[SOLR-8737] Managed synonym lists do not include the original term in the expand

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 5.5
• Fix Version/s: 5.5.1
• Component/s: Schema and Analysis
• Assignee: Jan Høydahl
• Reporter: Jan Høydahl
• Created: 25/Feb/16 13:18
• Updated: 26/Feb/16 17:57
• Resolved: 26/Feb/16 09:05

Description

Spinoff from discussion in solr-user list http://find.searchhub.org/document/8dfce8a277de0f2a

The managed synonyms filter does not behave the same way as the original synonym filter when a list is added. The original synonyms filter with default expand=true produces the following map when parsing a line:

Input:

  a, b, c

Becomes:

  a => a, b, c
  b => a, b, c
  c => a, b, c

But the managed filter excludes the original term in the mapping, so an input ["a", "b", "c"] becomes:

  a => b, c
  b => a, c
  c => a, b

This can also be seen in TestManagedSynonymFilterFactory.java where it is tested explicitly, while the tests for the file based synonymfilter expect an all-way expand including the original term.

This causes a query for "a" to not match documents with the term "a", but only those with term "b" or "c".

The offending line in ManagedSynonymFilterFactory is this

188:   treeTerms.remove(origTerm);

Issue Links

• is related to: SOLR-6878 solr.ManagedSynonymFilterFactory all-to-all synonym switch (aka. expand)