Fresh 3-repeat run — April 14, 2026 — Dart 3.11.0, macOS ARM64
How wall time grows with row count. Lower is better.
C-side JSON serialization vs Dart jsonEncode. Lower is better.
How many times faster resqlite is at each row count.
How many times faster resqlite is at each row count.
Time spent on the main (UI) thread. sqlite3 blocks it entirely.
resqlite: zero-copy Isolate.exit → 0.00 ms on main thread.
Performance across narrow, wide, text-heavy, numeric, and nullable schemas.
Individual INSERT statements, each with its own transaction.
Batched writes at 100 / 1K / 10K rows.
executeBatch() vs execute()-loop inside a write transaction.
Reading rows inside an open write transaction.
Total wall time for N parallel queries. Lower is better.
Reusing prepared statements with different bind values.
Initial emission, invalidation latency, fan-out, and churn.
63 benchmarks compared. All within noise.
| Benchmark | resqlite | sqlite3 | sqlite_async | resqlite speedup vs best competitor |
|---|
All times are wall-clock medians in ms (lower is better) from repeat 3/3. Point query is queries/sec (higher is better). Comparison baseline: 2026-04-13. Dart 3.11.0 stable, macOS ARM64.