How to process a DataFrame with millions of records using Python Pandas?

Question

I have a csv file containing around 8 Million records, but it is taking more than an hour to complete the process, so please could you please help me with this?

Note: There is no issue with the python code; it works very well without any errors. The only problem is that is taking too much time to load and process the 8M records.

Here is the code

import pandas as pd
import numpy as np
import ipaddress
from pathlib import Path
import shutil
import os
from time import time
start = time()

inc_path = 'C:/Users/phani/OneDrive/Desktop/pandas/inc'
arc_path = 'C:/Users/phani/OneDrive/Desktop/pandas/arc'
dropZone_path = 'C:/Users/phani/OneDrive/Desktop/pandas/dropZone'

for src_file in Path(dropZone_path).glob('XYZ*.csv*'):
  process_file = shutil.copy(os.path.join(dropZone_path, src_file), arc_path)

for sem_file in Path(dropZone_path).glob('XYZ*.sem'):
  semaphore_file = shutil.copy(os.path.join(dropZone_path, sem_file), inc_path)

 # rename the original file
 for file in os.listdir(dropZone_path):
file_path = os.path.join(dropZone_path, file)
shutil.copy(file_path, os.path.join(arc_path, "Original_" + file))

 for sema_file in 
   Path(arc_path).glob('Original_XYZ*.sem*'):
   os.remove(sema_file)

  ## Read CSVfile from TEMP folder
  df = pd.read_csv(process_file)
  df.sort_values(["START_IP_ADDRESS"], ascending=True,)

  i = 0
  while i < len(df) - 1:
     i += 1
    line = df.iloc[i:i + 1].copy(deep=True)
curr_START_IP_NUMBER = line.START_IP_NUMBER.values[0]
curr_END_IP_NUMBER = line.END_IP_NUMBER
prev_START_IP_NUMBER = df.loc[i - 1, 'START_IP_NUMBER']
prev_END_IP_NUMBER = df.loc[i - 1, 'END_IP_NUMBER']
# if no gap - continue
if curr_START_IP_NUMBER == (prev_END_IP_NUMBER + 1):
    continue
# else fill the gap
# new line start ip number
line.START_IP_NUMBER = prev_END_IP_NUMBER + 1
line.START_IP_ADDRESS = (ipaddress.ip_address(int(line.START_IP_NUMBER)))
# new line end ip number
line.END_IP_NUMBER = curr_START_IP_NUMBER - 1
line.END_IP_ADDRESS = (ipaddress.ip_address(int(line.END_IP_NUMBER)))
line.COUNTRY_CODE = ''
line.LATITUDE_COORDINATE = ''
line.LONGITUDE_COORDINATE = ''
line.ISP_NAME = ''
line.AREA_CODE = ''
line.CITY_NAME = ''
line.METRO_CODE = ''
line.ORGANIZATION_NAME = ''
line.ZIP_CODE = ''
line.REGION_CODE = ''
# insert the line between curr index to previous index
df = pd.concat([df.iloc[:i], line, df.iloc[i:]]).reset_index(drop=True)
df.to_csv(process_file, index=False)
for process_file in Path(arc_path).glob('XYZ*.csv*'):
   EREFile_CSV = shutil.copy(os.path.join(arc_path, process_file), inc_path)

Asked By: Phani Kumar

||

Source

Answer 1

if the csv file doesn’t change very often, and you always need to repeat the analysis perhaps during dev stage, the best way is to save it as a pickle .pkl file.

Answered By: Ji Wei

Answer 2

You can either read the .csv file in chunks using the Pandas library, and then process each chunk separately, or concat all chunks in a single dataframe (if you have enough RAM to accommodate all the data):

#read data in chunks of 1 million rows at a time
chunks = pd.read_csv(process_file, chunksize=1000000)

# Process each chunk
for chunk in chunks:
    # Process the chunk
    print(len(chunk))
    
# or concat the chunks in a single dataframe
#pd_df = pd.concat(chunks)

Alternatively, you can use the Dask library, which can handle large datasets by internally chunking the dataframe and processing it in parallel:

from dask import dataframe as dd
dask_df = dd.read_csv(process_file)

Answered By: Chris

How to process a DataFrame with millions of records using Python Pandas?

Question:

Answers: