When does minimum_should_match change default value?

Elastic Stack Elasticsearch
1

Hey guys :wave:

We've upgraded from 6.8 to 7.10 recently and are still finding out issues. It seems like minimum_should_match behaviour has been silently updated.

Consider the following repeatable example:

PUT /test-bench
{
  "mappings": {
    "properties": {
      "label": { "type": "text" },
      "object": {
        "type": "nested",
        "properties": {
          "name": {
            "type": "text"
          }
        }
      }
    }
  }
}

POST /test-bench/_doc/1
{
  "label": "how are you doing",
  "object": {
    "name": "how are you doing"
  }
}

GET /test-bench/_search
{
  "query": {
    "match_all": {}
  },
  "highlight": {
    "fields": {
      "label": {},
      "object.name": {}
    }
  }
}

GET /test-bench/_search
{
  "query": {
    "bool": {
      "should": [
        {
          "nested": {
            "path": "object",
            "query": {
              "match_phrase": {
                "object.name": {
                  "query": "how long does take"
                }
              }
            }
          }
        }
      ],
      "must": [
        {
          "term": {
            "_id": {
              "value": "1"
            }
          }
        }
      ]
    }
  },
  "highlight": {
    "fields": {
      "label": {},
      "object.name": {}
    }
  }
}

One would expect this to not return a result, but it does. Not only that, but the highlight doesn't make any sense - match_phrase shouldn't highlight partial matches.

{
  "took" : 3,
  "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-bench",
        "_type" : "_doc",
        "_id" : "1",
        "_score" : 1.0,
        "_source" : {
          "label" : "how are you doing",
          "object" : {
            "name" : "how are you doing"
          }
        },
        "highlight" : {
          "object.name" : [
            "<em>how</em> are you doing"
          ]
        }
      }
    ]
  }
}

Forcing minimum_should_match value to 1 makes this to behave as expected. I wonder why that might be. I see previous responses claiming this shouldn't be needed.

Could someone shed light on this?

Is it that starting with Elasticsearch 7.0 minimum_should_match changes default to 0 the moment we have must clause added? A prompt response would be appreciated.

Thanks,
Nick

2

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.