Schema builder

s := schema.New(db.DB, db.Driver)
err := s.Create("users", func(table *schema.Blueprint) {
    table.ID()
    table.String("name", 255)
    table.String("email", 255).Unique()
    table.Timestamps()
    table.SoftDeletes()
})

Types: ID, String, Text, Integer, BigInteger, Boolean, Float, Decimal, Date, DateTime, Timestamp, JSON, UUID and Binary. Columns default to NOT NULL. Modifiers: Nullable, NotNull, Default, Unique, Primary, AutoIncrement and Unsigned (MySQL only). Defaults are escaped literals or schema.CurrentTimestamp. Invalid types, lengths, precision or unsupported options return errors.

err = s.Table("users", func(table *schema.Blueprint) {
    table.String("phone", 50).Nullable()
    table.String("name", 300).Change()
    table.Index("email", "phone").Named("users_contact_index")
})

Table supports DropColumn, RenameColumn, DropIndex and DropForeign. Index, Unique and Primary accept composite lists; generated names are deterministic, shortened for server identifier limits, or customized with Named. Foreign(...).References(...). On(...).OnDelete(...).OnUpdate(...) defines foreign keys. ForeignID("user_id"). Constrained("users").CascadeOnDelete() also adds the integer column. Supported actions: cascade, restrict, set null, no action and set default (not MySQL). Match types and nullability with referenced keys/actions.

Operations include Drop, DropIfExists, Rename, HasTable and HasColumn. NewContext supports cancellation. Compilers quote identifiers and use dialect-specific identity, JSON, UUID and binary types; introspection binds values. PostgreSQL/MySQL/SQLite compiler constructors are available under schema/compilers.

SQLite changes requiring new constraints or column modifications rebuild the table using its documented procedure, preserving rows and indexes/triggers and checking foreign keys before commit. A rebuild inside a foreign-key-enabled user transaction is explicitly rejected; run it outside that transaction or use the migrator. Native drop/rename restrictions apply: drop dependent indexes/objects before dropping their columns. Advanced raw SQL definitions and dependent views need explicit review and migration code.

Create/native alter execute transactionally with a *sql.DB on SQLite/PostgreSQL. MySQL DDL atomicity is server-defined; migrations retain partial-failure markers.