Jifty::DBI::Schema(3) Use a simple syntax to describe a Jifty table.


package MyApp::Model::Page;
use Jifty::DBI::Schema;
use Jifty::DBI::Record schema {
# ... your columns here ...


Each Jifty Application::Model::Class module describes a record class for a Jifty application. Each "column" statement sets out the name and attributes used to describe the column in a backend database, in user interfaces, and other contexts. For example:

    column content =>
       type is 'text',
       label is 'Content',
       render as 'textarea';

defines a column called "content" that is of type "text". It will be rendered with the label "Content" (note the capital) and as a "textarea" in a HTML form.

Jifty::DBI::Schema builds a Jifty::DBI::Column. That class defines other attributes for database structure that are not exposed directly here. One example of this is the ``refers_to'' method used to create associations between classes.

It re-exports "defer" and "lazy" from Scalar::Defer, for setting parameter fields that must be recomputed at request-time:

    column name =>
        default is defer { Jifty->web->current_user->name };

See Scalar::Defer for more information about "defer".



All these functions are exported. However, if you use the "schema" helper function, they will be unimported at the end of the block passed to "schema".


Takes a block with schema declarations. Unimports all helper functions after executing the code block. Usually used at "BEGIN" time via this idiom:

    use Jifty::DBI::Record schema { ... };

If your application subclasses "::Record", then write this instead:

    use MyApp::Record schema { ... };


DEPRECATED. This method of defining columns will not work anymore. Please use the "schema {}" method documented above.

merge_params HASHREF HASHREF

Takes two hashrefs. Merges them together and returns the merged hashref.

    - Empty fields in subclasses don't override nonempty fields in superclass anymore.
    - Arrays don't merge; e.g. if parent class's valid_values is [1,2,3,4], and
      subclass's valid_values() is [1,2], they don't somehow become [1,2,3,4,1,2].

BUG: This should either be a private routine or factored out into Jifty::Util



Indicates that the column references an object or a collection of objects in another class. You may refer to either a class that inherits from Jifty::Record by a primary key in that class or to a class that inherits from Jifty::Collection.

referencing a record

Correct usage is "references Application::Model::OtherClass by 'column_name'", where Application::Model::OtherClass is a valid Jifty model, subclass of Jifty::Record, and 'column_name' is a distinct column of OtherClass. You can omit "by 'column_name'" and the column name 'id' will be used.

At this moment you must specify type of the column your self to match type of the column you refer to.

You can name a column as combination of 'name' and 'by', for example:

    column user_name => references App::Model::User by 'name', type is 'varchar(64)';

Then user, user_name and respective setters will be generated. user method will return object, user_name will return actual value. Note that if you're using some magic on load for user records then to get real name of loaded record you should use "$record->user->name" instead.

In the above case name of the column in the DB will be 'user_name'. If you don't like suffixes like '_id', '_name' and other in the DB then you can name column without suffix, for example:

    column user => references App::Model::User by 'name', type is 'varchar(64)';

In this case name of the column in the DB will be 'user', accessors will be the same as in above example.

referencing a collection

Correct usage is "references Application::Model::OtherCollection by 'column_name'", where Application::Model::OtherCollection is a valid Jifty model, subclass of Jifty::Collection, and 'column_name' is a column of records in OtherCollection. In this case "by 'column_name'" is not optional.

Columns that refers to a collection are virtual and can be changed. So such columns in a model doesn't create real columns in the DB, but instead it's way to name collection of records that refer to this records.


Simple model with users and multiple phone records per user:

    package TestApp::Model::User;
    use Jifty::DBI::Schema;
    use Jifty::DBI::Record schema {
        column name  => type is 'varchar(18)';
        column phones => references TestApp::Model::PhoneCollection by 'user';
    package TestApp::Model::Phone;
    use Jifty::DBI::Schema;
    use Jifty::DBI::Record schema {
        column user  => references TestApp::Model::User by 'id',
            is mandatory;
        column type  => ...;
        column value => ...;

From a user record you get his phones and do something:

    my $phones = $user->phones;
    while ( my $phone = $phones->next ) {

From a phone record you can get its owner or change it:

    my $user_object = $phone->user;
    my $user_id = $phone->user_id;
    $phone->set_user( $new_owner_object );
    $phone->set_user( 123 );    # using id
    $phone->set_user_id( 123 ); # the same, but only using id


Synonym for "references".


