[LUCENE-9091] UnifiedHighlighter HTML escaping should only escape essentials

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 8.5
• Component/s: modules/highlighter
• Lucene Fields: New, Patch Available
• Assignee: David Smiley
• Reporter: Nándor Mátravölgyi
• Created: 12/Dec/19 18:31
• Updated: 28/Aug/22 15:54
• Resolved: 24/Dec/19 04:18

Description

The unified highlighter does not use the org.apache.lucene.search.highlight.SimpleHTMLEncoder through org.apache.solr.highlight.HtmlEncoder. It has the HTML escaping feature re-implemented and embedded in the org.apache.lucene.search.uhighlight.DefaultPassageFormatter.

The HTML escaping done by the unified highlighter escapes characters that do not need it. This makes the result payload 50%+ more heavy with no benefit.

Here is a highlight snippet using the original highlighter:

A <em>filter</em> that stems words using a Snowball-generated stemmer. Available stemmers &amp; x are listed in org.tartarus.snowball.ext. Note: This <em>filter</em> is aware of the KeywordAttribute.

Here is the same highlight snippet using the unified highlighter:

A&#32;<em>filter</em>&#32;that&#32;stems&#32;words&#32;using&#32;a&#32;Snowball&#45;generated&#32;stemmer&#46;&#32;Available&#32;stemmers&#32;&amp;&#32;x&#32;are&#32;listed&#32;in&#32;org&#46;tartarus&#46;snowball&#46;ext&#46;&#32;Note&#58;&#32;This&#32;<em>filter</em>&#32;is&#32;aware&#32;of&#32;the&#32;KeywordAttribute&#46;

Maybe I'm missing the point why this is done the way it is. If this behaviour is desired for some use-case it should be a separate encoder, and the HTML encoder should only escape the necessary characters.

Affects all versions of Lucene-Solr since the addition of the UnifiedHighlighter. Here are the lines where the escaping are implemented differently:

• Escaping by the unified highlighter
• Escaping by the other highlighters