[TYPO3-mvc] [!!!] FYI: Implemented first version of single table inheritance.

[Thomas "Thasmo" Deinhamer]

You rock Jochen, thanks for the fast implementation,
I'm going to try out this sweet addition as soon as possible.

Regards,
Thomas

Am 15.04.2010 22:37, schrieb Jochen Rau:
> Hi.
>
> Single table inheritance is a beast ;-) That said, I can describe the
> current state I have committed with r2226.
>
> [!!!] But before we start, note that there is a breaking change. Until
> this rev. Extbase made a "best guess" for the table name if it was not
> the lowercased class name (simply by crawling the class hierarchy
> upwards trying to find a mapping rule or table). I have removed this
> "magic" because It was very hard to understand what was happening; esp.
> if there was an error.
>
> Now, let's have a class hierarchy like
>
> Party
> | |
> Organization Person
> | |
> Company ScientificInstitution
>
> The class Party is a class of another extension. All the other classes
> are part of my extension. The classes Party and Organization will never
> get instanciated but their concrete classes Person, Company and
> ScientificInstitution.
>
> There are different ways to map this class hierarchy onto the Relational
> Model of a database:
>
> 1) Map the entire class hierarchy to a single table.
> aka "Single Table Inheritance"
> 2) Map each concrete class to its own table.
> aka "Class Table Inheritance"
> 3) Map each class to its own table.
> aka "Concrete Table Inheritance"
> 4) Map the classes into a generic structure.
>
> Examples:
> 1) Single Table Inheritance: All the data of the concrete classes
> (Company, ScientificInstitution and Person) are stored in the single
> table "party". This table has an extra field (could be any name)
> indicating the record type (the class name by default). The name of the
> extra field must be defined in the "ctrl" section of the TCA of the
> table (key: "type").
>
> If you query for Company objects through a CompanyRepository, you only
> find objects of this particular type. But you can also have an
> OrganizationRepository that is responsible for Objects of type Company
> *and* ScientificInstitution. The configuration is
>
> config.tx_extbase.persistence.classes {
> Organization {
> mapping {
> tableName = party
> }
> subclasses {
> Company = Company
> ScientificInstitution = ScientificInstitution
> }
> }
> Person {
> mapping {
> tableName = party
> }
> }
> Company {
> mapping {
> tableName = party
> }
> }
> ScientificInstitution {
> mapping {
> tableName = party
> }
> }
> }
>
> As you can see, you have to define the table name for every class that
> is not stored in a table named as the lowercased class name.
>
> There is a new option "subclasses" where you can specify, which classes
> are subclasses and should be taken into account, too. This might be a
> subset of the "real" subclasses in the class hierarchy. It might be
> puzzling, that the statement is "Company = Company". The right part is
> the class name. The left part can be any unique name (suggestion: the
> class name).
>
> Sometimes it is not possible to store the class name as record type.
> think about mapping tt_content onto your class hierarchy. There is a
> field called "CType". You can define, that the CType "text" should be
> mapped onto your class "Tx_TemplaVoila_Domain_Model_Text". The
> configuration is:
>
> config.tx_extbase.persistence.classes {
> Tx_TemplaVoila_Domain_Model_Text {
> mapping {
> tableName = tt_content
> recordType = text
> }
> }
> }
>
> I am sure somebody is "furbo" enough to find this configuration:
>
> config.tx_extbase.persistence.classes {
> Client {
> mapping {
> tableName = party
> recordType = Person
> }
> }
> }
>
> in comparison to
>
> config.tx_extbase.persistence.classes {
> Person {
> mapping {
> tableName = party
> }
> }
> }
>
> But this is very bad style. Why? I am going to spend a beer on T3DD (or
> any other event) for the best explanation. ;-)
>
> 2) Class Table Inheritance: currently not supported by Extbase.
>
> 3) Concrete Table Inheritance: There are tables for each of the concrete
> classes Company, ScientificInstitution, and Person. No further
> configuration needed. You can have a repository for each of the concrete
> classes (but not for the class Organization, yet).
>
> 4) See FLOW3 ;-)
>
> Please test the new feature and give feedback. Are there any logical
> issues? Issues in the code? Any thoughts about the configuration options?
>
> Regards
> Jochen
>
>