[SOLR-10836] TextLogisticRegression parser (and others) assume non-empty index.

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 6.7, 7.0
• Assignee: David Smiley
• Reporter: David Smiley
• Created: 07/Jun/17 03:57
• Updated: 08/Jun/19 15:22
• Resolved: 10/Jun/17 04:26

Description

TextLogisticRegressionQParserPlugin, IGainTermsQParserPlugin and SignificantTermsQParserPlugin are similarly affected:

finish() contains this line of code:

Terms terms = MultiFields.getFields(searcher.getIndexReader()).terms(trainingParams.feature);

Instead I think it should be this:

Terms terms = ((SolrIndexSearcher)searcher).getSlowAtomicReader().terms(trainingParams.feature);

MultiFields can take some time to build and Solr (via SlowAtomicReader) has already done this work.

In addition, the code assumes Terms is non-null when in fact it will be null if there's no data.