On 4/9/2012 at 3:06 PM, [email protected] wrote:
> LUCENE-3969: [...] tenatively add posLen to ShingleFilter
> [...]
> +++ lucene/dev/branches/lucene3969/modules/analysis/common/src/java/org/
> +++ apache/lucene/analysis/shingle/ShingleFilter.java Mon Apr  9  19:05:47 
> 2012
> [...]
> @@ -319,6 +321,8 @@ public final class ShingleFilter extends
>            noShingleOutput = false;
>          }
>          offsetAtt.setOffset(offsetAtt.startOffset(), 
> nextToken.offsetAtt.endOffset());
> +        // nocommit is this right!?  i'm just guessing...
> +        posLenAtt.setPositionLength(builtGramSize);
>          isOutputHere = true;
>          gramSize.advance();
>          tokenAvailable = true;

+1 - looks right to me.

builtGramSize is the position length of the output shingle - missing positions 
(e.g. from stop words) are represented as "filler" tokens.

Steve

Reply via email to