On Tue, Aug 14, 2007 at 03:15:44PM -0400, Alvaro Herrera wrote:
> Bruce Momjian wrote:
> > A lot of work has been done to try to get /contrib/tsearch2 into the
> > core backend for 8.3, but we have hit a roadblock in how to handle
> > multiple text search configurations. (FYI, the documentation is at
> > http://momjian.us/expire/textsearch/HTML/textsearch.html.)
> >
> > There are three options for controlling text search configurations:
> >
> > 1) have a GUC variable which specifies the default configuration
> > 2) require the configuration to be always specified
> > 3) use the type system to automatically use the right configuration
> >
> > The problem with #1 is that is it error-prone (easy to mismatch
> > configurations). One idea was to have the GUC be super-user-only but
> > then restoring a dump as non-super-user is a problem.
>
> What is the worst consequence of mismatching configuration? Does it
> cause a system crash? A backend hang? A corrupted index? Lost data?
> Or does it, as I assume, just fail to return the exact result set that
> would be returned if the correct configuration was supplied?
>
> If the answer is the latter, I think this is not so huge a problem that
> FULL TEXT should be rejected for 8.3 on these grounds. We just tell
> people to use the correct query and be done with it.
>
I think that users of the full-text option would need to read the
documentation and we could include any needed caveats. This would
certainly do for the 8.3 release and would give us time to simplify
the management and use in the 8.4 release. I, for one, have been
waiting a long time for it to be integrated into the database.
Ken