common-library-golang

PHCP: common-library-golang

Common-library-golang is a collection of functional components used within the PHCP ecosystem, providing numerous components for microservice development, it provides some out-of-the-box functions, such as environment, log, database, etc. To make it more widely available, it is now open-sourced under the Apache license.

Requirements

• Go 1.25+

Installation

go get github.com/phcp-tech/common-library-golang

Build and Test

go mod tidy
go vet ./...
go build ./...

Short mode test — fast CI verification

Some tests start a real TCP listener (e.g. httpserver integration tests). Pass -short to skip them and keep the test run lightweight:

go test ./... -short

Full test run — including integration tests

go test ./...

Run test with coverage

# All packages
go test ./... -cover -timeout 60s

Packages

Group	Package	Import path	Description
Basic	env	.../common-library-golang/env	TOML config + environment variable loader
Basic	log	.../common-library-golang/log	Structured JSON logger with file rotation and ringbuf
Basic	health	.../common-library-golang/health	Composable health-check aggregator for /health endpoints
Basic	version	.../common-library-golang/version	Application version and build metadata for /version endpoints
Basic	shutdown	.../common-library-golang/shutdown	Block until OS signal or programmatic trigger, then continue for cleanup
Basic	ringbuf	.../common-library-golang/ringbuf	Lock-free ring buffers (SPSC and MPSC)
Basic	maps	.../common-library-golang/maps	Thread-safe generic concurrent maps with pluggable replacement strategies
Basic	cgroup	.../common-library-golang/cgroup	Read CPU and memory resource limits from cgroup v2 (Linux only; returns 0 on other platforms)
Basic	metrics	.../common-library-golang/metrics	Runtime and system metrics snapshot (CPU, memory, goroutines, cgroup limits, uptime)
Bootstrap	bootstrap	.../common-library-golang/bootstrap	Sequential Init + LIFO Close orchestrator; 1st Add() = env, 2nd Add() = log (convention)
Bootstrap	env/component	.../common-library-golang/env/component	IComponent adapter for the env package
Bootstrap	log/component	.../common-library-golang/log/component	IComponent adapter for the log package
Bootstrap	auth/component	.../common-library-golang/auth/component	IComponent adapter for the auth (casbin) package
Bootstrap	gin/component	.../common-library-golang/gin/component	IComponent adapter for the gin engine
Bootstrap	redis/component	.../common-library-golang/redis/component	IComponent adapter for the redis package
Bootstrap	dbgorm/clickhouse/component	.../common-library-golang/dbgorm/clickhouse/component	IComponent adapter for the GORM ClickHouse connection (pings on Init)
Bootstrap	dbgorm/mysql/component	.../common-library-golang/dbgorm/mysql/component	IComponent adapter for the GORM MySQL connection (pings on Init)
Bootstrap	dbgorm/postgres/component	.../common-library-golang/dbgorm/postgres/component	IComponent adapter for the GORM PostgreSQL connection (pings on Init)
Bootstrap	dbgorm/sqlite/component	.../common-library-golang/dbgorm/sqlite/component	IComponent adapter for the GORM SQLite connection
Bootstrap	dbsqlc/mysql/component	.../common-library-golang/dbsqlc/mysql/component	IComponent adapter for the MySQL connection (sql.Open, lazy)
Bootstrap	dbsqlc/postgres/component	.../common-library-golang/dbsqlc/postgres/component	IComponent adapter for the PostgreSQL pool
Bootstrap	dbsqlc/sqlite/component	.../common-library-golang/dbsqlc/sqlite/component	IComponent adapter for the SQLite connection
Bootstrap	dbsqlx/clickhouse/component	.../common-library-golang/dbsqlx/clickhouse/component	IComponent adapter for the sqlx ClickHouse connection (pings on Init)
Bootstrap	dbsqlx/clickhouse-native/component	.../common-library-golang/dbsqlx/clickhouse-native/component	IComponent adapter for the native-driver ClickHouse connection (lazy open)
Bootstrap	dbsqlx/mysql/component	.../common-library-golang/dbsqlx/mysql/component	IComponent adapter for the sqlx MySQL connection (pings on Init)
Bootstrap	dbsqlx/postgres/component	.../common-library-golang/dbsqlx/postgres/component	IComponent adapter for the sqlx PostgreSQL connection (pings on Init)
Bootstrap	dbsqlx/sqlite/component	.../common-library-golang/dbsqlx/sqlite/component	IComponent adapter for the sqlx SQLite connection
Bootstrap	httpserver/component	.../common-library-golang/httpserver/component	IComponent adapter for the HTTP server; ComponentWithRunner for custom runner injection
Bootstrap	httpserver/componentwithlambda	.../common-library-golang/httpserver/componentwithlambda	IComponent adapter with AWS Lambda support; selects runner via app.runmode
Database	redis	.../common-library-golang/redis	Redis client (standalone and cluster) with connection pool and key scan utilities
Database	dbgorm/clickhouse	.../common-library-golang/dbgorm/clickhouse	ClickHouse via GORM — eager ping on Open, shared process-wide *gorm.DB
Database	dbgorm/mysql	.../common-library-golang/dbgorm/mysql	MySQL via GORM — eager ping on Open, shared process-wide *gorm.DB
Database	dbgorm/postgres	.../common-library-golang/dbgorm/postgres	PostgreSQL via GORM — eager ping on Open, shared process-wide *gorm.DB
Database	dbgorm/sqlite	.../common-library-golang/dbgorm/sqlite	SQLite via GORM — embedded, no network, file auto-created on Open
Database	dbgorm/sqlite/vfs	.../common-library-golang/dbgorm/sqlite/vfs	SQLite via GORM over embedded FS — binary-embedded read-only databases
Database	dbsqlc/mysql	.../common-library-golang/dbsqlc/mysql	MySQL connection via standard database/sql and go-sql-driver
Database	dbsqlc/postgres	.../common-library-golang/dbsqlc/postgres	PostgreSQL connection pool via pgx/v5
Database	dbsqlc/sqlite	.../common-library-golang/dbsqlc/sqlite	SQLite connection via pure-Go modernc driver
Database	dbsqlc/sqlite/vfs	.../common-library-golang/dbsqlc/sqlite/vfs	SQLite over embedded FS via VFS (binary-embedded databases)
Database	dbsqlx/clickhouse	.../common-library-golang/dbsqlx/clickhouse	ClickHouse via sqlx (clickhouse-go/v2 database/sql driver) — eager ping on Open, shared process-wide *sqlx.DB
Database	dbsqlx/clickhouse-native	.../common-library-golang/dbsqlx/clickhouse-native	ClickHouse native TCP client via clickhouse-go/v2 — returns driver.Conn, not *sqlx.DB
Database	dbsqlx/mysql	.../common-library-golang/dbsqlx/mysql	MySQL via sqlx — eager ping on Open, shared process-wide *sqlx.DB
Database	dbsqlx/postgres	.../common-library-golang/dbsqlx/postgres	PostgreSQL via sqlx (pgx stdlib driver) — eager ping + SHOW search_path on Open
Database	dbsqlx/sqlite	.../common-library-golang/dbsqlx/sqlite	SQLite via sqlx — embedded, no network, file auto-created on Open
Database	dbsqlx/sqlite/vfs	.../common-library-golang/dbsqlx/sqlite/vfs	SQLite via sqlx over embedded FS — binary-embedded read-only databases
Network	network	.../common-library-golang/network	Network utilities helpers
Network	auth	.../common-library-golang/auth	Casbin RBAC authorisation middleware for Gin
Network	token	.../common-library-golang/token	JWT access/refresh token creation, parsing, and Gin middleware
Network	gin	.../common-library-golang/gin	Gin engine factory with slog request logging and CORS
Network	gin/pprof	.../common-library-golang/gin/pprof	Optional pprof profiling endpoints for Gin (explicit opt-in)
Network	httpclient	.../common-library-golang/httpclient	Resty-based HTTP client with retry, JWT auth, and JSON helpers
Network	httpclient/retryable	.../common-library-golang/httpclient/retryable	hashicorp/go-retryablehttp wrapper returning standard *http.Client
Network	httpserver	.../common-library-golang/httpserver	Production HTTP/HTTPS server with timeouts and graceful shutdown
Network	httpserver/lambda	.../common-library-golang/httpserver/lambda	AWS Lambda adapter implementing httpserver.IRunner (explicit opt-in)

