[ https://issues.apache.org/jira/browse/SOLR-10829?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16040483#comment-16040483 ]
Alan Woodward commented on SOLR-10829: -------------------------------------- Do you think it's worth creating a specific IDField class for this? That way we can enforce ID requirements (keyword analyzed, doc values for sorting, no point values) for the unique key field, and for _root_ and _parent_ and other special fields. The schema for 7.0 detects if any of these fields aren't using the IDField type and emits a warning, and 8.0 errors out. > IndexSchema should enforce that uniqueKey field must not be points based > ------------------------------------------------------------------------ > > Key: SOLR-10829 > URL: https://issues.apache.org/jira/browse/SOLR-10829 > Project: Solr > Issue Type: Bug > Security Level: Public(Default Security Level. Issues are Public) > Reporter: Hoss Man > > if a uniqueKey field is defined, then the IndexSchema should fail fast & hard > on startup if the field type of the specific uniqueKey uses "points" (ie: > {{FieldType.isPointField()}}) > The reason for this is because deleting by id, and overwriting existing > documents are predicated on being able to use > {{IndexWriter.deleteDocuments(Term...)}} and > {{IndexWriter.updateDocument(Term, Iterable<..docs..>)}} respectively -- but > Points based fields have no "Term" that can be based to these methods. > IndexSchema.readSchema should fail fast in this case with a clear error (just > like it does if the uniqueKey field is multivalued) -- This message was sent by Atlassian JIRA (v6.3.15#6346) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org