golang-lru
HashiCorp's generic-typed, thread-safe LRU cache for Go, with ARC, 2Q, and TTL-expiring variants.
Repository Health
Technical Analysis
golang-lru is HashiCorp’s fixed-size, thread-safe LRU (Least Recently Used) cache implementation for Go, built on generics for full type safety. It began as a wrapper around Groupcache’s cache logic and now ships four interchangeable cache variants in one module: a classic LRU, a 2Q cache that resists scan pollution, an ARC (Adaptive Replacement Cache) that balances recency and frequency automatically, and an expirable LRU with per-entry TTL.
The library is used throughout HashiCorp’s own tooling, including Vault, Consul, and Nomad, as well as many other Go projects that need bounded in-memory caching without standing up an external cache server. Each variant lives as a small, dependency-light package under a shared module, so teams pick exactly the eviction strategy their workload needs.
What You Get
- A generic, thread-safe Cache[K, V] type implementing classic LRU eviction
- An expirable subpackage adding per-cache TTL expiration on top of LRU semantics
- A 2Q cache variant that resists pollution from one-time scan access patterns
- An ARC (Adaptive Replacement Cache) variant that auto-balances recency vs frequency
- Optional eviction callbacks fired outside the internal lock for safe cleanup/metrics hooks
Common Use Cases
- In-process caching of expensive API or database responses to cut redundant calls
- Bounding memory for connection or session pools inside infrastructure daemons
- Building TTL-based rate limiters or short-lived counters with the expirable variant
- Scan-resistant read-through caching where hot keys must survive occasional bulk scans
Under The Hood
Architecture
The module is organized as a set of sibling packages sharing one generic doubly-linked-list implementation in internal/list.go (adapted from Go’s container/list, with an added expiry timestamp field). The root lru package wraps a non-thread-safe simplelru.LRU[K, V] behind a sync.RWMutex and an optional eviction-callback buffer that is drained outside the critical section (see Cache.Purge and Cache.onEvicted in lru.go) to avoid calling user code while holding the lock. The arc package composes four simplelru.LRUCache instances (T1/T2 for recent/frequent entries, B1/B2 as ghost eviction-history lists) to implement the ARC algorithm, while expirable reimplements the list with time-aware eviction and a background sweep goroutine. This layered composition means every variant reuses the same core list/map primitives rather than duplicating cache logic.
Tech Stack
Pure Go with no third-party runtime dependencies; go.mod declares go 1.19 to use generics throughout (K comparable, V any). The arc subpackage is its own Go module with its own go.mod, isolating it for versioning. Build/test tooling is a plain Makefile plus go test; CI runs via GitHub Actions (.github/workflows/ci.yml) with Dependabot keeping the module and Actions versions current, and a HashiCorp copywrite check enforces license headers.
Code Quality
Every package ships a matching _test.go file (lru_test.go, simplelru/lru_test.go, arc/arc_test.go, expirable/expirable_lru_test.go, 2q_test.go) exercising eviction order, capacity limits, callback firing, and expiry timing. Errors are returned explicitly (e.g. New returns an error for non-positive size) rather than panicking, and public types are documented with doc comments throughout. Locking is consistent: every thread-safe wrapper holds its own sync.RWMutex rather than sharing state across packages.
What Makes It Unique Rather than picking one eviction policy, the module exposes four algorithmically distinct caches behind a near-identical API, letting callers swap ARC for plain LRU or add TTL expiration without changing call sites. The ARC implementation in particular is a comparatively rare open-source instance of IBM’s patented Adaptive Replacement Cache algorithm made available under a permissive-enough license for general use, giving Go projects an alternative to manually tuning 2Q’s fixed parameters.
Used by 18 apps in this directory
Beta9
Developer Tools · AI Development · Data Engineering
Run AI workloads at scale with a Pythonic serverless runtime that handles GPU inference, background jobs, and sandboxes with zero infrastructure overhead.
Bytebase
Devops
An open-source database CI/CD and DevSecOps platform — schema migration review, GitOps-driven changes, data masking, and access control across MySQL, PostgreSQL, Oracle, Snowflake, MongoDB, and more.
Cozy Stack
File Storage · Productivity
Self-hosted personal cloud platform that unifies your files, apps, and devices in one private space you fully control.
Dolt
Databases · Data Engineering · Developer Tools
The SQL database you can branch, merge, diff, and clone — Git for your data, MySQL-compatible and ready for multi-agent AI workflows.
Gitea
Devops · Developer Tools · Project Management
Self-hosted DevOps in a single Go binary — Git hosting, GitHub Actions-compatible CI/CD, and 30+ package registries without any SaaS dependency.
GitLab
Devops · Developer Tools
The complete DevOps platform that unifies Git hosting, CI/CD, issue tracking, and security scanning into a single self-hostable application.
Grafana
Monitoring · Analytics
The open-source observability platform that unifies metrics, logs, and traces from any data source into dynamic, queryable dashboards.
Harness Open Source
Developer Tools · Devops · Code Editors
A unified open source DevOps platform combining Git hosting, CI/CD pipelines, cloud development environments, and artifact registries in a single self-hosted system.
Hatchet
AI Development · Developer Tools · Automation
A Postgres-backed orchestration engine for background tasks, AI agents, and durable workflows that replaces Redis queues and multi-datastore durable execution platforms with a single self-hostable service.