I am generally a fan of Domain Specific Languages. Especially when they are declarative (rather than imperative) in style. However, I don't like it when people build DSLs on top of SQL.
When using a DSL I like to know what happens one level below the DSL, i.e. the transformations which takes the declarative what to the imperative how. For example, when using a regex it is useful to know about the finite automata used to do the matching.
SQL is in itself a (declarative) DSL. Hence, when using SQL it is nice to have a feeling for what will happen in the DBMS when the SQL statement is executed, what indexes will be used etc. If I have another DSL on top of SQL I have to go two levels deep to get that insight. That is one level too much.