class Sequel::Schema::CreateTableGenerator

  1. lib/sequel/database/schema_generator.rb
Superclass: Object

Schema::CreateTableGenerator is an internal class that the user is not expected to instantiate directly. Instances are created by Sequel::Database#create_table. It is used to specify table creation parameters. It takes a Database object and a block of column/index/constraint specifications, and gives the Database a table description, which the database uses to create a table.

Schema::CreateTableGenerator has some methods but also includes #method_missing, allowing users to specify column type as a method instead of using the column method, which makes for a nicer DSL.

For more information on Sequel's support for schema modification, see the “Schema Modification” guide.


GENERIC_TYPES = %w'String Integer Float Numeric BigDecimal Date DateTime Time File TrueClass FalseClass'.freeze  

Classes specifying generic types that Sequel will convert to database-specific types.


columns [R]

Column hashes created by this generator

constraints [R]

Constraint hashes created by this generator

indexes [R]

Index hashes created by this generator

Public Class methods

add_type_method (*types)

Add a method for each of the given types that creates a column with that type as a constant. Types given should either already be constants/classes or a capitalized string/symbol with the same name as a constant/class.

[show source]
# File lib/sequel/database/schema_generator.rb, line 59
def self.add_type_method(*types)
  types.each do |type|
    case type
    when Symbol, String
      method = type
      type = Object.const_get(type)
      method = type.to_s

    define_method(method){|name, opts=OPTS| column(name, type, opts)}
new (db, &block)

Set the database in which to create the table, and evaluate the block in the context of this object.

[show source]
# File lib/sequel/database/schema_generator.rb, line 33
def initialize(db, &block)
  @db = db
  @columns = []
  @indexes = []
  @constraints = []
  @primary_key = nil
  instance_exec(&block) if block
  @columns.unshift(@primary_key) if @primary_key && !has_column?(primary_key_name)

Public Instance methods

Bignum (name, opts=OPTS)

Use custom #Bignum method to use :Bignum instead of #Bignum class, to work correctly in cases where #Bignum is the same as Integer.

[show source]
# File lib/sequel/database/schema_generator.rb, line 45
def Bignum(name, opts=OPTS)
  column(name, :Bignum, opts)
Fixnum (name, opts=OPTS)

Use custom #Fixnum method to use Integer instead of #Fixnum class, to avoid warnings on ruby 2.4+.

[show source]
# File lib/sequel/database/schema_generator.rb, line 51
def Fixnum(name, opts=OPTS)
  column(name, Integer, opts)
check (*args, &block)

Add an unnamed constraint, specified by the given block or args:

check(num: 1..5) # CHECK num >= 1 AND num <= 5
check{num > 5}   # CHECK num > 5
[show source]
# File lib/sequel/database/schema_generator.rb, line 79
def check(*args, &block)
  constraint(nil, *args, &block)
column (name, type, opts = OPTS)

Add a column with the given name, type, and opts:

column :num, :integer

column :name, String, null: false, default: 'a'
# name varchar(255) NOT NULL DEFAULT 'a'

inet :ip
# ip inet

You can also create columns via method missing, so the following are equivalent:

column :number, :integer
integer :number

The following options are supported:


The collation to use for the column. For backwards compatibility, only symbols and string values are supported, and they are used verbatim. However, on PostgreSQL, symbols are literalized as regular identifiers, since unquoted collations are unlikely to be valid.


The default value for the column.


For foreign key columns, this ensures referential integrity will work even if referencing table uses a foreign key value that does not yet exist on referenced table (but will exist before the transaction commits). Basically it adds DEFERRABLE INITIALLY DEFERRED on key creation. If you use :immediate as the value, uses DEFERRABLE INITIALLY IMMEDIATE.


Create an index on this column. If given a hash, use the hash as the options for the index.


For foreign key columns, the column in the associated table that this column references. Unnecessary if this column references the primary key of the associated table, except if you are using MySQL.


Mark the column as allowing NULL values (if true), or not allowing NULL values (if false). The default is to allow NULL values.


Specify the behavior of this column when being deleted (:restrict, :cascade, :set_null, :set_default, :no_action).


Specify the behavior of this column when being updated (:restrict, :cascade, :set_null, :set_default, :no_action).


Make the column as a single primary key column. This should not be used if you have a single, nonautoincrementing primary key column (use the #primary_key method in that case).


The name to give the primary key constraint


Overrides the type given as the argument. Generally not used by column itself, but can be passed as an option to other methods that call column.


Mark the column as unique, generally has the same effect as creating a unique index on the column.


The name to give the unique key constraint

MySQL specific options:


Specify a GENERATED ALWAYS AS column expression, if generated columns are supported.


Set the type of column when using :generated_always_as, should be :virtual or :stored to force a type.

