ActiveRecord 4.x Native Database Types Override

Define native database types and change default migration behavior in ActiveRecord/Rails.


In your ActiveRecord/Rails 4.x project, add this to your Gemfile:

gem 'activerecord-native_db_types_override'

Then run:

bundle install


Define configuration to be run prior to migration execution, such as in a Rails initializer or environment file.

Configure Adapters

The following symbols can be used can be used with the NativeDbTypesOverride.configure_adapters method to automatically require and load the right adapter class prior to configuring it:

Adapters included in ActiveRecord:

Adapters that require other gems:

You can also specify a custom adapter class that uses a native_database_types method to map types. See 'Other Adapters' for more information.


The following would:

  • use primary key type 'bigserial primary key' for in migrations
  • change add_reference behavior in migrations to use bigint type by default
  • use timestamptz for all datetimes and timestamps created by migrations*

    NativeDbTypesOverride.configure_adapters({ postgres: { datetime: { name: "timestamptz" }, timestamp: { name: "timestamptz" }, primary_key: { name: "bigserial primary key"} } })

    NativeDbTypesOverride.configure_migrations({ add_reference: :bigint })

See PostgreSQLAdapter source for the default types.


For the MySQL/MySQL2 adapters, maybe you could change boolean to a string type:

  mysql: {
    boolean: { name: "varchar", limit: 1 }

Alternatively, you could use the this Gem to override primary_key to overcome errors that occur when trying to run rake db:migrate that are caused by MySQL updates (i.e. MySQL 5.7) that no longer support DEFAULT NULL values for PRIMARY KEY. Simply create the following:


require 'active_record/connection_adapters/mysql2_adapter'
  ActiveRecord::ConnectionAdapters::Mysql2Adapter => {
    primary_key: "int(11) auto_increment PRIMARY KEY"


# Load monkey patches to prevent migration errors after importing legacy sql dumpfile
require File.expand_path('../initializers/abstract_mysql2_adapter.rb', __FILE__)

config/database.yml check contains 'adapter: mysql2' for db connection to be used

See [AbstractMysqlAdapter][mysql_adapter] for the default types. (Change code branch as needed.)


Maybe you need to extend the default string limit from 255 to 4096:

  sqlite: {
    string: { :name: "varchar", limit: 4096 }

See [SQLite3Adapter][sqlite_adapter] for the default types. (Change code branch as needed.)


Oracle enhanced adapter isn't included in Rails, so you have to include its gem in your Gemfile along with any other requirements it has.

In addition, it's native_database_types method can define boolean as VARCHAR2 (1 char) or NUMBER (0 or 1), so if that is all you are trying to do, then don't use this gem and just try this or look in its adapter/README to see how this could be done:

ActiveRecord::ConnectionAdapters::OracleEnhancedAdapter.emulate_booleans_from_strings = true

However, if you need to make another change like making datetime and timestamp store timezones and you want to emulate_booleans_from_strings, just ensure that you define the boolean shown in the following example rather than using OracleEnhancedAdapter's emulate_booleans_from_strings option:

  oracle: {
    datetime: { name: "TIMESTAMP WITH TIMEZONE" },
    timestamp: { name: "TIMESTAMP WITH TIMEZONE" },
    boolean: { name: "VARCHAR2", limit: 1 }

The reason for this is that Native Database Types Override gem tries to use the constant NATIVE_DATABASE_TYPES available on most adapters to get the existing hash before doing the overrides, but OracleEnhancedAdapter's emulate_booleans_from_strings option changes what is returned by the native_database_types method to NATIVE_DATABASE_TYPES but with :boolean changed to the value => "VARCHAR2", :limit => 1.

See oracle_enhanced_adapter link below for the default types.

Other Adapters

ActiveRecord adapter class that uses a native_database_types method to define the type mappings (as most do) should work. Just specify the fully-qualified name (e.g. ActiveRecord::ConnectionAdapters::MyDbAdapter) as the key in the options hash.

Be sure to add a require for the adapter, if it has not been loaded already prior to configuration, e.g.:

require 'active_record/connection_adapters/postgresql_adapter'
  ActiveRecord::ConnectionAdapters::PostgreSQLAdapter => {
    datetime: { name: "timestamptz" },
    timestamp: { name: "timestamptz" }


Turn on debug output:

NativeDbTypesOverride.debug = true

If not using one of the supported symbols you may need to add a require for the adapter before the configuration or you may get an uninitialized constant (adapter class) error, depending on whether something else loaded the adapter prior.

Test out a migration and include all the types defined in your adapter's NATIVE_DATABASE_TYPES if you're unsure whether it is defining things correctly, e.g.:

rails g model MyModel a_string:string a_text:text an_integer:integer a_float:float a_decimal:decimal a_datetime:datetime a_timestamp:timestamp a_time:time a_date:date a_binary:binary a_boolean:boolean a_xml:xml a_tsvector:tsvector
rake db:migrate

With Rails you can output the schema to see if the types were set correctly:

rake db:structure:dump

Then just to ensure it really is working, go into rails console:

rails c

And try to play with data (this is postgres-specific):

m =
m.a_string = 'test make this string as long as the limit'
m.an_integer = 12345678
m.a_float = 1.2345789
m.a_decimal = 1.2345789
m.a_datetime =
m.a_timestamp =
m.a_time =
m.a_date =
m.a_binary = 1
m.a_boolean = true
m.a_xml = '<testing>123</testing>'
m.a_tsvector = nil!



(c) 2012-2014 FineLine Prototyping, Inc., (c) 2016-2017 Proto Labs, Inc., released under the MIT license.