Multiple Connections

It is possible to register multiple connections, for example:

Granite::Connections << Granite::Adapter::Mysql.new(name: "legacy_db", url: "LEGACY_DB_URL")
Granite::Connections << Granite::Adapter::Pg.new(name: "new_db", url: "NEW_DB_URL")
​
class Foo < Granite::Base
  connection legacy_db
​
  # model fields
end
​
class Bar < Granite::Base
  connection new_db
​
  # model fields
end

In this example, we defined two connections. One to a MySQL database named "legacy_db", and another to a PG database named "new_db". The connection name given in the model maps to the name of a registered connection.

NOTE: How you store/supply each connection's URL is up to you; Granite only cares that it gets registered via Granite::Connections << adapter_object.

timestamps

The timestamps macro defines created_at and updated_at field for you.

class Bar < Granite::Base
  connection mysql
​
  # Other fields
  timestamps
end

Would be equivalent to:

class Bar < Granite::Base
  connection mysql
​
  column created_at : Time?
  column updated_at : Time?
end

Primary Keys

Each model is required to have a primary key defined. Use the column macro with the primary: true option to denote the primary key.

NOTE: Composite primary keys are not yet supported.

```crystal class Site < Granite::Base connection mysql

column id : Int64, primary: true column name : String end

`belongs_to` associations can also be used as a primary key in much the same way.
​
```crystal
class ChatSettings < Granite::Base
  connection mysql
​
  # chat_id would be the primary key
  belongs_to chat : Chat, primary: true
end

Custom

The name and type of the primary key can also be changed from the recommended id : Int64.

class Site < Granite::Base
  connection mysql
​
  column custom_id : Int32, primary: true
  column name : String
end

Natural Keys

Primary keys are defined as auto incrementing by default. For natural keys, you can set auto: false option.

class Site < Granite::Base
  connection mysql
​
  column custom_id : Int32, primary: true, auto: false
  column name : String
end

UUIDs

For databases that utilize UUIDs as the primary key, the type of the primary key can be set to UUID. This will generate a secure UUID when the model is saved.

class Book < Granite::Base
  connection mysql
​
  column isbn : UUID, primary: true
  column name : String
end
​
book = Book.new
book.name = "Moby Dick"
book.isbn # => nil
book.save
book.isbn # => RFC4122 V4 UUID string

Default values

A default value can be defined that will be used if another value is not specified/supplied.

class Book < Granite::Base
  connection mysql
​
  column id : Int64, primary: true
  column name : String = "DefaultBook"
end
​
book = Book.new
book.name # => "DefaultBook"

Generating Documentation

By default, running crystal docs will not include Granite methods, constants, and properties. To include these, use the granite_docs flag when generating the documentation. E.x. crystal docs -D granite_docs.

Doc block comments can be applied above the column macro.

# If the item is public.
column published : Bool

Annotations

Annotations can be a powerful method of adding property specific features with minimal amounts of code. Since Granite utilizes the property keyword for its columns, annotations are able to be applied easily. These can either be JSON::Field, YAML::Field, or third party annotations.

class Foo < Granite::Base
  connection mysql
  table foos
​
  column id : Int64, primary: true
​
  @[JSON::Field(ignore: true)]
  @[Bar::Settings(other_option: 7)]
  column password : String
​
  column name : String
  column age : Int32
end

Converters

Granite supports custom/special types via converters. Converters will convert the type into something the database can store when saving the model, and will convert the returned database value into that type on read.

Each converter has a T generic argument that tells the converter what type should be read from the DB::ResultSet. For example, if you wanted to use the JSON converter and your underlying database column is BLOB, you would use Bytes, if it was TEXT, you would use String.

Currently Granite supports various converters, each with their own supported database column types:

• Enum(E, T) - Converts an Enum of type E to/from a database column of type T. Supported types for T are: Number, String, and Bytes.

• Json(M, T) - Converters an Object of type M to/from a database column of type T. Supported types for T are: String, JSON::Any, and Bytes.

  • NOTE: M must implement #to_json and .from_json methods.

• PgNumeric - Converts a PG::Numeric value to a Float64 on read.

The converter is defined on a per field basis. This example has an OrderStatus enum typed field. When saved, the enum value would be converted to a string to be stored in the DB. Then, when read, the string would be used to parse a new instance of OrderStatus.

enum OrderStatus
  Active
  Expired
  Completed
end
​
class Order < Granite::Base
  connection mysql
  table foos
​
  # Other fields
  column status : OrderStatus, converter: Granite::Converters::Enum(OrderStatus, String) 
end

Serialization

Granite implements JSON::Serializable and YAML::Serializable by default. As such, models can be serialized to/from JSON/YAML via the #to_json/#to_yaml and .from_json/.from_yaml methods.