---

env — Configuration Management

Loads a TOML config file and merges OS environment variables on top. Built on koanf. Implements the singleton pattern: only the first InitEnv call takes effect.

import "github.com/phcp-tech/common-library-golang/env"

// Call once at startup before any Env() usage.
if err := env.InitEnv("config.toml"); err != nil {
    log.Fatal(err)
}

host := env.Env().String("server.host")
port := env.Env().Int("server.port")

For single-binary deployments the config file can be embedded at compile time:

//go:embed config.toml
var configFS embed.FS

env.InitEnv("config.toml", &configFS)

See full examples.

---

log — Structured Logging

Structured JSON logging via log/slog with UTC timestamps. File writes are fully asynchronous: the caller returns immediately after pushing the formatted entry into an internal RingMPSC buffer; a dedicated consumer goroutine performs the actual I/O. This makes the package suitable for high-throughput, latency-sensitive scenarios where blocking on disk I/O is unacceptable.

InitLog must be called once at application startup before any log function. Omit the argument for stdout output at INFO level; pass a Config to customise.

import "github.com/phcp-tech/common-library-golang/log"

// Stdout mode at default INFO level.
log.InitLog()
log.Info("application started")
log.Infof("listening on port %d", 8080)
log.InfoWith("request", "method", "GET", "path", "/api/v1", "status", 200)

// Stdout mode with custom level.
log.InitLog(&log.Config{Level: "debug"})

// File mode: set FilePath to enable rotating file logging.
log.InitLog(&log.Config{
    Level:      "info",
    FilePath:   "/var/log/app.log",
    MaxSizeMB:  100,
    MaxBackups: 7,
    MaxAgeDays: 30,
    Compress:   true,
})
defer log.Close() // flush async buffer and close file on shutdown

Available log functions: Debug / Info / Warn / Error and their f (format) and With (structured key-value) variants. Log level can be changed at runtime with SetLevel.

See full examples.

---

health — Composable Health Checks

Interface-based health-check aggregator for /health HTTP endpoints. Each infrastructure package supplies a Checker that owns its own component name and reachability status; Check runs all checkers and returns their combined results.

import (
    "github.com/phcp-tech/common-library-golang/health"
    db    "github.com/phcp-tech/common-library-golang/dbsqlc/postgres"
    cache "github.com/phcp-tech/common-library-golang/redis"
)

// Compose any number of checkers in a /health handler.
router.GET("/health", func(c *gin.Context) {
    c.JSON(http.StatusOK, health.Check(
        c.Request.Context(),
        db.HealthChecker(),
        cache.HealthChecker(),
    ))
})
// JSON response: [{"name":"postgres","status":1},{"name":"redis","status":1}]

Constant	Value	Meaning
health.StatusHealthy	1	Component is reachable
health.StatusUnhealthy	0	Component is unreachable or not initialised

See full examples.

---

version — Application Version Metadata

Returns application version and build metadata for /version HTTP endpoints. Reads app.name, app.version, and app.env.value from env.Env(), and populates GoVersion / BuildInfo from the embedded Go build info. Requires env.InitEnv to be called at application startup.

import "github.com/phcp-tech/common-library-golang/version"

router.GET("/version", func(c *gin.Context) {
    c.JSON(http.StatusOK, version.Get())
})
// JSON response:
// {
//   "name": "my-service",
//   "version": "1.2.3",
//   "environment": "production",
//   "goVersion": "go1.26.1",
//   "buildInfo": "v1.2.3"
// }

See full examples.

---

shutdown — Graceful Shutdown

Two primitives for application shutdown coordination:

• Wait blocks the calling goroutine until an OS signal (SIGINT, SIGTERM, SIGHUP, SIGQUIT) or Trigger is called. After it returns, the caller performs cleanup and exits.
• Trigger unblocks Wait programmatically from any goroutine — useful for a /shutdown HTTP endpoint or a metrics failure handler. Safe to call multiple times.

import "github.com/phcp-tech/common-library-golang/shutdown"

// In main: start services, then block until shutdown.
shutdown.Wait()

// Cleanup runs here (or via defer before Wait).
runner.Shutdown(ctx)

// From a /shutdown HTTP endpoint or metrics failure handler.
shutdown.Trigger()

Note: Trigger has no effect when the process is forcibly terminated by the OS (e.g. IDE stop button on Windows calls TerminateProcess — no signal is delivered).

See full examples.

---

ringbuf — Ring Buffers

High-performance fixed-capacity ring buffers for producer-consumer pipelines.

Type	Use case	Thread safety
RingSPSC	Single producer, single consumer	Lock-free (atomic only)
RingMPSC	Multiple producers, single consumer	Mutex on producer side

Both types support an optional ProcessFunc that starts a consumer goroutine automatically, or manual Pop / TryPop for caller-managed consumption. Push blocks when the buffer is full (backpressure); TryPush returns false immediately.

import "github.com/phcp-tech/common-library-golang/ringbuf"

// SPSC — single producer, automatic consumer goroutine
rb := ringbuf.NewRingSPSC(ringbuf.RingSPSCConfig[string]{
    Capacity:    1024,
    ProcessFunc: func(s string) { fmt.Println(s) },
})
rb.Push("hello")
rb.Close() // drain and wait for consumer to finish

// MPSC — multiple producers, automatic consumer goroutine
rb := ringbuf.NewRingMPSC(ringbuf.RingMPSCConfig[[]byte]{
    Capacity:    4096,
    ProcessFunc: func(b []byte) { os.Stdout.Write(b) },
})

See full examples.

Performance

Benchmarks run with go test -bench=. -benchtime=3s -benchmem. All operations produce zero heap allocations.

Test environment

CPU	11th Gen Intel® Core™ i7-11850H @ 2.50 GHz (8 cores / 16 threads)
RAM	32 GB
OS	Windows 11 Enterprise
Go	1.26.2 windows/amd64

Results

Benchmark	ns/op	Throughput	Allocs
SPSC Push (1P-1C, blocking)	76.65	~13.0 M ops/s	0
SPSC TryPush (1P-1C, non-blocking)	87.91	~11.4 M ops/s	0
SPSC ProducerConsumer (end-to-end with ProcessFunc)	83.72	~11.9 M ops/s	0
SPSC Push string (realistic log payload)	91.06	~11.0 M ops/s	0
MPSC Push (1 producer)	151.1	~6.6 M ops/s	0
MPSC Push (4 producers concurrent)	192.2	~5.2 M ops/s	0
MPSC Push (8 producers concurrent)	200.7	~5.0 M ops/s	0
MPSC ProducerConsumer (4P end-to-end)	197.5	~5.1 M ops/s	0
Go channel (buffered 4096, reference)	126.3	~7.9 M ops/s	0

Key observations

• SPSC is ~1.6× faster than a buffered Go channel for the same single-producer / single-consumer pattern.
• MPSC trades some throughput for multi-producer safety via a mutex; with 8 concurrent producers it remains above 5 M ops/s.
• Throughput scales gracefully under producer contention: going from 1 to 8 producers adds only ~33% latency per item.
• Both types maintain zero allocations per operation, making them suitable as the async I/O backbone for the log package.

