Environment: SQLite (aiosqlite), Python 3.10, Linux. Operations: 1000 items.
Results
| Operation | Implementation | Time (s) | Ops/sec | Notes |
|---|---|---|---|---|
| Insert | FluxCRUD (Batcher) | 0.12s | ~8,000 | Uses batch_writer (Chunked Inserts) |
| Raw SQLAlchemy (Bulk) | 0.21s | ~4,600 | session.add_all + commit | |
| FluxCRUD (Simple) | 0.76s | ~1,300 | Sequential await repo.create() | |
| FastCRUD | 0.80s | ~1,250 | Sequential await crud.create() | |
| Read | Raw SQLAlchemy | 0.27s | ~3,700 | Baseline |
| FluxCRUD (Direct) | 0.32s | ~3,100 | Fastest ORM-wrapper (Default) | |
| FastCRUD | 0.43s | ~2,300 | ||
| FluxCRUD (Loader) | 0.50s | ~2,000 | N+1 Safe (Opt-in) |
Analysis
1. High Performance Writes
FluxCRUD’s Batcher (repo.batch_writer) significantly outperforms sequential inserts (~4x faster) and rivals raw bulk inserts.
2. Read Performance
FluxCRUD is now Fast by Default. UseRepository(..., use_loader=False) (default) for maximum speed.
- Direct Read: By bypassing the DataLoader and inlining operations, FluxCRUD is ~35% faster than FastCRUD.
- Safety Option: You can opt-in to
use_loader=Truefor complex graph queries. This adds overhead (~50% slower than direct) but prevents N+1 issues automatically. - Low Overhead: FluxCRUD adds minimal overhead (~0.05s per 1000 items) over raw SQLAlchemy.
3. Developer Experience
- FluxCRUD provides the
Batcherout-of-the-box for high-throughput jobs. - FastCRUD and FluxCRUD have similar performance for basic sequential operations, but FluxCRUD offers more robust tools (Caching, Batching) for scaling.