------- Additional Comments From [hidden email] 2005-06-28 09:20 -------
(In reply to comment #3)
> Going forward I think it would be useful to try retain some of the features
> the existing highlighter (eg IDF weighted fragment scoring, fragSizes defined
> in bytes) and merge with your phrase-highlighting features.Adding span query
> support would be good too. What I'm less clear on right now is how this is
Given the possibility of nested span queries, it might be best
do this is by reindexing the field to be highlighted in ram, reuse
the span query on it for collecting the Spans (via getSpans())
and use the beginnings and the ends from this spans as
the basis for highlighting.
For efficiency during reindexing the analyzer used to assemble
the lucene document could ignore all tokens that will not match,
except for their positions.