Lucene多值字段-将TextField与StringField混合

Question

I query my index using one multivalued field. During indexing the field is created using few fields, some of them are TextField and some are StringField .

The problem I had was that by querying the index using query with more that one word I have got: java.lang.IllegalStateException: field "someField" was indexed without position data; cannot run PhraseQuery (term=someTerm)

I change the way the multivalued field is created, only TextField are used and the problem disapeard.

It seems that TextField and StringField should not be mixed in one multivalued field. Am I right? Could anybody give me some explanation why (or why not)?

Answer 1

StringField is explicitly set to store only docs ( IndexOptions.DOCS_ONLY ), which omits frequencies and positions from being stored. Since it is effectively a keyword field, and multiple words will be stored as a single token, then running a phrase query against it doesn't really make sense.

While it is certainly possible to mix different field types into the same field, this seems to invite confusion and unpredictable results to me. I would recommend being consistent about the types added to a particular field, and if you need values added with significantly different logic governing them, like the differences between TextField vs. StringField, it would probably be a much better idea to place them in different fields in the index.

If this is happening in some sort of catch-all, convenience field (like an the text field from this SOLR example ), then using a TextField for anything is probably a reasonable idea.