Helper for "references". Used to specify what column name should be used in the referenced model. See the documentation for "references".


type passed to our database abstraction layer, which should resolve it to a database-specific type. Correct usage is "type is 'text'".

Currently type is passed directly to the database. There is no intermediary mapping from abstract type names to database specific types.

The impact of this is that not all column types are portable between databases. For example blobs have different names between mysql and postgres.


Give a default value for the column. Correct usage is "default is 'foo'".


Used for default values, to connote that they should not be quoted before being supplied as the default value for the column. Correct usage is "default is literal 'now()'".


Defines a subroutine which returns a true value only for valid values this column can have. Correct usage is "validator is \&foo".


States that this column is not writable. This is useful for properties that are set at creation time but not modifiable thereafter, like 'created by'. Correct usage is "is immutable".


States that this column is not directly readable by the application using "$record->column"; this is useful for password columns and the like. The data is still accessible via "$record->_value('')". Correct usage is "is unreadable".


Sets a maximum max_length to store in the database; values longer than this are truncated before being inserted into the database, using Jifty::DBI::Filter::Truncate. Note that this is in bytes, not characters. Correct usage is "max_length is 42".


Mark as a required column. May be used for generating user interfaces. Correct usage is "is mandatory".


Same as ``mandatory''. This is deprecated. Correct usage would be "is not_null".


Mark as an autocompleted column. May be used for generating user interfaces. Correct usage is "is autocompleted".


Declares that a column should only have distinct values. This currently is implemented via database queries prior to updates and creates instead of constraints on the database columns themselves. This is because there is no support for distinct columns implemented in DBIx::DBSchema at this time. Correct usage is "is distinct".


Used to declare that a column references a collection, which hides it from many parts of Jifty. You probably do not want to set this manually, use "references" instead.


Declares that a column is not backed by an actual column in the database, but is instead computed on-the-fly using a method written by the application author. Such columns cannot (yet) be used in searching, sorting, and so on, only inspected on an individual record.


Declares an integer sort value for this column. By default, Jifty will sort columns in the order they are defined.


Alias for "sort_order".


Sets a list of input filters on the data. Correct usage is "input_filters are 'Jifty::DBI::Filter::DateTime'". See Jifty::DBI::Filter.


Sets a list of output filters on the data. Correct usage is "output_filters are 'Jifty::DBI::Filter::DateTime'". See Jifty::DBI::Filter. You usually don't need to set this, as the output filters default to the input filters in reverse order.


Sets a list of filters on the data. These are applied when reading and writing to the database. Correct usage is "filters are 'Jifty::DBI::Filter::DateTime'". See Jifty::DBI::Filter. In actuality, this is the exact same as ``input_filters'', since output filters default to the input filters, reversed.


What application version this column was last changed. Correct usage is "since '0.1.5'".


The version after this column was supported. The column is not available in the version named, but would have been in the version immediately prior.

Correct usage is "till '0.2.5'". This indicates that the column is not available in version 0.2.5, but was available in 0.2.4. The value specified for ``since'' must be less than this version.


A list of valid values for this column. Jifty will use this to automatically construct a validator for you. This list may also be used to generate the user interface. Correct usage is "valid_values are qw/foo bar baz/".

If you want to display different values than are stored in the DB you can pass a list of hashrefs, each containing two keys, display and value.

 valid_values are
  { display => 'Blue', value => 'blue' },
  { display => 'Red', value => 'red' }


Alias for "valid_values".


Designates a human-readable label for the column, for use in user interfaces. Correct usage is "label is 'Your foo value'".


A sentence or two to display in long-form user interfaces about what might go in this column. Correct usage is "hints is 'Used by the frobnicator to do strange things'".


The displayed length of form fields. Though you may be able to fit 500 characters in the field, you would not want to display an HTML form with a size 500 input box.


Used in user interface generation to know how to render the column.

The values for this attribute are the same as the names of the modules under Jifty::Web::Form::Field, i.e.

  • Button
  • Checkbox
  • Combobox
  • Date
  • Hidden
  • InlineButton
  • Password
  • Radio
  • Select
  • Textarea
  • Upload
  • Unrendered

You may also use the same names with the initial character in lowercase.

The ``Unrendered'' may seem counter-intuitive, but is there to allow for internal fields that should not actually be displayed.

If these don't meet your needs, you can write your own subclass of Jifty::Web::Form::Field. See the documentation for that module.


Alias for "render_as".


An index will be built on this column Correct usage is "is indexed"






This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.