.current_thread().name))

    # STEP 5) Deomonstrate function is working
    #compute_hash_tput(100, hashlib.sha256)



    # Demonstrate multiple threads doing work
    # Create a bunch of threads
    workers = []
    for i in range(num_workers):
        # Method 2 for defining a thread, specify the worker as the target
        # Note args is based in a tuple!
        t = threading.Thread(target=compute_hash_tput, args=(num_hashes, hashlib.sha256,) )
        # start the execution of the thread
        t.start()
        # save the thread so we can join later
        workers.append(t)

    # Wait for each worker to finish and join back with the main thread
    for w in workers:
        w.join()

    # Uncomment to demonstrate Method 1
    # Now show by craeting classes
    """
    class_workers = []
    for i in range(num_workers):
        t = HashPerformance(hashfn, num_hashes)
        t.start()
        class_workers.append(t)

    for w in class_workers:
        w.join()
    """