Bug Span near query: "Less than 2 subSpans.size():1"

[juretta]

Elasticsearch version: 5.5.0 (docker.elastic.co/elasticsearch/elasticsearch:5.5.0)

Description:

The documentation for the span near query states:

The clauses element is a list of one or more other span type queries

(emphasis mine)

Just supplying one clause though, causes an error to be returned iff the HTTP request method is POST instead of GET. E.g.

...
    "failures": [
      {
        "shard": 0,
        "index": "test",
        "node": "...",
        "reason": {
          "type": "illegal_argument_exception",
          "reason": "Less than 2 subSpans.size():1"
        }
      }
    ]

---

Steps to reproduce:

With ES 5.5.0 running:

1. Create a mapping:

curl -X "PUT" "http://localhost:9200/test" \
     -H "Content-Type: text/plain; charset=utf-8" \
     -u elastic:changeme \
     -d $'{
  "settings": {
    "analysis": {
      "analyzer": {
        "my_analyzer": {
          "tokenizer": "my_pattern_tokenizer"
        }
      },
      "tokenizer": {
        "my_pattern_tokenizer": {
          "type": "pattern",
          "pattern": "\\\\/"
        }
      }
    }
  },
  "mappings": {
    "bar": {
      "properties": {
        "path": {
          "type": "text",
          "analyzer": "my_analyzer"
        }
      }
    }
  }
}'

2. Add an example document:

curl -X "POST" "http://localhost:9200/test/bar" \
     -H "Content-Type: application/json" \
     -u elastic:changeme \
     -d $'{
  "path": "foo/bar/baz"
}'

3. Run a search using a span near query with a single clause:

curl -X "POST" "http://localhost:9200/test/bar/_search" \
     -H "Content-Type: application/json" \
     -u elastic:changeme \
     -d $'{
  "query": {
    "span_near": {
      "clauses": [
        {
          "span_first": {
            "match": {
              "span_term": {
                "path": "foo"
              }
            },
            "end": 1
          }
        }
      ],
      "slop": 2,
      "in_order": true
    }
  }
}'

If using an HTTP POST this fails with:

...
    "failures": [
      {
        "shard": 1,
        "index": "test",
        "node": "24CCgJNwRbCoxUlUbFXiuw",
        "reason": {
          "type": "illegal_argument_exception",
          "reason": "Less than 2 subSpans.size():1"
        }
      }
    ]
...

The same request using a GET request method succeeds.

According to https://www.elastic.co/guide/en/elasticsearch/reference/current/search-request-body.html it should be possible to use either POST or GET.

GET vs. POST was just a red herring. This fails in either case.

[cbuescher]

I cannot reproduce the difference between GET/POST you report. I also get the failure you are seeing, I think this might be an error in the documentation for the span near query. My understanding from reading the SpanNearQuery Lucene is that it always requires at least two clauses. Maybe @jpountz or @jimczi can confirm this? In this case we should update the documentation and maybe throw an error earlier in the Elasticsearch query builder if we only have one clause.

[jimczi]

I think it should be handled automatically in ES. I fixed the auto-generated case where another query creates a SpanNearQuery with a single clause but I did not change the SpanNearQueryBuilder itself.
We could also fix this in Lucene since I am not sure why it throws an error instead of just rewriting the query internally. I'll check

[juretta]

Ok, my bad. My test client (Paw) didn't actually sent the request body. Just using curl (as I added as an example, doh) does show that it fails regardless whether it is GET or POST. So I see the same @cbuescher is seeing.

Sorry for the confusion.