node-server

Hono's Node.js adapter was underperforming compared to h3, fastify, and koa — especially on body parsing. These changes bring Hono from 4th place to 2nd, virtually tied with h3. Addresses #76 On my macOS (M1-Max) Before After Improvement Benchmarked with bun-http-framework-benchmark using at 500 concurrent connections for 10s per test. Changes 1. Direct body reading from () The biggest bottleneck was body parsing. When or was called on the pseudo-Request, it triggered a long chain of allocations: This created 5+ intermediate objects and piped the body through multiple stream conversion layers on every request. The fix overrides , , , and on the request prototype to read directly from the Node.js using event-based I/O: If is later called (e.g. user accesses ), the cached buffer is reused as the body source to avoid double-consumption. Falls back to the original path when the full object has already been materialized. 2. URL construction fast-path () was called for every request purely for validation and normalization. Since Hono's and both do string-based parsing (no URL object needed), we skip for common requests: Fast path: validate host with a charCode loop, concatenate Slow path: fall back to only when is present in the path (needed for path normalization) 3. optimization () Added check before iterating headers. In the common case (no cookies), this skips the per-header string comparison against . Thanks for the great lib :)