Changelog

0.1.4

Changes since v0.1.3.

Reworked the router around compact per-method radix nodes and stable route-cache snapshots.
Improved exact static-route hits, parameterized routing, large mixed route trees, method mismatch handling, and high-cardinality cache behavior.
Reduced context reuse overhead while preserving zero request-time allocations across Zinc's primary static, parameter, and not-found dispatch benchmarks.
Improved feature paths without broad router changes: invalid JSON binding, JSON happy paths, validation failures, large JSON binding, and static-file handling.

Expanded the comparison suite with realistic large route trees, cold paths, scenario builds, 404 and 405 paths, parallel dispatch, route registration, and cache-matrix coverage.
Separated throughput/RPS benchmarks from the main suite so they remain an optional workload.
In the latest Apple M1 Pro comparison snapshot, Zinc records the lowest raw latency in 64/77 comparable rows and finishes first or second in all 77/77 rows against Gin, Echo, and Chi.

Added regression coverage for route precedence, wildcard routing, method handling, cache cardinality, route memory, context reuse, and concurrent dispatch.
Removed the accidentally tracked benchmark binary.
Preserved the existing public API while establishing a tested performance baseline for the next Zinc release cycle.