Should this library catch postgresql-simple's SqlErrors and make them available via PgNamedError?

[jhrcek]

I'd like to start a discussion about ${subj}

The fact that functions like query from postgresql-simple operate in IO monad, whereas the functions in this library use mtl's MonadError (e.g. queryNamed) which makes it a bit harder to handle SqlErrors (thrown by postgresql-simple via throwIO) when using wrappers from this library.

I recently used MonadBaseControl + lifted-base to be able to get access to SqlError thrown by a query and rethrow it in our application monad which also uses MonadError.

Would it be a good idea to take advantage of the fact that potgresql-functions run in IO
and modify the existing PgNamedError by adding new constructor like this?

data PgNameError = 
   ...
   SqlError Database.PostgreSQL.Simple.SqlError

... and then wrapping calls to postgresql-simple query functions in some exception catching code and rethrowing them inside MonadError?

Advantage: In that case we could work with SqlErrors directly via our App's MonadError instance without resorting to its MonadBaseControl IO instance

[turboMaCk]

I made a more general issue for this #40. We don't need to consider just exposing IO exceptions since there are many ways to address the primary concern.