bench

Benchmarks

python -m bench.benchmark            # default: 50,000 keys
python -m bench.benchmark --keys 200000 --commits 5000

The suite reports three things:

1. Throughput — bulk load, random point reads, and full scan for the B+ tree engine, with SQLite as a reference baseline. ImmuStore is pure Python, so it is naturally slower than SQLite's C engine; the goal is honest measurement and showing the operations scale, not claiming parity.

2. Why the balanced index matters — inserting sorted keys, the worst case for an unbalanced tree. The B+ tree stays shallow (height ~3 for thousands of keys); the naive binary tree degrades to a linked list (height == N) and gets thousands of times slower before it overflows the recursion stack entirely.

3. Durability cost — commit throughput with fsync on (full) vs off (none).

Numbers are machine- and Python-version-dependent; treat them as relative, not absolute. Representative results are quoted in the top-level README.