Hi,
> -----Mensaje original-----
> De: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED]]En nombre de Geoff
> Hutchison
> Enviado el: lunes, 02 de abril de 2001 0:28
> Para: Patrick
> CC: [EMAIL PROTECTED]
> Asunto: Re: [htdig-dev] htdig still counting factor:0 for
> title,description, and text
>
>
> At 3:20 PM -0400 4/1/01, Patrick wrote:
> >I will gladly investigate and write a patch
>
> ... it's most certainly a bug in the
> current htsearch/parser.cc code--I suspect it doesn't check if a word
> match has a zero scoring factor before it adds it to the list.
> Scoring is done in a separate step at which point the document is
> already added to the result list. (IIRC)
A quick inspection of the current parser makes me think that you're right,
Geoff.
To solve the bug, we could try to fight with the parser :-\, but IMHO it's
easier to insert a patch in Display::buildMatchList().
There the hits scored below a given threshold (e.g. 0) can be filtered from
the match list. Perhaps Patrick would like to do it.
>
> The htsearch query parser needs to be replaced by Quim's new code. I
> don't have the time for at least a month to do this, but would be
> glad to walk anyone through what needs to be done (I suspect it won't
> be too difficult).
Note that the new parser doesn't compute weighted scores, it deals only with
matches.
Scoring is left to the - not yet written - post-processing that will sort
results, format them and so on. The same problem would arise.
I'll be working on the new htsearch ASAP - that doesn't mean right now :) -
so I take note of this point.
Would it be a good idea to have a kind of "min_score" configuration
parameter?
Regards,
_______________________________________________
htdig-dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/htdig-dev
