Hi Elbek, Maybe go with ContainedByIntervalsSource? ContainingIntervalsSource is actually the big source filtered by small source, and ContainedByIntervalsSource is the opposite so it should give the expect behavior?
Best Patrick elbek kamoliddinov <elbek....@gmail.com> 于2021年6月2日周三 下午2:55写道： > Hello everyone, > > I am experimenting with Interval Queries to get phrase match count within > parts of an indexed field. > > ContainingIntervalsSource seemed to be the way to go but, it only considers > at most a single match per region. > Example: > Field value: "[a b c d e a c] e f g h [a c k]" (opening and closing square > braces are not part of the text, but shows region of the field I am > interested in) > > Within the regions in the field I am trying to find all phrase match > positions for say 'a c' with slop=1 > First region has 2 matches and the second region has 1. > ContainingIntervalsSource produces an iterator that produces the first > match position for the first region ([a b c d e a c]). but there are two > matches in this region. It seems this behavior is by design. Is it possible > to accomplish this with the existing interval sources or one should write a > custom one for this? > > On a related note, does it make sense for ContainingIntervalsSource to > produce multiple match positions for the first segment? > > Thanks, > Elbek. >
