We are using solr for offering search for our website. We want to add an anchor text field value (from incoming links to a webpage), a link popularity score field value, and other similar field values, that are computed separately from the programs page owners use to index their documents to Solr.
However, the problem is when we add these values, they are later lost when one of the many other indexing programs is run to add new documents or modifications because the entire document is cleared when the new documents are added. Since documents may be re-indexed or new documents added at any time, trying to synchronize populating these values with content indexing is not an option. We also want to avoid forcing all the indexers to change how they index. We are using Solr 3.6. What we would like is to have Solr keep any field values if they are not specified in a document to add, if that document was already existing in the index. Since these other indexing programs do not touch these fields at all, we'd like to keep (copy over) the stored values for these unspecified fields when a document with the same key is indexed. E.g., say a document has "popularity", "body", and "title" field and someone issues a delete command to solr to delete the document then adds a document with the same id but only "body" and "title" we want to have the previous "popularity" value copied over. Would this be possible using a custom UpdateRequestProcessor, or is there any easier/better way to accomplish this? Or is there perhaps already such a processor written? What I am thinking is having the UpdateRequestProcessor query solr for the id of the document to add for any fields not specified in the add request, then adding those values to the input document. Since deleting does not actual change the indexing until a commit is made it should be possible to get the old values, right? -- View this message in context: http://lucene.472066.n3.nabble.com/Keep-existing-field-values-that-are-not-specified-for-re-added-documents-tp4081993.html Sent from the Solr - User mailing list archive at Nabble.com.