---

maps — Thread-Safe Concurrent Maps

Thread-safe generic concurrent maps backed by orcaman/concurrent-map with pluggable replacement strategies.

Two implementations are provided:

Type	Key	Value	Strategy
CMap	string (fixed)	int64 (fixed)	Greater-value wins (built-in)
CMapGen[K, V]	any comparable	any	Configurable via SetDefaultCompare or SetDefaultStrategy

Built-in strategy types: NumericGreaterStrategy, AlwaysReplaceStrategy, TimestampStrategy.

import "github.com/phcp-tech/common-library-golang/maps"

// Generic map — configure a default compare function.
m := maps.NewCMapGen[string, int64]()
m.SetDefaultCompare(func(old, new int64) bool { return new > old })

m.Set("EURUSD", 10500)
m.Replace("EURUSD", 10600)       // stored: 10600 > 10500
m.ReplaceAlways("EURUSD", 9000)  // always stored
m.ReplaceIfNotExists("GBPUSD", 12800) // stored only if key absent

// Atomic read-modify-write.
m.UpsertWithCallback("USDJPY", 15100, func(exists bool, old, new int64) int64 {
    if !exists || new > old { return new }
    return old
})

// Fixed string→int64 map (no configuration needed).
cm := maps.NewCMap()
cm.Set("tick", 10000)
cm.Replace("tick", 10050) // stored: 10050 > 10000

See full examples.

Performance

Benchmarks run with go test -bench=. -benchtime=3s -benchmem.

Test environment — same machine as ringbuf benchmarks: Intel® Core™ i7-11850H @ 2.50 GHz · 8 cores / 16 threads · 32 GB RAM · Go 1.26.2 / Windows 11

CMap (fixed string→int64, built-in greater-value strategy)

Benchmark	ns/op	Throughput	B/op	Allocs
Set	33.01	~30.3 M ops/s	0	0
Get	15.68	~63.8 M ops/s	0	0
Replace	43.42	~23.0 M ops/s	0	0
Set (parallel, 16 goroutines)	90.67	~11.0 M ops/s	0	0
Replace (parallel, 16 goroutines)	434.8	~2.3 M ops/s	0	0

CMapGen (generic [K comparable, V any])

Benchmark	ns/op	Throughput	B/op	Allocs	Note
Set	94.20	~10.6 M ops/s	0	0
Get	60.54	~16.5 M ops/s	0	0
Replace (default strategy)	137.0	~7.3 M ops/s	32	3	⚠ fmt.Sprint fallback allocates
ReplaceWithCompare	52.94	~18.9 M ops/s	0	0	recommended
ReplaceAlways	29.78	~33.6 M ops/s	0	0
UpsertWithCallback	58.68	~17.0 M ops/s	0	0
Set (parallel, 16 goroutines)	74.32	~13.5 M ops/s	0	0
Replace (parallel, 16 goroutines)	81.61	~12.3 M ops/s	32	3	⚠ see above
Mixed read/write (parallel)	26.08	~38.3 M ops/s	10	1

sync.Map (stdlib reference)

Benchmark	ns/op	Throughput	B/op	Allocs
Store	60.96	~16.4 M ops/s	56	1
Load	10.66	~93.8 M ops/s	0	0
Store (parallel, 16 goroutines)	103.7	~9.6 M ops/s	56	1

Key observations

• CMap.Set is ~1.8× faster than sync.Map.Store and allocates nothing.
• CMapGen.Replace with a typed compare function (ReplaceWithCompare / SetDefaultCompare) is zero-allocation; the no-argument Replace() fallback uses fmt.Sprint internally and produces 3 allocations — always call SetDefaultCompare or SetDefaultStrategy at construction time.
• CMapGen.ReplaceAlways is the fastest write path (~33.6 M ops/s) when no conditional logic is needed.
• Under 16-goroutine parallel write contention CMap.Replace degrades to ~2.3 M ops/s due to single-key hot-spot; spreading writes across many keys restores throughput.

---

cgroup — Linux Resource Limits

Reads CPU and memory resource limits from the cgroup v2 unified hierarchy (/sys/fs/cgroup). Intended for containerised workloads (Kubernetes, Docker) where the process runs inside a cgroup with configured CPU and memory constraints.

On non-Linux platforms (macOS, Windows) all functions return (0, nil) — no cgroup filesystem is available and no error is raised.

Function	cgroup v2 file	Returns
CPULimitMilli()	cpu.max	CPU limit in millicores; 0 = unlimited
CPURequestMilli()	cpu.weight	CPU request in millicores via weight→shares→mCPU
MemoryLimitBytes()	memory.max	Memory limit in bytes; 0 = unlimited
MemoryRequestBytes()	memory.low	Memory soft limit in bytes; 0 = not set

import "github.com/phcp-tech/common-library-golang/cgroup"

// Read CPU limit and cap GOMAXPROCS accordingly.
if milli, err := cgroup.CPULimitMilli(); err == nil && milli > 0 {
    cpus := milli / 1000
    if cpus < 1 {
        cpus = 1
    }
    runtime.GOMAXPROCS(cpus)
}

// Size an in-process cache as 25 % of the container memory limit.
if limitBytes, err := cgroup.MemoryLimitBytes(); err == nil && limitBytes > 0 {
    cacheBytes := limitBytes / 4
    cache.SetMaxSize(cacheBytes)
}

See full examples.

---

metrics — Runtime Metrics Snapshot

Collects a one-shot snapshot of key runtime and system metrics as a flat []NameValue slice, suitable for a /metrics or /health HTTP endpoint.

Each call samples CPU usage over one second via gopsutil, so it is intended for periodic polling (e.g. every 10–30 s), not hot-path use.

Metric name	Description
cpuPercent	Process CPU usage (%) across all cores, sampled over 1 s
memorySize	Resident set size (RSS) in MiB
threads	OS thread count
goroutines	Live goroutine count
gomaxprocs	runtime.GOMAXPROCS(0)
numCPU	runtime.NumCPU()
cpuRequest	cgroup v2 CPU request in millicores (0 if not set or non-Linux)
cpuLimit	cgroup v2 CPU limit in millicores (0 if unlimited or non-Linux)
memoryRequest	cgroup v2 memory soft limit in bytes (0 if not set or non-Linux)
memoryLimit	cgroup v2 memory limit in bytes (0 if unlimited or non-Linux)
age	Process uptime formatted as Xd Xh Xm Xs

import "github.com/phcp-tech/common-library-golang/metrics"

// Periodic metrics poll — call from a background goroutine.
snapshot := metrics.GetMetrics()

// Look up a specific entry.
for _, nv := range snapshot {
    if nv.Name == "goroutines" {
        slog.Info("runtime", "goroutines", nv.Value)
    }
}

// Serialize to JSON for a /health endpoint.
b, _ := json.Marshal(snapshot)
w.Write(b)

See full examples.

---

bootstrap — Application Lifecycle Orchestrator

bootstrap orchestrates the sequential initialization and LIFO cleanup of application components at startup.

Component registration order convention

The first Add() call MUST register the env component. The second Add() call MUST register the log component. All subsequent calls (Add, AddParallel, PreReady) are free.

