[ https://issues.apache.org/jira/browse/LUCENE-6609?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14648066#comment-14648066 ]
Hoss Man commented on LUCENE-6609: ---------------------------------- I think the same thing would make sense for {{SortedSetFieldSource.getSortField()}} -- it can return an instance of {{SortedSetSortField}}. > FieldCacheSource (or it's subclasses) should override getSortField > ------------------------------------------------------------------ > > Key: LUCENE-6609 > URL: https://issues.apache.org/jira/browse/LUCENE-6609 > Project: Lucene - Core > Issue Type: Improvement > Reporter: Hoss Man > > {{ValueSource}} defines the following method... > {code} > public SortField getSortField(boolean reverse) { > return new ValueSourceSortField(reverse); > } > {code} > ...where {{ValueSourceSortField}} builds up a {{ValueSourceComparator}} > containing a {{double[]}} based on the {{FunctionValues}} of the original > {{ValueSource}}. > meanwhile, the abstract {{FieldCacheSource}} exists as a base implementation > for classes like {{IntFieldSource}} and {{DoubleFieldSource}} which wrap a > {{ValueSource}} around {{DocValues}} for the specified field. > But neither {{FieldCacheSource}} nor any of it's subclasses override the > {{getSortField(boolean)}} method -- so attempting to sort on something like > an {{IntFieldSource}} winds up using a bunch of ram to build that > {{double[]}} to give users a less accurate sort (because of casting) then if > they just sorted directly on the field. > is there any good reason why {{FieldCacheSource}} subclases like > {{IntFieldSource}} shouldn't all override {{getSortField}} with something > like... > {code} > public SortField getSortField(boolean reverse) { > return new SortField(field, Type.INT, reverse); > } > {code} > ? -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org