benchmark.py

import pandas as pd
import numpy as np
import time
import fast_parser
import matplotlib.pyplot as plt

FILENAME = "market_data.bin"
NUM_RECORDS = 40_000_000

def run_benchmark():
    p = fast_parser.FastParser(FILENAME)
    print("--- Starting Benchmark ---")

    # Normal Pandas/Numpy Method
    start_pd = time.time()
    data = np.fromfile(FILENAME, dtype=[('ts', 'u8'), ('sym', 'S8'), ('price', 'f8')])
    df = pd.DataFrame(data)
    pd_mean = df['price'].mean()
    pd_time = time.time() - start_pd
    print(f"Pandas/Numpy Time: {pd_time:.4f}s | Mean: {pd_mean:.2f}")

    # Zero-Copy Method
    start_zc = time.time()
    prices = p.get_prices()
    zc_mean = np.mean(prices)
    zc_time = time.time() - start_zc
    print(f"Zero-Copy Time:   {zc_time:.4f}s | Mean: {zc_mean:.2f}")

    speedup = pd_time / zc_time
    print(f"\n Your parser is {speedup:.1f}x faster!")

    # Visualization
    labels = ['Pandas / NumPy', 'Zero-Copy Parser']
    actual_times = [pd_time, zc_time]
    actual_throughput = [(NUM_RECORDS / t) / 1e6 for t in actual_times]

    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(15, 6))
    colors = ['#E74C3C', '#3498DB'] 

    
    bars1 = ax1.bar(labels, actual_times, color=colors, width=0.6)
    ax1.set_title('Execution Latency\n(Lower is Better)', fontsize=14, pad=15)
    ax1.set_ylabel('Time (Seconds)', fontsize=12)
    ax1.grid(axis='y', linestyle='--', alpha=0.6)
    for bar in bars1:
        height = bar.get_height()
        ax1.text(bar.get_x() + bar.get_width()/2., height, f'{height:.4f}s', 
                 ha='center', va='bottom', fontsize=11, fontweight='bold')

   
    bars2 = ax2.bar(labels, actual_throughput, color=colors, width=0.6)
    ax2.set_title('Data Throughput\n(Higher is Better)', fontsize=14, pad=15)
    ax2.set_ylabel('Millions of Records / Second', fontsize=12)
    ax2.grid(axis='y', linestyle='--', alpha=0.6)
    for bar in bars2:
        height = bar.get_height()
        ax2.text(bar.get_x() + bar.get_width()/2., height, f'{height:.1f}M', 
                 ha='center', va='bottom', fontsize=11, fontweight='bold')

    plt.suptitle(f'Project Performance: {speedup:.1f}x Speedup\nDataset: 40 Million Records ({40*24/1e6:.1f} MB)', 
                 fontsize=16, fontweight='bold', y=1.08)

    plt.tight_layout()
    fig.savefig('benchmark_results.png', dpi=300, bbox_inches='tight')
    print("\n Benchmark image saved as 'benchmark_results.png' in the sidebar!")

if __name__ == "__main__":
    run_benchmark()