This convention exists because:

• every component's Init() reads configuration via env.Env(), so env must initialize first;
• Go's slog package has a usable default instance from program start (writes to stderr), so Init() failures at any stage — including before log.Init() — are captured by slog;
• env.Close() is a no-op, so LIFO naturally makes log.Close() the last meaningful shutdown operation.

import (
    "github.com/phcp-tech/common-library-golang/bootstrap"
    envComp  "github.com/phcp-tech/common-library-golang/env/component"
    logComp  "github.com/phcp-tech/common-library-golang/log/component"
    dbComp   "github.com/phcp-tech/common-library-golang/dbsqlc/postgres/component"
    ginComp  "github.com/phcp-tech/common-library-golang/gin/component"
    httpComp "github.com/phcp-tech/common-library-golang/httpserver/component"
)

func main() {
    var router *gin.Engine
    bootstrap.New().
        Add(envComp.Component("config/app.toml", &configFS)). // 1st — MUST be env
        Add(logComp.Component()).                              // 2nd — MUST be log
        AddParallel(dbComp.Component()).
        PreReady(migrate).
        PreReady(initServices).
        Add(ginComp.Component(func(r *gin.Engine) { router = r; mount(r) })).
        Add(httpComp.Component(func() http.Handler { return router })).
        PostReady(func() { slog.Info("server ready") }).
        Run()
}

API

Method	Signature	Description
New	() *App	Creates an empty orchestrator
Add	(cs ...IComponent) *App	Sequential phase: Init in registration order, Close in LIFO order. 1st call = env, 2nd call = log
AddParallel	(cs ...IComponent) *App	Concurrent phase: all Init calls run in parallel; all must succeed before the next step starts
PreReady	(fn func() error) *App	Inline setup function in the startup sequence; non-nil error aborts startup; no Close, not tracked in LIFO
PostReady	(fn func()) *App	Notification hook invoked after all steps succeed; multiple calls accumulate in registration order
Run	()	Executes all steps, waits for SIGINT/SIGTERM, then closes in LIFO order

IComponent interface

type IComponent interface {
    Name() string   // displayed in log messages
    Init() error    // non-nil error aborts startup and exits with code 1
    Close()         // called during shutdown in LIFO order; must never panic
}

Use bootstrap.Func to wrap a function pair as an IComponent when Close logic exists but no struct is needed:

bootstrap.Func("worker", startWorker, stopWorker)

Startup and shutdown sequence

New()
  → Add(env).Init   ← slog (default handler writes to stderr before log.Init)
  → Add(log).Init   ← slog
  → steps (in registration order)
        ├── stepPhase    → Init, added to closeStack
        └── stepPreReady → fn(), not added to closeStack
  → PostReady callbacks
  → wait for signal
        ↓
  ← single closeAll (LIFO)
        custom components (reverse order)
        → log.Close   (last meaningful close — env.Close is a no-op)
        → env.Close   (no-op)

PreReady

PreReady registers a function into the startup sequence, sharing the same ordered list as Add/AddParallel. A non-nil error aborts startup identically to a failed component Init(). PreReady steps have no Close and do not participate in LIFO shutdown.

.AddParallel(dbComp.Component()).
PreReady(migrate).        // runs after DB is ready
PreReady(initServices).   // runs after migrate
Add(ginComp.Component(mount)).

	bootstrap.Func(..., nil)	PreReady(fn)
Appears in step list	✓ as stepPhase	✓ as stepPreReady
Participates in LIFO closeStack	✓ (Close is a no-op)	✗
Semantic	"a component without Close"	"one-shot code in the startup sequence"

PostReady

PostReady can be called multiple times. Each call appends a callback; all run in registration order after every step succeeds and before the process blocks on an OS signal. Suitable for actions that do not affect request-handling correctness:

.PostReady(func() { slog.Info("server ready", "addr", ":8080") }).
PostReady(func() { discovery.Register(serviceID) })

Code placed after Run() in main, or in a defer before Run(), executes after the full shutdown sequence completes and is suitable for post-shutdown teardown.

See full examples.

---

Bootstrap Component Packages

Each base package ships a companion component/ sub-package that adapts it to the IComponent interface. Component packages read all configuration from env.Env() inside Init() — never in the constructor — so they are safe to construct before bootstrap.New() is called.

Component package	Env keys read on Init
env/component	(none — reads the config file itself)
log/component	log.level, log.file.path, log.file.max.size.mb, log.file.max.backups, log.file.max.age.days, log.file.compress
auth/component	(model and policy passed as constructor arguments)
gin/component	app.env.value, cors.allow.origins.prod, cors.allow.origins.dev
redis/component	redis.clusters, redis.database, redis.password
dbgorm/clickhouse/component	db.host, db.port, db.name, db.username, db.password, db.max.open.conns, db.max.idle.conns, db.conn.max.lifetime, db.conn.max.idletime
dbgorm/mysql/component	db.host, db.port, db.name, db.username, db.password, db.max.open.conns, db.max.idle.conns, db.conn.max.lifetime, db.conn.max.idletime
dbgorm/postgres/component	db.host, db.port, db.name, db.schema, db.username, db.password, db.max.open.conns, db.max.idle.conns, db.conn.max.lifetime, db.conn.max.idletime
dbgorm/sqlite/component	db.sqlite.path
dbsqlc/mysql/component	db.host, db.port, db.name, db.username, db.password, db.max.open.conns, db.max.idle.conns, db.conn.max.lifetime, db.conn.max.idletime
dbsqlc/postgres/component	db.host, db.port, db.name, db.schema, db.username, db.password, db.pool.*
dbsqlc/sqlite/component	db.sqlite.path
dbsqlx/clickhouse/component	db.host, db.port, db.name, db.username, db.password, db.max.open.conns, db.max.idle.conns, db.conn.max.lifetime, db.conn.max.idletime
dbsqlx/clickhouse-native/component	db.host, db.port, db.name, db.username, db.password, db.max.open.conns, db.max.idle.conns, db.conn.max.lifetime
dbsqlx/mysql/component	db.host, db.port, db.name, db.username, db.password, db.max.open.conns, db.max.idle.conns, db.conn.max.lifetime, db.conn.max.idletime
dbsqlx/postgres/component	db.host, db.port, db.name, db.schema, db.username, db.password, db.max.open.conns, db.max.idle.conns, db.conn.max.lifetime, db.conn.max.idletime
dbsqlx/sqlite/component	db.sqlite.path
httpserver/component	http.server.port
httpserver/componentwithlambda	app.runmode ("aws_lambda" → Lambda runner), http.server.port (other modes)

---

redis — Redis Client

Thread-safe Redis client supporting both standalone and cluster modes, backed by go-redis/v9. Connection pool settings (PoolSize, MinIdleConns) are configurable via Config with sensible defaults (100 / 5). The caller reads values from env.Env() at the composition root — this package has no dependency on env.

import "github.com/phcp-tech/common-library-golang/redis"

// Standalone mode (single address).
cli := redis.NewRedisClient(&redis.Config{
    Clusters: []string{env.Env().String("redis.addr")},
    Password: env.Env().String("redis.password"),
    DB:       env.Env().Int("redis.db"),
})
defer cli.Close()

// Cluster mode — more than one address enables cluster automatically.
cli := redis.NewRedisClient(&redis.Config{
    Clusters:     env.Env().Strings("redis.clusters"),
    PoolSize:     200,
    MinIdleConns: 10,
})

// Basic operations.
cli.Set(ctx, "key", "value", 5*time.Minute)
val, err := cli.Get(ctx, "key")
cli.Del(ctx, "key")
cli.Unlink(ctx, "key") // async deletion — preferred for large keys

// Remove all keys starting with a prefix.
cli.CleanCache(ctx, "user:42:")

// Count matching keys across all nodes.
n, err := cli.GetKeysCount(ctx, "session:*")

Implements the singleton pattern via InitDefault / Default.

