Turkish
durak / benchmarks /benchmark_rust_vs_python.py
Taklaxbr's picture
Archive open-source github repository (cdliai/durak)
7c16940 verified
Raw History Blame Contribute Delete
3.53 kB
"""Benchmark comparing Rust vs Python implementations.
Measures performance differences between pure Python implementations
and Rust-accelerated functions.
"""
import time
def benchmark(func, *args, iterations=10000):
"""Run a benchmark and return average execution time."""
start = time.perf_counter()
for _ in range(iterations):
func(*args)
end = time.perf_counter()
return (end - start) / iterations * 1000 # Return ms per call
def main():
import durak
print("=" * 70)
print("Rust vs Python Performance Benchmark")
print("=" * 70)
# Test text
test_text = "İstanbul'da Merhaba Dünya! Bu bir TEST cümlesidir."
large_text = test_text * 100
# 1. Normalization Benchmark
print("\n1. Text Normalization")
print("-" * 70)
# Python version
def python_normalize(text):
return durak.normalize_case(text)
# Rust version
try:
from durak import _durak_core
def rust_normalize(text):
return _durak_core.fast_normalize(text)
py_time = benchmark(python_normalize, test_text)
rust_time = benchmark(rust_normalize, test_text)
print(f"Python normalize: {py_time:.4f} ms per call")
print(f"Rust normalize: {rust_time:.4f} ms per call")
print(f"Speedup: {py_time / rust_time:.2f}x")
except ImportError:
print("Rust extension not available. Run: maturin develop")
# 2. Tokenization Benchmark
print("\n2. Tokenization with Offsets")
print("-" * 70)
# Python version
def python_tokenize(text):
return durak.tokenize_with_offsets(text)
# Rust version
try:
from durak import _durak_core
def rust_tokenize(text):
return _durak_core.tokenize_with_offsets(text)
py_time = benchmark(python_tokenize, large_text, iterations=1000)
rust_time = benchmark(rust_tokenize, large_text, iterations=1000)
print(f"Python tokenize: {py_time:.4f} ms per call")
print(f"Rust tokenize: {rust_time:.4f} ms per call")
print(f"Speedup: {py_time / rust_time:.2f}x")
except ImportError:
print("Rust extension not available")
# 3. Resource Loading Benchmark
print("\n3. Resource Loading")
print("-" * 70)
try:
from durak import _durak_core
# File-based loading
def load_from_file():
return durak.load_stopword_resource("base/turkish")
# Embedded Rust loading
def load_from_rust():
return _durak_core.get_stopwords_base()
file_time = benchmark(load_from_file, iterations=100)
rust_time = benchmark(load_from_rust, iterations=100)
print(f"File-based load: {file_time:.4f} ms per call")
print(f"Embedded Rust load: {rust_time:.4f} ms per call")
print(f"Speedup: {file_time / rust_time:.2f}x")
except ImportError:
print("Rust extension not available")
# 4. Full Pipeline Benchmark
print("\n4. Complete Processing Pipeline")
print("-" * 70)
# Keep data types compatible across steps: normalize keeps string,
# tokenize returns list.
pipeline = durak.Pipeline(["clean", "normalize", "tokenize", "remove_stopwords"])
pipeline_time = benchmark(pipeline, large_text, iterations=100)
print(f"Full pipeline: {pipeline_time:.4f} ms per call")
print("\n" + "=" * 70)
print("Benchmark complete!")
print("=" * 70)
if __name__ == "__main__":
main()