Question

To set the context I have a directory with 200-300 files, each file ranges in size (# of lines). I pase the files and export them to a csv file. I think the last time I ran it the csv file had over 340,000 rows. On top of that the first 8 files are constantly being written to so I lose data while parsing sometimes.

Now, each file is set up like this:

DateTime Message Action ActionDetails

I have code in place to take go through all the files, parse them and then output to a csv file:

for infile in listing:
    _path2 = _path + infile
    f = open(_path2, 'r')
    labels = ['date', 'message', 'action', 'details']
    reader = csv.DictReader(f, labels, delimiter=' ', restkey='rest')

    for line in reader:
        if line.get('rest'):
            line['details'] += ' %s' % (' '.join(line['rest']))
        out_file.write(','.join([infile,line['date'], line['message'], line['action'], line['details']]) + '\n')

    f.close()
out_file.close()

I was wondering what the “best” way to go about copying the first 8 files so I don’t lose data while parsing would be. By best I mean take the least amount of time as the total time to run the python script at the moment is about 35-45 seconds.

Answer 1

I got a little bored. Try this on for size. I didn’t actually have a chance to check if it was parsing and writting correctly but other than that I believe it should run given some info. This problem is a good opportunity to use queueing. Let me know how fast it runs!

from threading import Thread
import Queue
import os
import time
import sys

# declare some global items
# queue that an author thread can write line items to a csv
write_q = Queue.Queue()

# queue filled with files to parse 
read_q = Queue.Queue()

# queue filled with files that have size change during read. Can
# preload this queue to optimize however program should handle any
# file that changes during operation
moving_q = Queue.Queue()

# given csv labels
labels = ['date', 'message', 'action', 'details']

# global for writer thread so it knows when to close
files_to_parse = True

# parsing function for any number of threads
def file_parser():    
    # Each parser thread will run until the read_q is empty
    while True:
        moving = False
        # Test for a file from the read queue or moving queue 
        try:
            if not moving_q.empty():
                try:
                    f_path = moving_q.get(False)
                    moving = True
                # if the moving queue is empty after trying to read
                # might have been snatched by different thread. Ignore error
                except Queue.Empty:
                    pass
            else:
                # No items left in moving queue so grab non moving file
                f_path = read_q.get(False)
        # all files have been dealt with
        except Queue.Empty:
            print "Done Parsing"
            sys.exit()

        # Following will parse a file and test that the file is not being
        # modified during the read
        with open(f_path, 'r') as f:
            # csv reader setup
            reader = csv.DictReader(f, labels, delimiter=' ', restkey='rest')

            # initillized file size (when we started reading)
            pre = os.path.getsize(f_path)

            # store output items in a list so if file is updated during read
            # we can just ignore those items and read file later
            line_items = []

            # parse the file line by line
            for line in reader:
                # Check that file hasn't been updated
                post = os.path.getsize(f_path)
                if pre != post:
                    # if file has changed put the file back on the queue and clear the output lines
                    moving_q.put(f_path)
                    line_items = None
                    break
                # parse the line and add it to output list
                else:
                    if line.get('rest'):
                        line['details'] += ' %s' % (' '.join(line['rest']))
                        line_items.append(','.join([infile,line['date'], line['message'], line['action'], line['details']]) + '\n')

            # don't want to do reading and writing in same thread. Push
            # all line items onto the write thread for the author to deal with    
            if line_items and moving:
                write_q.put(line_items)
                moving_q.task_done()
            elif line_items and not moving:
                write_q.put(line_items)
                read_q.task_done()

# author thread that will write items to a file as other threads complete
# tasks. Should help speed up IO bound processing
def file_author(out_file):
    with open(out_file,'w') as f:
        # parse files until all the parser threads are running
        while files_to_parse or not read_q.empty():
            # only one writer thread so write as items are put into thread
            if not read_q.empty():
                line_items = write_q.get(False)
                for line_item in line_items:
                    f.write(line_item)
                write_q.task_done()
            # sleep in the downtime so we dont overload PC
            else:
                time.sleep(.1)
    print "Done writting"


if __name__ == "__main__":
    # list of file names as you had before
    listing = []
    outfile = "MyNewCSVfile.csv"

    # You can optimize parsing by adding known "moving files" directly
    # to the moving_queue, however program should handle either way
    for infile in listing:
        _path2 = _path + infile
        write_q.put(_path2)

    # make a writer thread
    t = Thread(target = file_author, args = (outfile,))
    t.daemon = True
    t.start()

    # make some parse threads
    for i in range(10):
        t = Thread(target = file_parser)
        t.daemon = True
        t.start()

    # wait for parser threads to finish work
    read_q.join()
    moving_q.join()

    # close author
    files_to_parse = False
    time.sleep(.1)
    print "Complete"