Health check

HealthChecker() returns a health.Checker that pings the default client. Reports StatusUnhealthy when no client has been initialised or when the ping fails.

import (
    "github.com/phcp-tech/common-library-golang/health"
    "github.com/phcp-tech/common-library-golang/redis"
)

results := health.Check(c.Request.Context(), redis.HealthChecker())
// → []health.Result{{Name: "redis", Status: health.StatusHealthy}}

See full examples.

---

dbgorm/clickhouse — GORM ClickHouse

GORM-based ClickHouse adapter. GORM's driver pings the server during Open — InitDefault returns a non-nil error immediately when the database is unreachable, causing bootstrap to abort startup. The default instance is a process-wide *gorm.DB shared via dbgorm.Default() / dbgorm.SetDefault().

Note: PrepareStmt is automatically disabled for ClickHouse because the driver maps Prepare() to batch INSERT; it must be off for regular SELECT queries. This is handled transparently inside dbgorm.Open.

import (
    dbgorm     "github.com/phcp-tech/common-library-golang/dbgorm"
    "github.com/phcp-tech/common-library-golang/dbgorm/clickhouse"
)

err := clickhouse.InitDefault(&clickhouse.Config{
    Host:     "localhost",
    Port:     "9440",   // native TCP+TLS; use 8123/8443 for HTTP
    Database: "mydb",
    Username: "user",
    Password: "pass",
})
if err != nil {
    log.Fatal(err)
}

db := dbgorm.Default() // *gorm.DB, ready for GORM operations

Health check

Use dbgorm.HealthChecker() from the root package — the checker is dialect-agnostic and works with any driver initialised via InitDefault.

import (
    dbgorm "github.com/phcp-tech/common-library-golang/dbgorm"
    "github.com/phcp-tech/common-library-golang/health"
)

results := health.Check(c.Request.Context(), dbgorm.HealthChecker())
// → []health.Result{{Name: "database", Status: health.StatusHealthy}}

See full examples.

---

dbgorm/mysql — GORM MySQL

GORM-based MySQL adapter. Unlike dbsqlc/mysql, GORM's driver pings the server during Open — InitDefault returns a non-nil error immediately when the database is unreachable, which causes bootstrap to abort startup. The default instance is a process-wide *gorm.DB shared via dbgorm.Default() / dbgorm.SetDefault() (no sync.Once — can be replaced at runtime).

import (
    dbgorm "github.com/phcp-tech/common-library-golang/dbgorm"
    "github.com/phcp-tech/common-library-golang/dbgorm/mysql"
)

// Open and store as the process-wide default.
err := mysql.InitDefault(&mysql.Config{
    Host:     "localhost",
    Port:     "3306",
    Database: "mydb",
    Username: "user",
    Password: "pass",
})
if err != nil {
    log.Fatal(err) // server was unreachable — GORM pinged on Open
}

db := dbgorm.Default() // *gorm.DB, ready for GORM operations

For cases that require multiple databases, use NewMySQL directly and store the returned *gorm.DB in explicit fields rather than the shared default.

Health check

Use dbgorm.HealthChecker() from the root package — the checker is dialect-agnostic and works with any driver initialised via InitDefault.

import (
    dbgorm "github.com/phcp-tech/common-library-golang/dbgorm"
    "github.com/phcp-tech/common-library-golang/health"
)

results := health.Check(c.Request.Context(), dbgorm.HealthChecker())
// → []health.Result{{Name: "database", Status: health.StatusHealthy}}

See full examples.

---

dbgorm/postgres — GORM PostgreSQL

GORM-based PostgreSQL adapter. GORM's driver pings the server and runs SHOW search_path during Open — InitDefault returns a non-nil error immediately when the database is unreachable, causing bootstrap to abort startup. Supports an optional SearchPath for schema isolation. The default instance is a process-wide *gorm.DB shared via dbgorm.Default() / dbgorm.SetDefault().

import (
    dbgorm "github.com/phcp-tech/common-library-golang/dbgorm"
    "github.com/phcp-tech/common-library-golang/dbgorm/postgres"
)

err := postgres.InitDefault(&postgres.Config{
    Host:       "localhost",
    Port:       "5432",
    Database:   "mydb",
    Username:   "user",
    Password:   "pass",
    SearchPath: "myschema", // optional
})
if err != nil {
    log.Fatal(err) // server was unreachable
}

db := dbgorm.Default() // *gorm.DB, ready for GORM operations

Health check

Use dbgorm.HealthChecker() from the root package — the checker is dialect-agnostic and works with any driver initialised via InitDefault.

import (
    dbgorm "github.com/phcp-tech/common-library-golang/dbgorm"
    "github.com/phcp-tech/common-library-golang/health"
)

results := health.Check(c.Request.Context(), dbgorm.HealthChecker())
// → []health.Result{{Name: "database", Status: health.StatusHealthy}}

See full examples.

---

dbgorm/sqlite — GORM SQLite

GORM-based SQLite adapter backed by glebarez/sqlite (pure Go, no CGO). SQLite is an embedded database — no network connection is involved. InitDefault opens (and auto-creates) the file; it fails only when the path is empty or the file cannot be created. MaxOpenConns is fixed at 1 to prevent database is locked errors under concurrent writes.

import (
    dbgorm "github.com/phcp-tech/common-library-golang/dbgorm"
    "github.com/phcp-tech/common-library-golang/dbgorm/sqlite"
)

// File-based database.
err := sqlite.InitDefault(&sqlite.Config{
    Path: "file:app.db?cache=shared",
})
if err != nil {
    log.Fatal(err)
}

db := dbgorm.Default() // *gorm.DB, ready for GORM operations

// In-memory database (tests, ephemeral data).
db, _ := sqlite.NewSQLite(&sqlite.Config{
    Path: "file::memory:?cache=shared",
})

No HealthChecker: SQLite is embedded — if it fails, the application is already broken. A health endpoint for an embedded database provides no operational value.

See full examples.

---

dbgorm/sqlite/vfs — GORM Embedded SQLite via VFS

Opens a SQLite database embedded inside the Go binary using modernc.org/sqlite/vfs and Go's embed.FS. The embedded file must reside at config/sqlite.db inside the provided embed.FS.

Import this sub-package only when distributing a SQLite database as part of the binary. For regular file-based databases use dbgorm/sqlite instead.

InitDefault uses sync.Once — the first call wins; subsequent calls are silently ignored.

import (
    dbgorm "github.com/phcp-tech/common-library-golang/dbgorm"
    sqlvfs "github.com/phcp-tech/common-library-golang/dbgorm/sqlite/vfs"
)

//go:embed config/sqlite.db
var sqliteFS embed.FS

// Open directly (no singleton).
db, err := sqlvfs.New(&sqliteFS)

// Or store as the process-wide default (sync.Once).
if err := sqlvfs.InitDefault(&sqliteFS); err != nil {
    log.Fatal(err)
}
db := dbgorm.Default() // *gorm.DB, ready for GORM operations

See full examples.

---

dbsqlc/mysql — MySQL Connection

Standard database/sql connection for use with sqlc (sql_package: "database/sql"), backed by go-sql-driver/mysql. Connection pool settings (MaxOpenConns, MaxIdleConns, ConnMaxLifetime, ConnMaxIdletime) are configurable via Config with sensible defaults. Implements the singleton pattern via InitDefault / Default.

