Data Mapper Design Pattern and Gateways - Newbie question

Please, correct me if I'm wrong:

If we use a Dao/Vo pattern or a TDG pattern we will have a nice code organization by having for each (or at least for a lot of) tables a related class.

The problem with this approach is that or data IS NOT closed inside a given table. We have some domain specific data, like findDogBreed(); or findBookBestSellerAuthor(); and the above patterns don't seem to deal with this nicely.

Once sol开发者_JS百科ution is to use Mappers. Mappers will contain a set of methods and properties related to one table BUT they will not be closed to that table only nor will they be related to a specific SQL Schema.

The problem is, if we start to abstract all those things, we will NOT have access to SQL syntax. What if we need our database administrator to work on it ? And on more complex queries, using mappers could lead to a really messy abstraction "thing".

Is this correct ? If so, I'm wondering what paths do we have in order to find a middle term here.

You don't have to lose the option to write SQL manually when you abstract the functionality, even on multiple levels abstraction.

E.g. look at Doctrine, which is Hibernate-inspired ORM for PHP. It allows you to write queries in DQL (Doctrine Query Language) that translates to SQL and automatically maps your entities, but you can also write native SQL (most often for performance optimization), but you need to define the result mapping by yourself.