Bruce Momjian wrote:
> > The problem with that theory is this. If the frontend application just
> > executes some random query, such as "select * from table", you really do not
> > know anything about what is coming back. You must rely on the little bit of
> > information the protocol gives you. In the case of Postgres, it gives you
> > the fieldname, datatype, and size for each column in the result.
> > Unfortunately, for varchar and char(n), the size reports -1. This is not
> > very helpful for describing the result set.
> >
> > Your above example works fine (in fact we use that already) when you know the
> > table and column name, as in metadata functions such as SQLColumns() in the
> > ODBC driver.
>
> Yep. We could pass back atttypmod as part of the PGresult. I can add
> that to the TODO list. Would that help?
Yes, that would do it!
Thank you for listening to our ravings on this issue.
Byron