Note: sql.Open is lazy — no connection is established during InitDefault. Init() always returns nil. Use an explicit PingContext in a PreReady step if an eager connectivity check is required at startup.

import "github.com/phcp-tech/common-library-golang/dbsqlc/mysql"

// Singleton mode: call once at startup.
err := mysql.InitDefault(&mysql.Config{
    Host:     "localhost",
    Port:     "3306",
    Database: "mydb",
    Username: "user",
    Password: "pass",
})
if err != nil {
    log.Fatal(err)
}

db := mysql.Default() // *sql.DB, pass to sqlc Queries

For cases that require multiple connections, use NewMySQL directly instead of the singleton.

Health check

HealthChecker() returns a health.Checker that pings the default connection. Reports StatusUnhealthy when no connection has been initialised or when the ping fails.

import (
    "github.com/phcp-tech/common-library-golang/health"
    "github.com/phcp-tech/common-library-golang/dbsqlc/mysql"
)

results := health.Check(c.Request.Context(), mysql.HealthChecker())
// → []health.Result{{Name: "database", Status: health.StatusHealthy}}

See full examples.

---

dbsqlc/postgres — PostgreSQL Connection Pool

pgx/v5 connection pool for use with sqlc (sql_package: "pgx/v5"). Pool creation is lazy: NewPostgres returns immediately without establishing any connections, so no live server is required at startup. Implements the singleton pattern via InitDefault / Default.

import "github.com/phcp-tech/common-library-golang/dbsqlc/postgres"

// Singleton mode: call once at startup.
err := postgres.InitDefault(&postgres.Config{
    Host:            "localhost",
    Port:            "5432",
    Database:        "mydb",
    Username:        "user",
    Password:        "pass",
    MaxOpenConns:    100,
    MaxIdleConns:    25,
    ConnMaxLifetime: 60, // minutes
    ConnMaxIdletime: 10, // minutes
    SearchPath:      "myschema", // optional
})
if err != nil {
    log.Fatal(err)
}

pool := postgres.Default() // *pgxpool.Pool, pass to sqlc Queries

For cases that require multiple pools, use NewPostgres directly instead of the singleton.

Health check

HealthChecker() returns a health.Checker that pings the default pool. Reports StatusUnhealthy when no pool has been initialised or when the ping fails.

import (
    "github.com/phcp-tech/common-library-golang/health"
    db "github.com/phcp-tech/common-library-golang/dbsqlc/postgres"
)

results := health.Check(c.Request.Context(), db.HealthChecker())
// → []health.Result{{Name: "postgres", Status: health.StatusHealthy}}

See full examples.

---

dbsqlc/sqlite — SQLite Connection

Pure-Go SQLite driver (modernc.org/sqlite, no CGO required). For use with sqlc (sql_package: "database/sql"). Implements the singleton pattern via InitDefault / Default.

import "github.com/phcp-tech/common-library-golang/dbsqlc/sqlite"

// In-memory database (tests and short-lived operations).
db, err := sqlite.NewSQLite(&sqlite.Config{Path: ":memory:"})

// File-based database with WAL mode and foreign key enforcement.
err := sqlite.InitDefault(&sqlite.Config{
    Path: "file:app.db?_journal_mode=WAL&_foreign_keys=on",
})
if err != nil {
    log.Fatal(err)
}

db := sqlite.Default() // *sql.DB, pass to sqlc Queries

SQLite allows only one writer at a time. NewSQLite calls SetMaxOpenConns(1) automatically to prevent database is locked errors when WAL mode is not in use.

No HealthChecker: SQLite is embedded — if it fails, the application is already broken. A health endpoint for an embedded database provides no operational value.

See full examples.

---

dbsqlc/sqlite/vfs — Embedded SQLite via VFS

Opens a SQLite database that is embedded inside the Go binary using modernc.org/sqlite/vfs and Go's embed.FS. The database file must reside at config/sqlite.db inside the embedded filesystem.

Import this sub-package only when the database is distributed as part of the binary. For regular file-based databases use dbsqlc/sqlite instead.

import sqlitevfs "github.com/phcp-tech/common-library-golang/dbsqlc/sqlite/vfs"

//go:embed config/sqlite.db
var sqliteFS embed.FS

// Singleton mode: call once at startup.
if err := sqlitevfs.InitDefault(&sqliteFS); err != nil {
    log.Fatal(err)
}

db := sqlitevfs.Default() // *sql.DB, pass to sqlc-generated Queries

For cases that require multiple VFS connections, use New directly instead of the singleton.

See full examples.

---

dbsqlx/clickhouse — sqlx ClickHouse

ClickHouse adapter built on github.com/vinovest/sqlx, via clickhouse-go/v2's database/sql driver (registers as "clickhouse") — the same driver mode used internally by gorm.io/driver/clickhouse. dbsqlx.Open pings the server eagerly — InitDefault returns a non-nil error immediately when the database is unreachable, which causes bootstrap to abort startup. The default instance is a process-wide *sqlx.DB shared via dbsqlx.Default() / dbsqlx.SetDefault().

import (
    "github.com/phcp-tech/common-library-golang/dbsqlx"
    "github.com/phcp-tech/common-library-golang/dbsqlx/clickhouse"
)

err := clickhouse.InitDefault(&clickhouse.Config{
    Host:     "localhost",
    Port:     "9440", // native TCP+TLS
    Database: "mydb",
    Username: "user",
    Password: "pass",
})
if err != nil {
    log.Fatal(err) // server was unreachable — Open pinged eagerly
}

db := dbsqlx.Default() // *sqlx.DB, ready for Get/Select/Exec

Health check

Use dbsqlx.HealthChecker() from the root package — the checker is dialect-agnostic and works with any driver initialised via InitDefault.

import (
    "github.com/phcp-tech/common-library-golang/dbsqlx"
    "github.com/phcp-tech/common-library-golang/health"
)

results := health.Check(c.Request.Context(), dbsqlx.HealthChecker())
// → []health.Result{{Name: "database", Status: health.StatusHealthy}}

See full examples.

---

dbsqlx/clickhouse-native — ClickHouse Native Driver

Native TCP ClickHouse client backed by clickhouse-go/v2. Returns driver.Conn — not *sqlx.DB — so it does not integrate with dbsqlx.Open/Default/Exec/Transact from the parent dbsqlx package. Use this package for direct, high-performance native protocol access; use dbsqlx/clickhouse instead when you want the shared *sqlx.DB conveniences.

Lazy open: clickhousenative.NewClickHouse only configures the connection; the TCP dial happens on the first operation (Ping/Query/etc.). InitDefault therefore always returns nil. Use a PreReady step with conn.Ping(ctx) for an eager check.

import "github.com/phcp-tech/common-library-golang/dbsqlx/clickhouse-native"

// Singleton mode — call once at startup.
err := clickhousenative.InitDefault(&clickhousenative.Config{
    Host:     "localhost",
    Port:     "9440", // native TCP+TLS
    Database: "mydb",
    Username: "user",
    Password: "pass",
})
if err != nil {
    log.Fatal(err)
}

conn := clickhousenative.Default() // driver.Conn, ready to use

Health check

HealthChecker() returns a health.Checker that calls conn.Ping(ctx) on the default client. Reports StatusUnhealthy when no client has been initialised or when the ping fails.

import (
    "github.com/phcp-tech/common-library-golang/dbsqlx/clickhouse-native"
    "github.com/phcp-tech/common-library-golang/health"
)

results := health.Check(c.Request.Context(), clickhousenative.HealthChecker())
// → []health.Result{{Name: "clickhouse-native", Status: health.StatusHealthy}}

See full examples.

