[KinoSearch] boosting doc score based on integer data

[Marvin Humphrey]

Dermot:

> I have a integer field (sales), it's the number of sales a doc has
> had. I don't want to sort the hits by this field so sort_spec is not
> the solution. I would like to use the integer value to moderate the
> final results. In effect, if a doc has 0 sales, it would get a boost
> of 0.8 and those docs with 5+ would get a 1.2 boost.

I suggest setting a document boost at index time, with the value determined by
your "sales" field:

  my $doc = KinoSearch::Doc->new( 
    fields => \%fields, 
    boost  => calc_boost( $fields{sales} )
  );

> I have stored and indexed the field and it isn't being analyzed. I'm
> not clear on how to proceed after that. Do I need to create a custom
> Compiler/Query/Scorer?

I can envision a design which would allow you to affect scores using a custom
function fed by the integer value for each document.  The advantage would be
that you could tweak at search-time rather than having to re-index. 

However, this would have to wait for KS to add an integer field type, and, if
scaling to large size is important, a C API.

> A little guidance would be much appreciated. There are a lot of docs
> under KS::Search but I am not sure how to mould a custom
> Query-to-Scorer compiler or if that is the best path for me to take.

There's a Cookbook document on the subject of creating custom Query
subclasses:

    http://xrl.us/o3z43 (Link to www.rectangular.com)

However, I recommend that you go with document boosts instead.

Marvin Humphrey