About producing output progressively. May be smth like there -
http://mmd.en.momondo.com/ ? When you search for flights the results are not
coming simultaneously. They are showed to you as they are found and every
result is put where it should be.
On Fri, Jun 25, 2010 at 11:51 PM, Matthew Toseland <
t...@amphibian.dyndns.org> wrote:
> On Friday 25 June 2010 19:49:32 Ximin Luo wrote:
> > On 25/06/10 13:17, Tanya Pyatigorskaya wrote:
> > > Hey!
> > >
> > > I've decided to compose you e-mail now not to forget what I want to
> ask.
> > > Answer when you can, after good Graduation party =)
> > >
> > > Now I'm trying to understand bug 0004111: Library API: support for
> > > aggregation of results from multiple indexes.
> > > I know multiple indexes only in relational databases.In databases
> multiple
> > > index is index to set of columns.
> > > First of all, I don't understand what is really 'multiple index' - it's
> > > meant index by set of words or index by different params?
> > > What params we have here for indexing? I've seen in interface only
> words.
> >
> > "multiple index" just means more than one index. an index on freenet is a
> > {keyword -> (pages)} mapping.
> >
> > > And what is aggregation itself? It's storing results in memory, smth
> like
> > > cache?
> >
> > erm, you put this as part of your project proposal.
> >
> > "c) download and aggregate (for search purposes) indexes published by
> other users;"
> >
> > what did you mean by it here?
> >
> > for what i meant in bug 0004111, consider:
> >
> > - lookup term T in index H1, H2 to get results R1, R2.
> > - lookup term T1, T2 in index H to get results R1, R2.
> >
> > aggregating R1, R2, means presenting a single coherent result-set R to
> the user.
> >
> > in the first case we assume the operation is OR. in the second case this
> might
> > be "T1 OR T2" or "T1 AND T2" etc, depending on what the user wanted.
> >
> > a basic mechanism for this exists already which uses simple set
> operations like
> > union/intersection. i haven't looked deeply at the code for this, but
> there are
> > some bugs and it's not very efficient.
> >
> > in the future, we will want to score pages. so index H1 might give page X
> a
> > score of 0.3, and index H2 might give page X a score of 0.5. in the final
> > result-set R, we need to combine these scores (somehow). this might be
> too
> > complicated for you to do right now, but your architecture should aim to
> be
> > *extensible* so we can add this later.
> >
> > you should have a look at the existing code and decide how you want to
> proceed.
> >
> > Ximin
> >
> > (also sent to devl@ to inform other people)
>
> We have both scores and combination support. It can no doubt be improved.
> One particularly interesting question is whether we can produce output
> progressively even when we are combining indexes and/or searching for
> multiple terms, so that we don't have to display all the results at once, or
> even download them all before displaying.
>
_______________________________________________
Devl mailing list
Devl@freenetproject.org
http://freenetproject.org/cgi-bin/mailman/listinfo/devl