---

dbsqlx/mysql — sqlx MySQL

MySQL adapter built on github.com/vinovest/sqlx (a database/sql extension with reflection-based struct scanning and Go generics), via the go-sql-driver/mysql driver. Unlike dbsqlc/mysql, dbsqlx.Open pings the server eagerly — InitDefault returns a non-nil error immediately when the database is unreachable, which causes bootstrap to abort startup. The default instance is a process-wide *sqlx.DB shared via dbsqlx.Default() / dbsqlx.SetDefault() (no sync.Once — can be replaced at runtime).

import (
    "github.com/phcp-tech/common-library-golang/dbsqlx"
    "github.com/phcp-tech/common-library-golang/dbsqlx/mysql"
)

// Open and store as the process-wide default.
err := mysql.InitDefault(&mysql.Config{
    Host:     "localhost",
    Port:     "3306",
    Database: "mydb",
    Username: "user",
    Password: "pass",
})
if err != nil {
    log.Fatal(err) // server was unreachable — Open pinged eagerly
}

db := dbsqlx.Default() // *sqlx.DB, ready for Get/Select/Exec

For cases that require multiple databases, use NewMySQL directly and store the returned *sqlx.DB in explicit fields rather than the shared default.

Health check

Use dbsqlx.HealthChecker() from the root package — the checker is dialect-agnostic and works with any driver initialised via InitDefault.

import (
    "github.com/phcp-tech/common-library-golang/dbsqlx"
    "github.com/phcp-tech/common-library-golang/health"
)

results := health.Check(c.Request.Context(), dbsqlx.HealthChecker())
// → []health.Result{{Name: "database", Status: health.StatusHealthy}}

See full examples.

---

dbsqlx/postgres — sqlx PostgreSQL

PostgreSQL adapter built on github.com/vinovest/sqlx, via the pgx/v5/stdlib driver (registers as "pgx" with database/sql). dbsqlx.Open pings the server and runs SHOW search_path eagerly — InitDefault returns a non-nil error immediately when the database is unreachable. Supports an optional SearchPath for schema isolation. The default instance is a process-wide *sqlx.DB shared via dbsqlx.Default() / dbsqlx.SetDefault().

import (
    "github.com/phcp-tech/common-library-golang/dbsqlx"
    "github.com/phcp-tech/common-library-golang/dbsqlx/postgres"
)

err := postgres.InitDefault(&postgres.Config{
    Host:       "localhost",
    Port:       "5432",
    Database:   "mydb",
    Username:   "user",
    Password:   "pass",
    SearchPath: "myschema", // optional
})
if err != nil {
    log.Fatal(err) // server was unreachable
}

db := dbsqlx.Default() // *sqlx.DB, ready for Get/Select/Exec

Health check

Use dbsqlx.HealthChecker() from the root package — the checker is dialect-agnostic and works with any driver initialised via InitDefault.

import (
    "github.com/phcp-tech/common-library-golang/dbsqlx"
    "github.com/phcp-tech/common-library-golang/health"
)

results := health.Check(c.Request.Context(), dbsqlx.HealthChecker())
// → []health.Result{{Name: "database", Status: health.StatusHealthy}}

See full examples.

---

dbsqlx/sqlite — sqlx SQLite

SQLite adapter built on github.com/vinovest/sqlx, via the pure-Go modernc.org/sqlite driver (no CGO). SQLite is an embedded database — no network connection is involved. InitDefault opens (and auto-creates) the file and its parent directory; it fails only when the path is empty or the file cannot be created. Every connection gets the standard PRAGMA settings (WAL, foreign keys, busy timeout, 32 MB cache). MaxOpenConns is fixed at 4 / MaxIdleConns at 2 — SQLite allows only one writer at a time, so a larger pool provides no benefit even under WAL.

import (
    "github.com/phcp-tech/common-library-golang/dbsqlx"
    "github.com/phcp-tech/common-library-golang/dbsqlx/sqlite"
)

// File-based database.
err := sqlite.InitDefault(&sqlite.Config{
    Path: "file:app.db?cache=shared",
})
if err != nil {
    log.Fatal(err)
}

db := dbsqlx.Default() // *sqlx.DB, ready for Get/Select/Exec

// In-memory database (tests, ephemeral data).
db, _ = sqlite.NewSQLite(&sqlite.Config{
    Path: "file::memory:?cache=shared",
})

No HealthChecker: SQLite is embedded — if it fails, the application is already broken. A health endpoint for an embedded database provides no operational value.

See full examples.

---

dbsqlx/sqlite/vfs — sqlx Embedded SQLite via VFS

Opens a SQLite database embedded inside the Go binary using modernc.org/sqlite/vfs and Go's embed.FS. The embedded file must reside at config/sqlite.db inside the provided embed.FS.

Import this sub-package only when distributing a SQLite database as part of the binary. For regular file-based databases use dbsqlx/sqlite instead.

InitDefault uses sync.Once — the first call wins; subsequent calls are silently ignored.

import (
    "github.com/phcp-tech/common-library-golang/dbsqlx"
    sqlvfs "github.com/phcp-tech/common-library-golang/dbsqlx/sqlite/vfs"
)

//go:embed config/sqlite.db
var sqliteFS embed.FS

// Open directly (no singleton).
db, err := sqlvfs.New(&sqliteFS)

// Or store as the process-wide default (sync.Once).
if err := sqlvfs.InitDefault(&sqliteFS); err != nil {
    log.Fatal(err)
}
db = dbsqlx.Default() // *sqlx.DB, ready for Get/Select/Exec

See full examples.

---

network — Network Utilities

Helpers for request IP extraction, local interface inspection, and IPv4 ↔ uint32 conversion.

IP address helpers

import "github.com/phcp-tech/common-library-golang/network"

// Real client IP — inspects X-Forwarded-For → X-Real-IP → RemoteAddr.
// Returns the first address in X-Forwarded-For when multiple proxies are chained.
// "::1" is normalised to "127.0.0.1".
ip := network.GetRemoteIp(req)

// All non-loopback IPv4 addresses on the local machine.
addrs := network.GetLocalIpAddress()

IPv4 ↔ uint32 conversion

Two byte orders are supported:

Function	Byte order	Typical use
Ip2IntWithBigEndian	Big-endian (network order)	Standard TCP/IP, databases
Int2IpWithBigEndian	Big-endian	Same as above
Ip2IntWithLittleEndian	Little-endian	MT4 / MT5 trading platforms
Int2IpWithLittleEndian	Little-endian	Same as above

// Big-endian (network byte order): "1.2.3.4" → 0x01020304
n := network.Ip2IntWithBigEndian("1.2.3.4")   // → 16909060
ip := network.Int2IpWithBigEndian(0x01020304)  // → "1.2.3.4"

// Little-endian (MT4/MT5 format): "1.2.3.4" → 0x04030201
n := network.Ip2IntWithLittleEndian("1.2.3.4")   // → 67305985
ip := network.Int2IpWithLittleEndian(0x04030201)  // → "1.2.3.4"

// Both return 0 / "" for empty, invalid, or IPv6 input.

Address validation

// IsValidAddr checks if the string is a valid IP address (with or without port)
// or a resolvable hostname.
network.IsValidAddr("192.168.1.1:8080") // true
network.IsValidAddr("192.168.1.1")      // true
network.IsValidAddr("example.com:443")  // true (DNS lookup)
network.IsValidAddr("999.999.999.999")  // false

See full examples.

