February 19, 2026
Benchmarks, brawls, and browser builds
Stoolap/Node: A Native Node.js Driver That's Surprisingly Fast
Rust DB crashes into Node; fans hype, skeptics call foul
TLDR: A Rust database, Stoolap, launched a native Node.js driver and beat SQLite in 47 of 53 tests, especially on complex queries. The community split: some love the speed and features, others dismiss micro-bench wins, question “no overhead” claims, and poke at AI-written vibes—while curious devs rush to the browser demo.
A Rust-built database called Stoolap just dropped a native Node.js driver, promising direct, super-fast access from JavaScript via NAPI-RS. The dev’s scoreboard screams 47 wins out of 53 against the popular SQLite driver, with eye-popping numbers like 138x faster on “counting unique things” and huge leads on complex queries. The crowd went wild… and then immediately started arguing.
One side is cheering the “no middleman” setup and big-boy features like MVCC (that’s a way to let readers and writers play nice without locks), smarter query planning, and even time-travel queries with “AS OF.” The other side rolled its eyes at micro-benchmarks: does anyone care about 0.001 milliseconds vs 0.01? Some questioned the “no serialization overhead” claim, poking at whether native bindings are truly friction-free. Meanwhile, a spicy subplot: a commenter accused the blog of being written by an AI, joked about the project’s “Go → Rust → Node speedrun,” and asked if this is really “native” or just binding to Rust.
Amid the drama, curious testers asked for browser support—then found a live playground and said, “I’ll try it.” The vibe: big performance flex, bigger skepticism, and plenty of memes about “new kid dunks on SQLite” vs “show me real-world workloads”
Key Points
- •@stoolap/node is a native Node.js driver for Stoolap, built with NAPI-RS to provide direct bindings without HTTP or serialization overhead.
- •Stoolap (Rust-based embedded SQL) features MVCC, a cost-based optimizer, parallel execution, semantic query caching, and AS OF temporal queries.
- •A benchmark of 53 identical SQL operations (in-memory, same data/machine) compared @stoolap/node with better-sqlite3 (SQLite).
- •Stoolap won 47 of 53 tests, with large speedups on complex operations like COUNT DISTINCT (138x) and subqueries (up to 57x).
- •SQLite was slightly faster in simple single-row operations (e.g., select/update by ID, inserts), typically by 1.0x–1.6x.