Soup Benchmark¶

Apple-to-apple performance comparison between zerodep soup and beautifulsoup4.

Implementations¶

Data Sizes Tested¶

Parse + Query Performance (Mean)¶

Serialization Performance (Mean)¶

Tree Operations Performance (Mean)¶

CSS Select Performance (Mean)¶

Pseudo-Select Performance (Mean)¶

Key Takeaways¶

• 2.5-3.7x faster across all sizes and operations -- zerodep builds a minimal DOM tree directly from html.parser without the abstraction layers (NavigableString, PageElement hierarchy, soupsieve integration) that BeautifulSoup carries.
• Serialization shows the largest speedup -- 3.2-3.7x faster, as zerodep's lightweight node structure has less overhead during tree-to-string conversion.
• CSS select and pseudo-selectors are ~3x faster -- despite BS4 delegating to the optimized soupsieve library, zerodep's inline selector engine avoids the cross-library dispatch overhead.
• Consistent advantage across workloads -- parsing (2.7-2.8x), serialization (3.2-3.7x), tree operations (2.5-2.9x), and CSS selectors (2.8-3.1x) all show strong gains.
• Zero pip dependencies -- zerodep uses only re and html.parser from the standard library. BeautifulSoup requires soupsieve and optionally lxml or html5lib.

Run It Yourself¶

pip install pytest pytest-benchmark beautifulsoup4
pytest soup/test_soup_benchmark.py --benchmark-only -v

Latest CI Results¶

Updated automatically on each release via Benchmark CI.