Question on Xesam capabilities

[Evgeny Egorochkin]

On Friday 10 August 2007 17:16:44 Luca Dionisi wrote:
> On 8/10/07, Evgeny Egorochkin <phreedom.stdin at gmail.com> wrote:
> > On Friday 10 August 2007 13:25:09 Luca Dionisi wrote:
> > > Is it possible with the Xesam API to "search within results"
> > > à la Google or even better?
> > >
> > > I mean, will I be able, with a Xesam compliant search tool,
> > > to do things like the following?
> > >
> > > - search all files of type development (*.c; *.h; *.java; ...)
> > >
> > > - within result set, search for  files containing some text
> > >   e.g. "implements"
> > >
> > > - within new result set, search for  files containing some
> > >   other text, e.g. "get_width ("
> >
> > This is easily implemented on the client side by ANDing the 1st, 2nd etc
> > queries.
> >
> > > - hilight the words that match the last search, and not
> > >   the words used for previous filtering stages.
> >
> > We don't have to provisions for this. Maybe a good idea.
>
> Indeed, I think it could be useful.  That's why I asked.  :P
>
> I thought that the "2-stage search" (as opposed to ANDing
> the criteria) could be used for that matter.

Actually this only concerns the full-text snippet, that's need to be returned. 
This can be solved by adding hints to field criterias. Highlighting is done 
by client so it knows better what to highlight.

> And from this, the second question arises:
> > > And could I expect fast results as for a classic query
> > > with a indexed search engine?
>
> Having a result-set (for the preliminary search) as a
> criterion for the final query, is it suitable for an indexed
> search? Theoretically?

2-stage search won't work well in general. Suppose your first search returns 
2/3 of your index. It's easier to re-do the query.

-- Evgeny