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

On Fri, Mar 21, 2014 at 06:10:15PM -0400, Bruce Momjian wrote:
> On Fri, Mar 21, 2014 at 10:45:18AM -0400, Tom Lane wrote:
> > Bruce Momjian <bruce@momjian.us> writes:
> > > OK, combined patch attached.  Is this correct?  I keep hoping someone
> > > else will chime in with an opinion on this, but at this point I think
> > > what you have done is clear enough for me to commit for 9.4.
> >
> > That whole function is seriously undercommented; I suppose it's not this
> > patch's charter to fix that, but we could at least write
> >
> > !                           /* ignore words without positions */
> > !                 entry++;
> > !                 continue;
> >
> > The proposed new documentation text seems pretty badly written.  How about
> >
> > This function requires lexeme positional information to perform its
> > calculation.  Therefore it ignores any <quote>stripped</> lexemes in the
> > <type>tsvector</>.  If there are no unstripped lexemes in the input, the
> > result will be zero.
> >
> > The parenthetical "See" text is ok.
>
> OK, thanks.  Updated patch attached.  If Alexander Hill would like to
> add more C comments to any of the ts*.c files, or docs explaining what
> "cover desity" is, I would be glad to include those in the patch as
> well.

Patch applied with this commit message:

    Fix ts_rank_cd() to ignore stripped lexemes

    Previously, stripped lexemes got a default location and could be
    considered if mixed with non-stripped lexemes.

    BACKWARD INCOMPATIBILITY CHANGE

--
  Bruce Momjian  <bruce@momjian.us>        http://momjian.us
  EnterpriseDB                             http://enterprisedb.com

  + Everyone has their own god. +