As with everything, it depends. I've implemented a large system managing billions of records using exactly this approach of cutting out a lot of boilerplate in

As with everything, it depends. I've implemented a large system managing billions of records using exactly this approach of cutting out a lot of boilerplate in the application layer. The most important thing is to be pragmatic. This approach works great for CRUD as well as many types of complex analytical queries. However, for cases where the the complexity or performance of the SQL was unacceptable, we also decoded data into application data structures for further processing. When done well, you can get the best of both worlds as needed.