Re: BUG #8354: stripped positions can generate nonzero rank in ts_rank_cd

Hi Bruce,

In normal use this won't affect any data that ends up in an index.

The only exception would be where somebody has created an index over the
result of ts_rank_cd - indexing search rank for the most common search
terms, maybe?

But that's true of any function whose result is indexed. What's the upgrade
policy in that case?

Cheers,
Alex


On Tue, Feb 11, 2014 at 11:45 PM, Bruce Momjian <bruce@momjian.us> wrote:

> On Fri, Feb  7, 2014 at 02:07:59AM +0800, Alexander Hill wrote:
> > Hi Bruce, all,
> >
> > I think this can be solved (if it's agreed that it's a bug) in a pretty
> > straightforward way: when creating the document representation used in
> > calculating cover density rank, we can just skip lexemes with no position
> > entirely.
> >
> > Fix and tests here:
> https://github.com/AlexHill/postgres/compare/bug_8354
> >
> > As a patch file here: https://github.com/AlexHill/postgres/commit/
> > cd522b254d166d569b86803115f0f499864e949b.patch
>
> OK, great.  I am attaching this patch.  Would this change the contents
> of any indexes?  That would be a problem for pg_upgrade.
>
> --
>   Bruce Momjian  <bruce@momjian.us>        http://momjian.us
>   EnterpriseDB                             http://enterprisedb.com
>
>   + Everyone has their own god. +
>