pandas vs polars for reading csv files

benchmark.py

# /// script
# requires-python = ">=3.12,<3.13"
# dependencies = [
#     "pandas",
#     "polars",
# ]
# ///

# note: the header of this script was generated by the command
# uv add --script benchmark.py pandas polars

import csv
import random
import time
import pandas as pd
import polars as pl

# Constants
NUM_ROWS = 10_000_000
OUTPUT_FILE = "random_coordinates.csv"

# Step 1: Generate the CSV file
print(f"Generating {NUM_ROWS} random coordinates and writing to {OUTPUT_FILE}...")

start_time = time.time()
with open(OUTPUT_FILE, mode="w", newline="") as file:
    writer = csv.writer(file)
    writer.writerow(["x", "y"])  # Header row

    for _ in range(NUM_ROWS):
        x = random.uniform(-1_000, 1_000)  # Random float between -1000 and 1000
        y = random.uniform(-1_000, 1_000)
        writer.writerow([x, y])
end_time = time.time()

csv_generation_time = end_time - start_time
print(f"CSV file {OUTPUT_FILE} created successfully in {csv_generation_time:.2f} seconds.\n")

# Step 2: Benchmark pandas read_csv
print("Benchmarking pandas read_csv...")

start_time = time.time()
data = pd.read_csv(OUTPUT_FILE)
end_time = time.time()

pandas_time = end_time - start_time
print(f"Pandas read_csv completed in {pandas_time:.2f} seconds.\n")

# Step 3: Benchmark Polars read_csv
print("Benchmarking Polars read_csv...")

start_time = time.time()
data_polars = pl.read_csv(OUTPUT_FILE)
end_time = time.time()

polars_time = end_time - start_time
print(f"Polars read_csv completed in {polars_time:.2f} seconds.\n")

# Final summary
print("----- Benchmark Summary -----")
print(f"CSV generation time: {csv_generation_time:.2f} seconds")
print(f"Pandas read_csv time: {pandas_time:.2f} seconds")
print(f"Polars read_csv time: {polars_time:.2f} seconds")
print("------------------------------")