embeddedpostgres

embedded-postgres v2

Run real PostgreSQL in your tests. Each test gets its own database, an available port, and automatic cleanup. A CLI provides the same lifecycle for other languages. The production Go packages have zero third-party dependencies.

v2 is available as an alpha release. Requires Go 1.26 or newer. Upgrading from v1? Start with the migration guide.

A database scoped to a test

go get github.com/fergusstrange/embedded-postgres/v2@v2.0.0-alpha.1

package app_test

import (
    "testing"

    "github.com/fergusstrange/embedded-postgres/v2/eptest"
)

func TestRepository(t *testing.T) {
    pg := eptest.Start(t)
    connectionURL := pg.ConnectionURL()
    _ = connectionURL // Open your chosen Go driver/pool here.
    // Register pool cleanup after eptest.Start, so the pool closes first.
}

The first start downloads PostgreSQL and the matching supervisor; later starts reuse the cache. PostgreSQL needs the native libraries for your OS. Choose your own SQL driver, and close connection pools before test cleanup.

For configuration, import github.com/fergusstrange/embedded-postgres/v2 as postgres and pass postgres.DefaultConfig().Port(0).Database("app_test") to eptest.Start. Use eptest.Run to share one instance across TestMain. See testing patterns for parallel tests and suite setup. Source checkouts, local replacements and vendored tests need an explicitly selected supervisor; see the migration guide.

Use the CLI with any test runner

curl -fsSL https://raw.githubusercontent.com/fergusstrange/embedded-postgres/master/install/install.sh | bash
export PATH="$HOME/.local/bin:$PATH"
embedded-postgres exec -- npm test

exec starts PostgreSQL, supplies DATABASE_URL and the usual PG* variables, runs your command, then stops the database and returns the command's exit code. Replace npm test with your test command. The CLI chooses an available port and a random password. See the CLI guide for foreground and background instances, Windows installation, configuration files and wrapper integration.

Direct lifecycle use

pg := postgres.NewDatabase(postgres.DefaultConfig().Port(0))
if err := pg.StartContext(ctx); err != nil {
    return err
}
defer pg.Close()
// pg.ConnectionURL() and pg.GetPort() now contain the actual endpoint.

Canceling the lifetime context also closes the server. Close is idempotent; Start and Stop preserve the familiar v1 calling pattern. StopContext can stop waiting when its context expires while cleanup continues. Done closes when cleanup finishes, Err reports a lifecycle failure, and Logs retains a bounded, password-redacted tail for diagnostics.

An independent supervisor watches a dedicated owner pipe. An unrecovered panic, os.Exit, or forced owner termination closes that pipe and triggers PostgreSQL shutdown. The library installs no process-wide signal handlers. Userspace cannot guarantee graceful shutdown after whole-machine failure or simultaneous forced termination of the supervisor and PostgreSQL. Persistent data still requires the same care as any PostgreSQL cluster.

Storage and extension points

config := postgres.DefaultConfig().Port(0).Storage(postgres.Storage{
    CacheDir: "/path/to/shared-cache",
    WorkDir:  "/path/to/temporary-parent",
    DataDir:  "/path/to/persistent-cluster", // Omit for disposable test data.
})

• CacheDir holds verified, immutable installations shared under process leases. Stopping an instance retains reusable binaries. PruneCache removes unused recognized entries and preserves active installations and unknown files.
• WorkDir is a parent for private, uniquely named temporary children. The caller's parent is never erased. Each normal lifecycle removes its own child.
• DataDir is persistent and never automatically removed. A nonempty unrelated directory or a cluster from a different PostgreSQL major is rejected.

Replace the whole acquisition policy with BinaryProvider, use LocalProvider for preinstalled PostgreSQL, or customize DownloadProvider for mirrors, HTTP transports and offline caches. Lifecycle hooks support preparation and migrations with reverse-order cleanup. Native extensions such as pgvector/PostGIS can live in caller-owned distributions without becoming core dependencies. See provider and hook contracts and compiled reference implementations.

Root callers must select an existing non-root Unix identity with RunAs(postgres.User{UID: 10001, GID: 10001}). The caller retains its identity; PostgreSQL and its utilities run as the selected account. Windows uses a restricted token and process Job Objects. Follow the non-root reference guide for container setup, directory ownership, and native library requirements.

Defaults and supported versions

V15, V16, V17, V18 pin 15.19.0, 16.15.0, 17.11.0, 18.6.0. PostgreSQL 14 and older are excluded. PostgreSQL uses a five-year major-version support policy, not a separate LTS edition. Binary checksums are pinned in the reviewed manifest; automatic library releases do not silently change those pins.

Native CI covers Linux and macOS AMD64/ARM64, Windows AMD64, and Windows ARM64 with x64 PostgreSQL emulation. Alpine 3.21 AMD64 has a separate musl integration job (the upstream bundle requires ICU 74; see the platform guide). Other Linux architectures can use a source-built companion when a pinned bundle exists. FreeBSD needs a custom distribution and companion; it is not in the supported release matrix.

Development

Contributing explains local tests and coverage. Build and release policy describes the native GitHub Actions matrix, 90% production coverage gate, security checks and automatic master releases. Download published binaries and read release notes on the releases page.

Thanks to theseus-rs/postgresql-binaries for gzip distributions, the PostgreSQL project, and zonkyio/embedded-postgres-binaries for powering v1. The logo uses Gopherize Me.