Skip to content
Advertisement

Create a separate logger for each process when using concurrent.futures.ProcessPoolExecutor in Python

I am cleaning up a massive CSV data dump. I was able to split the single large file into smaller ones using gawk initially using a unix SE Query as a following flow:

     BIG CSV file -> use gawk script + bash -> Small CSV files based on columns

I have about 12 split csv files that are created using the above mentioned flow and each with ~170K lines in them.

I am using python3.7.7 on a Windows 10 machine.

Code

def convert_raw_data(incoming_line, f_name, line_counter):
     # do some decoding magic
     # catch exception and try to log it into the a logger file under `f_name.log`


def convert_files(dir_name, f_name, dest_dir_name):
    # Open the CSV file
    # Open the Destination CSV file to store decoded data
    line_counter = 1
    for line in csv_reader:
       # convert raw HEX to Floating point values using `convert_raw_data` function call
       line_counter = line_counter + 1
       status = convert_raw_data(csv)
       if status:
          return f'All good for {f_name}.'
       else:
          return f'Failed for {f_name}'

def main():

    # Parse Arguments Logic here

    # get CSV Files and their respective paths
    csv_files = get_data_files_list(args.datasets)

    # decode raw data from each split csv file as an individual process
    with concurrent.futures.ProcessPoolExecutor() as executor:
        results = [ executor.submit(convert_files, dir_name, f_name, dest_dir) for dir_name, f_name in csv_files ]

        for f in concurrent.futures.as_completed(results):
            print(f.result())

Requirements

I wish to set a logging logger with the name f_name.log within each process spawned by the ProcessPoolExecutor and want to store the logs with the respective parsed file name. I am not sure if I should use something like:

def convert_raw_data(...., logger):
    logger.exception(raw_data_here)


def convert_files(....):
    logger = logging.basicConfig(filename=f_name, level=logging.EXCEPTION)

or are there caveats for using logging modules in a multiprocessing environment?

Advertisement

Answer

Found out a simple way to achieve this task:

import logging

def create_log_handler(fname):
    logger = logging.getLogger(name=fname)
    logger.setLevel(logging.ERROR)

    fileHandler = logging.FileHandler(fname + ".log")
    fileHandler.setLevel(logging.ERROR)

    logger.addHandler(fileHandler)

    formatter = logging.Formatter('%(name)s %(levelname)s: %(message)s')

    fileHandler.setFormatter(formatter)

    return logger

I called the create_log_handler within my convert_files(.....) function and then used logger.info and logger.error` accordingly.

by passing the logger as a parameter to convert_raw_data I was able to log even the erroneous data point in each of my csv file on each process.

User contributions licensed under: CC BY-SA
3 People found this is helpful
Advertisement