[TYPO3-mvc] Discussion: Storage Backend not compatible to DBAL

[Jochen Rau]

Hi Steffen,

sorry for being absent in this discussion started by me.

Steffen Ritter wrote:
>> There is an open issue "Typo3DBBackend.php not compatible to DBAL" on
>> forge.
>>
>> http://forge.typo3.org/issues/show/5374

[...]

> since the "fast Version" of rewriting did not get into 4.3 (sry i did
> not have the time for finishing patch due to an "clients emergency"
> (server break)) i would say for 4.4 we should provide a clean solution.
>
> BUT: providing backends for all instead of using DBAL is not a solution
> to me.
>
> Why?
> 1. mixed envirenments would not be supported (as of DBAL)

They could be easily supported, if we provide a solution to select the 
Storage Backend for each class. Dependency Injection comes in mind.

> 2. table Mappings would not be supported (mapping feature of extbase
> works only within one database)

This is related to 1. and is only a matter of configuration.

> 3. reinvent the wheel? better join forces an optimize DBAL

No. I don't want to reinvent the wheel. There is a lot of knowledge 
inside DBAL. I suggest to reuse this knowledge and build it upon a clean 
architecture.

Currently DBAL is structured like this (very simplified)

function INSERT () {
	switch(type)	{
		case 'native':
			// some stuff
			break;
		case 'adodb':
			// a lot of stuff
			break;
		case 'userdefined':
			//some stuff
			break;
	}
}

function UPDATE () {
	switch(type)	{
		case 'native':
			// some stuff
			break;
		case 'adodb':
			// a lot of stuff
			break;
		case 'userdefined':
			//some stuff
			break;
	}
}

function SELECT () {
	switch(type)	{
		case 'native':
			// some stuff
			break;
		case 'adodb':
			// a lot of stuff
			break;
		case 'userdefined':
			//some stuff
			break;
	}
}


And I still propose to have

StorageBackendInterface
- AbstractStorageBackend <- some common functionality
     - AbstractAnsiSqlBackend <- does most of the parsing
         - MySqlBackend
         - AdodbBackend
         - PostgresBackend
         - PdoBackend
         - [userdefinded]

which is much cleaner IMO.

Is it worth the effort? I think it is. Why?

It's far more attractive for other developers to extend and improve the 
system because they only have to dig into a small classes (like 
PostgresBackend). It's clean and modular.

Regards
Jochen

-- 
Every nit picked is a bug fixed