Using empty string in range query 'gt' returns no documents

[FlorianKalinna]

I have created a new Index and then added two documents to it.

PUT test_index
PUT /test_index/_doc/1 { "A" : "5" }
PUT /test_index/_doc/2 { "B" : "5" }

After that I searched for every document with A greater than emty in that index,

GET test_index/_search { "query": { "range": { "A": { "gt": "" } } }, "size": 2000 }

and received no document.

{ "took" : 0, "timed_out" : false, "_shards" : { "total" : 1, "successful" : 1, "skipped" : 0, "failed" : 0 }, "hits" : { "total" : { "value" : 0, "relation" : "eq" }, "max_score" : null, "hits" : [ ] } }

On the other hand when I searched for every document with A greater than equal empty,

GET test_index/_search { "query": { "range": { "A": { "gte": "" } } }, "size": 2000 }

I received one document.

{ "took" : 0, "timed_out" : false, "_shards" : { "total" : 1, "successful" : 1, "skipped" : 0, "failed" : 0 }, "hits" : { "total" : { "value" : 1, "relation" : "eq" }, "max_score" : 1.0, "hits" : [ { "_index" : "test_index", "_type" : "_doc", "_id" : "1", "_score" : 1.0, "_source" : { "A" : "5" } } ] } }

Why does the greater than comparison behave different and does not return a document, too?

[elasticmachine]

Pinging @elastic/es-search (:Search/Search)

[cbuescher]

This is indeed odd and to me seems a bit counterintuitive. At least for the lower bound of a text or keyword field I would assume ever value being larger than the empty string. The problem with this edge case is that the empty string should be considered an open lower bound that matches everything. This works as expected for "inclusice" lower bound (i.e. "gte") but not for "exclusive" (i.e. "gt") because its not completely clear which term to exclude. Internally we use Lucenes TermRangeQuery which does the right thing for "null" values here implicitely sets the inclusion flag to "true" in this case (see https://github.com/apache/lucene-solr/blob/master/lucene/core/src/java/org/apache/lucene/search/TermRangeQuery.java#L52 and https://github.com/apache/lucene-solr/blob/master/lucene/core/src/java/org/apache/lucene/search/TermRangeQuery.java#L81).
I think we should treat the empty string in the same way when constructing the range query and implicitely change the inclusion flag in this case to avoid this confusion.

[cbuescher]

Should be fixed in 7.11.