-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
I still want to use HaskellDB (or at least the relational calculus part of it), so I was thinking of splitting it into two pieces: a library that submits SQL queries and returns the results (database interface), and a library that constructs queries with the relational calculus and generates SQL from them. Obviously the database interface must be written for each DBMS product, while the relational calculus library ought to be independent of DBMS (although it might be wise to have SQL compatibility flags for the SQL it generates. For example: Oracle 8 (what I'm using now) only supports SQL-92).
Ideally, we would have the following. 1/ A high level combinator library for relational calculus built on, 2/ A standard low-level Haskell API i.e. functions like Connection -> String -> ( a -> b -> IO b) -> IO b 3/ Database specific bindings. The problem is that projections and cartesian products generate new types, at least in the way it was done in HaskellDB. The form the calculus takes would need to be substantially reworked. Tom -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.2 (GNU/Linux) iD8DBQE/O6DBYha8TWXIQwoRAmkCAKCwZkNpU3TjX5bj7rOZHx5DXRjPhwCgihnY sAWwEKX+5hZ1Tiu4wfSGo7Y= =ALnJ -----END PGP SIGNATURE-----