---

auth — Casbin RBAC Authorisation

Role-based access control (RBAC) Gin middleware backed by Casbin. Enforces policy using method-2 semantics: every role in the user's role list must pass the policy check — if any role is denied, the request is rejected with HTTP 403 Forbidden.

InitCasbin must be called once at application startup before Authorize is registered. Model and policy can be loaded from in-memory strings (fs=true) or from files on disk (fs=false).

import "github.com/phcp-tech/common-library-golang/auth"

// Load model and policy from in-memory strings (e.g. embedded at build time).
if err := auth.InitCasbin(true, modelString, policyString); err != nil {
    log.Fatal(err)
}

// Load model and policy from files.
if err := auth.InitCasbin(false, "model.conf", "policy.csv"); err != nil {
    log.Fatal(err)
}

// Register as a Gin middleware after token.Authenticate.
// It reads roles from the "userInfo" context key set by token.Authenticate.
r := gin.New()
r.Use(token.Authenticate())
r.Use(auth.Authorize())

See full examples.

---

token — JWT Authentication

HS256 JWT access and refresh token creation, parsing, and a Gin bearer-token middleware. Built on golang-jwt/jwt.

InitToken must be called once at application startup before any token function. The secrets and issuer are typically read from env.Env() after env.InitEnv().

import "github.com/phcp-tech/common-library-golang/token"

// Initialise once at startup (composition root), after env.InitEnv().
token.InitToken(
    env.Env().String("jwt.issuer"),             // e.g. "phcp"
    env.Env().String("jwt.access.secretcode"),  // access token signing key
    env.Env().String("jwt.refresh.secretcode"), // refresh token signing key
)

// Create a short-lived access token (valid for 1 hour).
tok, err := token.CreateToken(userId, username, productId, roles, time.Hour)

// Validate and parse an access token from an incoming request.
user, err := token.ParseToken(tok)
fmt.Println(user.Username, user.UserId, user.Roles)

// Create a long-lived refresh token (no roles embedded).
refresh, err := token.CreateRefreshToken(userId, username, productId, 24*time.Hour)

// Register as a Gin middleware; stores LoginUser in context under key "userInfo".
r := gin.New()
r.Use(token.Authenticate())

Authenticate aborts with HTTP 401 if the Authorization: Bearer <token> header is missing, malformed, or carries an invalid/expired token.

See full examples.

---

gin — Gin Engine Factory

Pre-configured Gin engine with structured slog-based request logging (slog-gin) and optional CORS support. Always runs in ReleaseMode.

InitGin is the single entry point. Pass a slice of allowed origins to enable CORS; entries containing * are treated as single-level wildcard patterns (e.g. https://*.example.com matches https://api.example.com but not https://a.b.example.com).

import libgin "github.com/phcp-tech/common-library-golang/gin"

// No CORS.
router := libgin.InitGin(nil)

// Exact origins and wildcard patterns can be mixed.
router := libgin.InitGin([]string{
    "https://app.example.com",   // exact
    "https://*.example.com",     // wildcard: root + any single subdomain
})

See full examples.

---

gin/pprof — Profiling Endpoints

Registers Go runtime pprof endpoints on a Gin engine via gin-contrib/pprof.

Import this sub-package only in services that require profiling. Importing it is an explicit opt-in — it does not affect services that only import gin.

Two route groups are mounted:

Path	Purpose
/debug/pprof/*	Standard Go pprof path for direct access
<path>/admin/pprof/*	API-gateway-friendly alias

import (
    libgin   "github.com/phcp-tech/common-library-golang/gin"
    ginpprof "github.com/phcp-tech/common-library-golang/gin/pprof"
)

router := libgin.InitGin(nil)
ginpprof.Mount(router, "/api/v1") // mounts /debug/pprof/* and /api/v1/admin/pprof/*

See full examples.

---

httpclient — Resty HTTP Client

Feature-rich HTTP client built on go-resty/resty with pre-configured retry, JWT bearer-token support, and automatic JSON handling. Suitable for structured service-to-service calls.

import "github.com/phcp-tech/common-library-golang/httpclient"

// Default settings: Timeout=10s, RetryMax=3.
cli := httpclient.NewHttpClient()

// Custom settings — zero-value fields fall back to defaults.
cli := httpclient.NewHttpClient(httpclient.Config{
    Timeout:            15 * time.Second,
    RetryMax:           5,
    InsecureSkipVerify: true, // only for internal self-signed certs
})

resp, err := cli.Get(url, jwtToken, nil)
resp, err := cli.Post(url, jwtToken, body)
resp, err := cli.Put(url, jwtToken, body)
resp, err := cli.Delete(url, jwtToken, body)

// Access the underlying resty.Client for advanced use.
restyClient := cli.Client()

See full examples.

---

httpclient/retryable — Retryable HTTP Client

Wraps hashicorp/go-retryablehttp and exposes it as a standard *http.Client with configurable retry and timeout. Use this sub-package when existing code already uses net/http directly and you need to add retry behaviour without switching to resty.

Import this sub-package only when needed — it pulls in the hashicorp library independently of the parent httpclient package.

import "github.com/phcp-tech/common-library-golang/httpclient/retryable"

cli := retryable.NewHttpClient(retryable.Config{
    Timeout:  15 * time.Second,
    RetryMax: 5,
})

// Obtain a standard *http.Client with retry built in.
stdClient := cli.Client().StandardClient()
resp, err := stdClient.Do(req)

See full examples.

---

httpserver — Production HTTP/HTTPS Server

A production-ready HTTP/HTTPS server built on net/http with secure defaults. It wraps http.Server directly (not gin.Run()) to provide:

• Configurable timeouts — ReadTimeout, WriteTimeout, IdleTimeout, ReadHeaderTimeout
• TLS 1.2+ with strong cipher suites — activated when CrtFile and KeyFile are set
• Graceful shutdown — Shutdown(ctx) drains in-flight requests before stopping

The IRunner interface unifies HTTP and Lambda modes so the composition root selects the backend and the rest of the application code is unchanged.

import "github.com/phcp-tech/common-library-golang/httpserver"

// Plain HTTP with package defaults.
runner := httpserver.NewHttpServer(httpserver.Config{Port: "8080"})

// HTTPS with custom timeouts.
runner := httpserver.NewHttpServer(httpserver.Config{
    Port:        "8443",
    CrtFile:     "/etc/ssl/server.crt",
    KeyFile:     "/etc/ssl/server.key",
    ReadTimeout: 15 * time.Second,
    WriteTimeout: 0, // 0 = unlimited (required for file downloads)
})

// Start in a goroutine; block until OS signal, then shut down gracefully.
go func() { _ = runner.Start(ginRouter) }()
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
defer cancel()
_ = runner.Shutdown(ctx)

See full examples.

---

httpserver/lambda — AWS Lambda Adapter

Implements httpserver.IRunner for AWS Lambda via aws-lambda-go-api-proxy. It bridges APIGatewayProxyRequest events to any http.Handler (including *gin.Engine) with no handler changes required.

Import this sub-package only for Lambda deployments — it pulls in the AWS Lambda SDK. Services that run as plain HTTP servers are unaffected.

import (
    "github.com/phcp-tech/common-library-golang/httpserver"
    lambdarunner "github.com/phcp-tech/common-library-golang/httpserver/lambda"
)

// Composition root selects the runner based on the deployment environment.
var runner httpserver.IRunner
if isLambda {
    runner = lambdarunner.NewHttpServer()
} else {
    runner = httpserver.NewHttpServer(httpserver.Config{Port: port})
}
_ = runner.Start(ginRouter) // same call regardless of mode

See full examples.