No, you are not weird. Yours is exactly my approach: Hand written queries with help from libraries for writing and, especially for retrieving data from the query back into the OO world.
My interpretation is that the relational and OO models are incompatible. Not 100% incompatible, but with enough dissimilarity to make writing a complete and efficient ORM a near impossible task.
Every ORM can easily do the simplest of mappings. Classes to tables, rows to instances is the very basic stuff. It is when you have complex data structures that the task of the ORM is close to impossible. Either you have a DSL for describing this mapping, or you fall back into the general language the rest of the program is written in.
I see little value in a DSL for describing relational to OO mappings, so I write the mapping in the general programming language already in use, hence I do not find any added value in ORMs.
This is the basis for iBATIS in the Java world. You provide the custom SQL and iBATIS knows how to wrap the results up into objects (as long as you provide the mapping).
My interpretation is that the relational and OO models are incompatible. Not 100% incompatible, but with enough dissimilarity to make writing a complete and efficient ORM a near impossible task.
Every ORM can easily do the simplest of mappings. Classes to tables, rows to instances is the very basic stuff. It is when you have complex data structures that the task of the ORM is close to impossible. Either you have a DSL for describing this mapping, or you fall back into the general language the rest of the program is written in.
I see little value in a DSL for describing relational to OO mappings, so I write the mapping in the general programming language already in use, hence I do not find any added value in ORMs.