2
\$\begingroup\$

This is a follow-up to https://stackoverflow.com/questions/71194832/converting-json-based-log-into-column-format-i-e-one-file-per-column .

My task is to optimize this code whose function is to convert the content of this JSON-formatted log file to columnar-formatted .txt file for each column. An example for a log file is:

{"timestamp": "2022-01-14T00:12:21.000", "Field1": 10, "Field_Doc": {"f1": 0}}
{"timestamp": "2022-01-18T00:15:51.000", "Field_Doc": {"f1": 0, "f2": 1.7, "f3": 2}}

It will generate 5 files:

  1. timestamp.column
  2. Field1.column
  3. Field_Doc.f1.column
  4. Field_Doc.f2.column
  5. Field_Doc.f3.column

The column file format is as follows:

  • string fields are separated by a new line '\n' character. Assume that no string value contains new line characters, so no need to worry about escaping them
  • double, integer & boolean fields are represented as a single value per line
  • null, undefined & empty strings are represented as an empty line

Example content of timestamp.column:

2022-01-14T00:12:21.000
2022-01-18T00:15:51.000

Note: The fields in the log will be dynamic, do not assume that these are the expected properties.

My current code for this is:

import json
import os
def flatten_dict(data, prefix=""):
    result = {}
    for key, value in data.items():
        if prefix:
           key = prefix + "." + key        
        if isinstance(value, dict):
            result.update( flatten_dict(value, key) )
        else:            
            if value is None:
                result[key] = "\n"           
            elif value is "":
                result[key] = "\n"
            else:
                result[key] = value
    return result

path = input("Enter the path for the log file: ")
# Checking if path exists to the selected log
assert os.path.exists(path), "I did not find the file at, "+str(path)
file_obj = open(path)   # emulate file in memory

for line in file_obj:
    data = json.loads(line) 
    data = flatten_dict(data)
    for key, value in data.items():
        with open(key + '.column', "a") as f:
            f.write(str(value) + "\n")

This code is currently very slow. It takes about 45 min just to parse 256mb of data which should have taken about 30sec or max a minute.

Can anyone guide me how can I optimize the efficiency? Also how can I print the usage of CPU and RAM for the same? File size to parse may vary up to 2 GB.

Any help would be appreciated.

\$\endgroup\$
5
  • 1
    \$\begingroup\$ Why are the columns separated into different files? This is a strange and inconvenient format. Do you control this? \$\endgroup\$
    – Reinderien
    Commented Mar 12, 2022 at 17:31
  • \$\begingroup\$ i don't control this, I'm just asked to do this \$\endgroup\$ Commented Mar 12, 2022 at 17:37
  • \$\begingroup\$ Is it work, homework, an interview problem, or a programming challenge? \$\endgroup\$
    – Reinderien
    Commented Mar 12, 2022 at 17:38
  • \$\begingroup\$ programming challenge \$\endgroup\$ Commented Mar 12, 2022 at 18:25
  • 2
    \$\begingroup\$ No, it's not: it's an interview question \$\endgroup\$
    – Reinderien
    Commented Mar 12, 2022 at 21:11

1 Answer 1

2
\$\begingroup\$

You're re-opening the column files every line. Open them once and then look the file handles up by key.

file_obj = open(path)   # emulate file in memory

for line in file_obj:
    data = json.loads(line) 
    data = flatten_dict(data)
    for key, value in data.items():
        with open(key + '.column', "a") as f:
            f.write(str(value) + "\n")

becomes ( not tested )

column_files = {}

with open(path) as jsonl_file:
    for line in jsonl_file:
        data = json.loads(line) 
        data = flatten_dict(data)
        for key, value in data.items():
            if key not in column_files:
                column_files[key] = open(key+'.column','w')

            column_files[key].write(str(value) + "\n")

for f in column_files: f.close()
\$\endgroup\$
1
  • 2
    \$\begingroup\$ Instead of using a dict of file handles, you can use a contextlib.ExitStack. Also the last line should be for f in column_files.values(): f.close(). \$\endgroup\$ Commented Mar 13, 2022 at 10:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.