go-migrate-lite

A tiny, dependency-light SQL migration runner for Postgres, written in Go.

It runs *.sql files in lexical order, tracks applied versions in schema_migrations, and works both as a CLI tool and as a Go library (including with embed.FS).

---

Features

• Runs migrations in filename order (e.g., 001_init.sql, 002_add_index.sql).
• Tracks applied versions in schema_migrations — safe to run repeatedly.
• Accepts any fs.FS: filesystem path, embed.FS, or custom.
• Works with PostgreSQL and TimescaleDB.

---

CLI

Run directly

export DATABASE_URL="postgres://user:pass@localhost:5432/mydb?sslmode=disable"
export MIGRATIONS_DIR=./migrations

go run github.com/jsness/go-migrate-lite/cmd/migrate-lite@latest

Build binary

go install github.com/jsness/go-migrate-lite/cmd/migrate-lite@latest
migrate-lite

Environment

Variable	Required	Default	Description
DATABASE_URL	yes	—	Postgres connection string
MIGRATIONS_DIR	no	migrations	Path to migration files

---

Library

Import the package to run migrations at application startup — useful with embed.FS so no external CLI or migration directory is needed at runtime.

import (
    "embed"

    migratelite "github.com/jsness/go-migrate-lite"
)

//go:embed migrations/*.sql
var migrationsFS embed.FS

// In main, after connecting to the database:
if err := migratelite.Run(ctx, pool, migrationsFS); err != nil {
    log.Fatalf("migrations: %v", err)
}

Run applies any pending .sql files from the given fs.FS in lexical order and records each applied version in schema_migrations. It is idempotent — already-applied versions are skipped.

The version key is the numeric prefix of the filename (e.g. "001" from 001_init.sql), matching the CLI, so both can run against the same database without conflict.

---

Migration Rules

• Filenames must start with a numeric version prefix followed by an underscore (e.g. 001_init.sql).
• Files are applied in lexical order.
• Each migration is applied exactly once.

migrations/
  001_init.sql
  002_add_index.sql

---

License

MIT