atlas-best-practices（阿特拉斯最佳实践）

Atlas Best Practices

Atlas is a language-independent tool for managing database schemas using declarative or versioned workflows.

Two Workflows

Declarative (Terraform-like): Atlas compares current vs desired state and generates migrations automatically.

atlas schema apply --url "postgres://..." --to "file://schema.hcl" --dev-url "docker://postgres/15"

Versioned: Atlas generates migration files from schema changes, stored in version control.

atlas migrate diff add_users --dir "file://migrations" --to "file://schema.sql" --dev-url "docker://postgres/15"
atlas migrate apply --dir "file://migrations" --url "postgres://..."

Dev Database

Atlas requires a dev database for schema validation, diffing, and linting. Use the docker driver for ephemeral containers:

# PostgreSQL
--dev-url "docker://postgres/15/dev?search_path=public"

# MySQL
--dev-url "docker://mysql/8/dev"

# SQLite
--dev-url "sqlite://dev?mode=memory"

Schema-as-Code

HCL Schema (Recommended)

Use database-specific file extensions for editor support: .pg.hcl (PostgreSQL), .my.hcl (MySQL), .lt.hcl (SQLite).

schema "public" {
  comment = "Application schema"
}

table "users" {
  schema = schema.public
  column "id" {
    type = bigint
  }
  column "email" {
    type = varchar(255)
    null = false
  }
  column "created_at" {
    type    = timestamptz
    default = sql("now()")
  }
  primary_key {
    columns = [column.id]
  }
  index "idx_users_email" {
    columns = [column.email]
    unique  = true
  }
}

table "orders" {
  schema = schema.public
  column "id" {
    type = bigint
  }
  column "user_id" {
    type = bigint
    null = false
  }
  column "total" {
    type = numeric
    null = false
  }
  foreign_key "fk_user" {
    columns     = [column.user_id]
    ref_columns = [table.users.column.id]
    on_delete   = CASCADE
  }
  check "positive_total" {
    expr = "total > 0"
  }
}

SQL Schema

Use standard SQL DDL files:

CREATE TABLE "users" (
  "id" bigint PRIMARY KEY,
  "email" varchar(255) NOT NULL UNIQUE,
  "created_at" timestamptz DEFAULT now()
);

Project Configuration

Create atlas.hcl for environment configuration:

variable "db_url" {
  type = string
}

env "local" {
  src = "file://schema.pg.hcl"
  url = var.db_url
  dev = "docker://postgres/15/dev?search_path=public"

  migration {
    dir = "file://migrations"
  }

  format {
    migrate {
      diff = "{{ sql . \"  \" }}"
    }
  }
}

env "prod" {
  src = "file://schema.pg.hcl"
  url = var.db_url

  migration {
    dir = "atlas://myapp"  # Atlas Registry
  }
}

Run with environment:

atlas schema apply --env local --var "db_url=postgres://..."

Migration Linting

Atlas analyzes migrations for safety. Configure in atlas.hcl:

lint {
  destructive {
    error = true  # Fail on DROP TABLE/COLUMN
  }
  data_depend {
    error = true  # Fail on data-dependent changes
  }
  naming {
    match   = "^[a-z_]+$"
    message = "must be lowercase with underscores"
    index {
      match   = "^idx_"
      message = "indexes must start with idx_"
    }
  }
  # PostgreSQL: require CONCURRENTLY for indexes (Pro)
  concurrent_index {
    error = true
  }
}

Key analyzers:

• DS: Destructive changes (DROP SCHEMA/TABLE/COLUMN)
• MF: Data-dependent changes (ADD UNIQUE, NOT NULL)
• BC: Backward incompatible (rename table/column)
• PG (Pro): Concurrent index, blocking DDL

Lint migrations:

atlas migrate lint --env local --latest 1

Suppress specific checks in migration files:

-- atlas:nolint destructive
DROP TABLE old_users;

Schema Testing

Write tests in .test.hcl files:

test "schema" "user_constraints" {
  parallel = true

  exec {
    sql = "INSERT INTO users (id, email) VALUES (1, 'test@example.com')"
  }

  # Test unique constraint
  catch {
    sql   = "INSERT INTO users (id, email) VALUES (2, 'test@example.com')"
    error = "duplicate key"
  }

  assert {
    sql = "SELECT COUNT(*) = 1 FROM users"
    error_message = "expected exactly one user"
  }

  cleanup {
    sql = "DELETE FROM users"
  }
}

# Table-driven tests
test "schema" "email_validation" {
  for_each = [
    {input: "valid@test.com", valid: true},
    {input: "invalid",        valid: false},
  ]

  exec {
    sql    = "SELECT validate_email('${each.value.input}')"
    output = each.value.valid ? "t" : "f"
  }
}

Run tests:

atlas schema test --env local schema.test.hcl

Transaction Modes

Control transaction behavior per-file with directives:

-- atlas:txmode none
CREATE INDEX CONCURRENTLY idx_users_email ON users (email);

Modes: file (default, one tx per file), all (one tx for all), none (no tx).

Pre-Execution Checks (Pro)

Block dangerous operations in atlas.hcl (requires Atlas Pro):

env "prod" {
  check "migrate_apply" {
    deny "too_many_files" {
      condition = length(self.planned_migration.files) > 3
      message   = "Cannot apply more than 3 migrations at once"
    }
  }
}

Common Commands

# Generate migration from schema diff
atlas migrate diff migration_name --env local

# Apply pending migrations
atlas migrate apply --env local

# Validate migration directory integrity
atlas migrate validate --env local

# View migration status
atlas migrate status --env local

# Push to Atlas Registry
atlas migrate push myapp --env local

# Declarative apply (no migration files)
atlas schema apply --env local --auto-approve

# Inspect current database schema
atlas schema inspect --url "postgres://..." --format "{{ sql . }}"

# Compare schemas
atlas schema diff --from "postgres://..." --to "file://schema.hcl"

CI/CD Integration

GitHub Actions setup:

- uses: ariga/setup-atlas@v0
  with:
    cloud-token: ${{ secrets.ATLAS_CLOUD_TOKEN }}

- name: Lint migrations
  run: atlas migrate lint --env ci --git-base origin/main

Baseline for Existing Databases

When adopting Atlas on existing databases:

# Create baseline migration reflecting current schema
atlas migrate diff baseline --env local --to "file://schema.hcl"

# Mark baseline as applied (skip execution)
atlas migrate apply --env prod --baseline "20240101000000"

ORM Integration

Atlas supports loading schemas from ORMs via external providers:

data "external_schema" "gorm" {
  program = [
    "go", "run", "-mod=mod",
    "ariga.io/atlas-provider-gorm",
    "load", "--path", "./models",
    "--dialect", "postgres",
  ]
}

env "local" {
  src = data.external_schema.gorm.url
}

Supported: GORM, Sequelize, TypeORM, Django, SQLAlchemy, Prisma, and more.

Instructions

• Always use a dev database for migrate diff and schema apply; it validates schemas safely.
• Enable strict linting in CI to catch destructive and data-dependent changes early.
• Use versioned migrations for production; declarative workflow suits development/testing.
• Test schemas with .test.hcl files; validate constraints, triggers, and functions.
• Push migrations to Atlas Registry for deployment; avoid copying files manually.
• Use -- atlas:txmode none for PostgreSQL concurrent index operations.
• Configure naming conventions in lint rules; consistency prevents errors.