[show source]
# File lib/sequel/database/schema_generator.rb, line 139
def column(name, type, opts = OPTS)
  columns << {:name => name, :type => type}.merge!(opts)
  if index_opts = opts[:index]
    index(name, index_opts.is_a?(Hash) ? index_opts : OPTS)
constraint (name, *args, &block)

Adds a named constraint (or unnamed if name is nil), with the given block or args. To provide options for the constraint, pass a hash as the first argument.

constraint(:blah, num: 1..5)
# CONSTRAINT blah CHECK num >= 1 AND num <= 5
constraint({name: :blah, deferrable: true}, num: 1..5)
[show source]
# File lib/sequel/database/schema_generator.rb, line 155
def constraint(name, *args, &block)
  opts = name.is_a?(Hash) ? name : {:name=>name}
  constraints << opts.merge(:type=>:check, :check=>block || args)
foreign_key (name, table=nil, opts = OPTS)

Add a foreign key in the table that references another table. See column for available options.

foreign_key(:artist_id) # artist_id INTEGER
foreign_key(:artist_id, :artists) # artist_id INTEGER REFERENCES artists
foreign_key(:artist_id, :artists, key: :id) # artist_id INTEGER REFERENCES artists(id)
foreign_key(:artist_id, :artists, type: String) # artist_id varchar(255) REFERENCES artists(id)

Additional Options:


The name to give the foreign key constraint

If you want a foreign key constraint without adding a column (usually because it is a composite foreign key), you can provide an array of columns as the first argument, and you can provide the :name option to name the constraint:

foreign_key([:artist_name, :artist_location], :artists, name: :artist_fk)
# ADD CONSTRAINT artist_fk FOREIGN KEY (artist_name, artist_location) REFERENCES artists
[show source]
# File lib/sequel/database/schema_generator.rb, line 179
def foreign_key(name, table=nil, opts = OPTS)
  opts = case table
  when Hash
  when NilClass
  return composite_foreign_key(name, opts) if name.is_a?(Array)
  column(name, Integer, opts)
full_text_index (columns, opts = OPTS)

Add a full text index on the given columns.

PostgreSQL specific options:


Can be set to :gist to use a GIST index instead of the default GIN index.


Set a language to use for the index (default: simple).

Microsoft SQL Server specific options:


The KEY INDEX to use for the full text index.

[show source]
# File lib/sequel/database/schema_generator.rb, line 201
def full_text_index(columns, opts = OPTS)
  index(columns, opts.merge(:type => :full_text))
has_column? (name)

True if the generator includes the creation of a column with the given name.

[show source]
# File lib/sequel/database/schema_generator.rb, line 206
def has_column?(name)
  columns.any?{|c| c[:name] == name}
index (columns, opts = OPTS)

Add an index on the given column(s) with the given options. General options:


The name to use for the index. If not given, a default name based on the table and columns is used.


The type of index to use (only supported by some databases)


Make the index unique, so duplicate values are not allowed.


Create a partial index (only supported by some databases)

PostgreSQL specific options:


Create the index concurrently, so it doesn't block operations on the table while the index is being built.


Use a specific operator class in the index.


Include additional column values in the index, without actually indexing on those values (PostgreSQL 11+).


Specify tablespace for index.

Microsoft SQL Server specific options:


Include additional column values in the index, without actually indexing on those values.

index :name
# CREATE INDEX table_name_index ON table (name)

index [:artist_id, :name]
# CREATE INDEX table_artist_id_name_index ON table (artist_id, name)
[show source]
# File lib/sequel/database/schema_generator.rb, line 239
def index(columns, opts = OPTS)
  indexes << {:columns => Array(columns)}.merge!(opts)
method_missing (type, name = nil, opts = OPTS)

Add a column with the given type, name, and opts. See column for available options.

[show source]
# File lib/sequel/database/schema_generator.rb, line 246
def method_missing(type, name = nil, opts = OPTS)
  name ? column(name, type, opts) : super
primary_key (name, *args)

Adds an autoincrementing primary key column or a primary key constraint. To just create a constraint, the first argument should be an array of column symbols specifying the primary key columns. To create an autoincrementing primary key column, a single symbol can be used. In both cases, an options hash can be used as the second argument.

If you want to create a primary key column that is not autoincrementing, you should not use this method. Instead, you should use the regular column method with a primary_key: true option.

If an array of column symbols is used, you can specify the :name option to name the constraint.



For non-composite primary keys, respects the existing order of columns, overriding the default behavior of making the primary key the first column.


primary_key(:id, type: :Bignum, keep_order: true)
primary_key([:street_number, :house_number], name: :some constraint_name)
[show source]
# File lib/sequel/database/schema_generator.rb, line 277
def primary_key(name, *args)
  return composite_primary_key(name, *args) if name.is_a?(Array)
  column = @db.serial_primary_key_options.merge({:name => name})
  if opts = args.pop
    opts = {:type => opts} unless opts.is_a?(Hash)
    if type = args.pop
      opts = opts.merge(:type => type)

  @primary_key = column
  if column[:keep_order]
    columns << column
primary_key_name ()

The name of the primary key for this generator, if it has a primary key.

[show source]
# File lib/sequel/database/schema_generator.rb, line 299
def primary_key_name
  @primary_key[:name] if @primary_key
respond_to_missing? (meth, include_private)

This object responds to all methods.

[show source]
# File lib/sequel/database/schema_generator.rb, line 251
def respond_to_missing?(meth, include_private)
spatial_index (columns, opts = OPTS)

Add a spatial index on the given columns.

[show source]
# File lib/sequel/database/schema_generator.rb, line 304
def spatial_index(columns, opts = OPTS)
  index(columns, opts.merge(:type => :spatial))
unique (columns, opts = OPTS)

Add a unique constraint on the given columns.

unique(:name) # UNIQUE (name)

Supports the same :deferrable option as column. The :name option can be used to name the constraint.

[show source]
# File lib/sequel/database/schema_generator.rb, line 314
def unique(columns, opts = OPTS)
  constraints << {:type => :unique, :columns => Array(columns)}.merge!(opts)