Sometimes instead real class name we can use custom handler name. Then this class can't be initialized and we will get this exception ```ReflectionException: Class custom_handler_name does not exist```
It would be nice if there was no description the form label was used instead. In the future I think it would be even better to have the label as an header, and the description as 'additional' instructions.
This is the result of https://github.com/nelmio/NelmioApiDocBundle/issues/410.
This PR aims to provide a uniform way of declaring data-types of parameters for
parsers and handlers to follow. In turn, this would allow formatters to
determine data-types in a cleaner and less volatile manner. (See use-case that
can be improved with this PR:
https://github.com/nelmio/NelmioApiDocBundle/blob/master/Formatter/AbstractFormatter.php#L103)
This is possible by the addition two properties to each property item in
`response`, and `parameters` fields in each API endpoint produced by the
`ApiDocExtractor`:
* `actualType` Contains a value from one of the `DataTypes` class constants.
* `subType` Can contain either `null`, or any other `DataTypes` class constant.
This is relevant when the `actualType` is a `DataTypes::COLLECTION`, wherein
`subType` would specify the type of the collection items. It is also relevant
when `actualType` is a `DataTypes::MODEL`, wherein `subType` would contain an
identifier of the model (the FQCN or anything the parser would wish to specify)
Examples:
```php
array(
'id' => array(
'dataType' => 'integer',
'actualType' => DataTypes::INTEGER,
'subType' => null,
),
'profile' => array(
'dataType' => 'object (Profile)',
'actualType' => DataTypes::MODEL,
'subType' => 'Foo\Entity\Profile',
'children' => array(
'name' => array(
'dataType' => 'string',
'actualType' => DataTypes::STRING,
'subType' => null,
),
'birthDate' => array(
'dataType' => 'date',
'actualType' => DataTypes::DATE,
'subType' => null,
),
)
),
'languages' => array(
'dataType' => 'array of strings',
'actualType' => DataTypes::COLLECTION,
'subType' => DataTypes::STRING,
),
'roles' => array(
'dataType' => 'array of choices',
'actualType' => DataTypes::COLLECTION,
'subType' => DataTypes::ENUM,
),
'groups' => array(
'dataType' => 'array of objects (Group)',
'actualType' => DataTypes::COLLECTION,
'subType' => 'Foo\Entity\Group',
),
'profileRevisions' => array(
'dataType' => 'array of objects (Profile)',
'actualType' => DataTypes::COLLECTION,
'subType' => 'Foo\Entity\Profile',
),
'address' => array(
'dataType' => 'object (a_type_a_custom_JMS_serializer_handler_handles)',
'actualType' => DataTypes::MODEL,
'subType' => 'a_type_a_custom_JMS_serializer_handler_handles',
),
);
```
When a formatter omits the `dataType` property or leaves it blank, it is
inferred within `ApiDocExtractor` before everything is passed to formatters.
With JMS Serializer you can set up own type and use it later with custom handler for this property - Parser allways try to call that custom property type like a class. This fix that.
When no groups are given in the ApiDoc, there should be no GroupExclusion and the Entity should be parsed wihtout exclusions.
Otherwise only the "Default" group of the Entity would be parsed, which may not be used.
The ApiDoc should not enforce the Entity to be grouped with "Default", to generate a "full-view" of it.
