Philip Warner wrote:
> At 14:38 11/07/00 +0200, Jan Wieck wrote:
> >> Can I suggest that we also put out a warning when defining an index using a
> >> field with a (potentially) unlimited size? Indexing a text field will
> >> mostly be a bizarre thing to do, but, eg, indexing the first 255 chars of a
> >> text field (via substr) might not be.
> >
> > Marking it BOLD somewhere in the release notes, the CREATE
> > INDEX doc and some other places should be enough. Such a
> > message at every CREATE INDEX is annoying.
>
> The suggestion was only if the index contained a text, lztext etc field,
> but no problem. The way I read your suggestion was that I'd get a real
> error when doing an insert if the text was too large.
Yes, that's what I'm after. It's too fragile IMHO to check on multi column indices with char(n) or so if
resulting index tuples will fit in the future.
The atttypmod field on NUMERIC columns for example doesn't tell the easy way how big the internal
representation might grow. And what about variable size user defined types that are marked toastable? Can you
estimatethe maximum internal storage size for them?
Jan
--
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck@Yahoo.com #