691421f287
By default the unified highlighter splits the input into passages using a sentence break iterator. However we don't check if the field is tokenized or not so `keyword` field also applies the break iterator even though they can only match on the entire content. This means that by default we'll split the content of a `keyword` field on sentence break if the requested number of fragments is set to a value different than 0 (default to 5). This commit changes this behavior to ignore the break iterator on non-tokenized fields (keyword) in order to always highlight the entire values. The number of requested fragments control the number of matched values are returned but the boundary_scanner_type is now ignored. Note that this is the behavior in 6x but some refactoring of the Lucene's highlighter exposed this bug in Elasticsearch 7x. |
||
|---|---|---|
| .. | ||
| licenses | ||
| src | ||
| build.gradle | ||