February 24, 2026
SQL diet or snake oil?
Show HN: SNKV – SQLite's B-tree as a key-value store (C/C++ and Python bindings)
SQLite goes on an SQL diet; devs debate “faster” key‑value magic
TLDR: SNKV turns SQLite into a lean key‑value store by skipping SQL and claims big speed gains. The community is split: hype says it could replace Redis, while skeptics question benchmarks and warn random access may be slower, making this a flashy idea with a reality‑check brawl.
SQLite just hit the gym. SNKV strips out the whole SQL layer and talks straight to SQLite’s storage engine—the B‑tree—turning it into a simple key‑value box. The pitch: keep SQLite’s rock‑solid safety (think crash protection and reliable transactions) while being ~50% faster for mixed reads/writes/deletes. In plain speak: fewer middlemen, more speed. That’s the claim, anyway.
Cue the Hacker News drama. One camp is hyped—m00dy joked this means “you don’t need Redis anymore,” turning a beloved server cache into yesterday’s snack. Another camp is throwing flags. Retr0id says they expected SQL overhead to be tiny and demands receipts: “Where does the +104% on random deletes come from?” Meanwhile rurban’s reality check lands hard: a plain hashtable still wins on random lookups, while ordered reads and range scans are where SNKV shines. Then the spice level hits max with d1l calling it “Vibe coded trash,” launching a meme storm about benchmarks measured in vibes-per-second. A few nerds dropped related links like lite3, but the thread’s heartbeat is the tug‑of‑war between “Redis killer!” and “slow where it matters.” The verdict so far: promising architecture, punchy claims, and a comments section doing a full‑contact fact check.
Key Points
- •SNKV is an ACID-compliant key–value store built directly on SQLite’s B‑Tree layer, bypassing the SQL stack.
- •It removes SQL parsing, compilation, VDBE execution, query optimization, index and schema management to cut overhead.
- •Core SQLite components retained include the B‑Tree engine, pager module (journaling, cache), and OS interface (I/O, locks).
- •Claims ~50% faster performance on mixed KV workloads (70% reads, 20% writes, 10% deletes) due to reduced SQL overhead.
- •Provides a simple C API with examples for Put/Get/Delete and transactions; tests and benchmarks are in the tests/ directory.