database-publishing.md

Benchmark Database Publishing

The benchmark system now supports publishing results directly to the SQLite tracking database. This allows you to:

1. Track performance over time - See how your optimizations affect performance
2. Compare with previous runs - Leverage the existing performance comparison system
3. Historical analysis - Use the same data for statistics and history commands

Usage

Main CLI

# Benchmark and publish to database
python main.py 2025 1 --benchmark --benchmark-publish

# Benchmark specific part and publish
python main.py 2025 1 --benchmark --benchmark-publish --part 1

# Benchmark entire year and publish
python main.py --benchmark-year 2025 --benchmark-publish

# Benchmark everything and publish (careful!)
python main.py --benchmark-all --benchmark-publish

Quick Benchmark Tool

# Fast benchmark with database publishing
python benchmark_quick.py fast 2025 1 --publish

# Normal benchmark with database publishing
python benchmark_quick.py normal 2025 1 --publish

# Thorough benchmark with database publishing
python benchmark_quick.py thorough 2025 1 --publish

# Year benchmarks with database publishing
python benchmark_quick.py fast-year 2025 --publish

Database Integration

When --benchmark-publish or --publish is used:

• Each benchmark run is stored as a separate record in the runs table
• Results include execution time, success/failure status, and error messages
• Input and code hashes are stored for change tracking
• is_sample is set to False (benchmarks use real input data)
• Results can be viewed using existing history and stats commands

Notes

• Database publishing requires tracking to be enabled (don't use --no-tracking)
• Only successful benchmark runs contribute to statistics
• Warmup runs are not published to the database
• Each individual run within a benchmark session is recorded separately

Example Workflow

# Run baseline benchmark and publish
python main.py 2025 1 --benchmark --benchmark-publish --benchmark-runs 10

# Make code optimizations
# ...

# Run new benchmark and compare
python main.py 2025 1 --benchmark --benchmark-publish --benchmark-runs 10

# View performance history
python main.py 2025 1 --history

# Check statistics
python main.py --stats