On Thu, 2012-08-16 at 19:18 -0400, Leon Smith wrote:
Ok, I really want to get the TypeInfo structure more or less "correct" before we release this work.
After reviewing the documentation regarding the pg_type table, I think it would be a mistake to "mandate" an interpretation of the table in the new TypeInfo structure, as my suggested TypeInfo would do. Rather, I think it would be better to represent (part of) the pg_type table faithfully, closer to what Bas originally did. However, I do think it should be a properly recursive type.
It doesn't look like TypeInfo is returned to users, so I don't see a major problem with using a more convenient representation if we find one. What am I missing?
So my new suggestion for the TypeInfo structure is this:
data TypeInfo = TypeInfo { typoid :: {-# UNPACK #-} !PQ.Oid , typcategory :: {-# UNPACK #-} !Char , typdelim :: {-# UNPACK #-} !Char , typname :: !ByteString , typelem :: Maybe TypeInfo }
We'll be omitting some information there, notably the subtypes for range types and the subtypes for composite types. I assume that you intend to add those organically as the need arises (e.g. keep adding in more parts of the PG catalogs)? Regards, Jeff Davis