should nested within must

[gitemconte]

I am curious if it is possible with this module to set up a complex elastic query that has should queries nested within a must.

Simply, I am trying to recreate this SQL query:
SELECT * FROM test where status = 'active' and (city = 'New York' or city = 'Toronto')

I would like to get back something similar to the below, but I have only been able to have the MUST and the SHOULD on the same level. Is this functionality possible? It seems like a simple query case, but maybe I am wrong.

{
    "query": {
        "bool": {
            "must": {
                "bool": {
                    "must": [
                        {
                            "match": {
                                "status": {
                                    "query": "active"
                                }
                            }
                        },
                        {
                            "bool": {
                                "should": [
                                    {
                                        "match": {
                                            "city": {
                                                "query": "New York"
                                            }
                                        }
                                    },
                                    {
                                        "match": {
                                            "city": {
                                                "query": "Toronto"
                                            }
                                        }
                                    }
                                ]
                            }
                        }
                    ]
                }
            }
        }
    }
}

[danpaz]

Nested queries haven't been developed yet, but I think it is possible to add this. #16 is a good example of how this was done for nesting aggregations and I think a similar approach could work here.

[jhn134910]

I may have misinterpreted the question but is the question about supporting nested bool structure rather than a nested query accessing a nested object? By the way -- really love this project.

[chiragob]

is there a way to build has_child with "or" in bodybuilder

{
"query":
{
"bool":
{
"should": [
{
"has_child": {
"inner_hits":{},
"type": "emp",
"query": {
"match": {
"name": "xyz"
}
}
}
},
{
"bool":
{
"must":
[
{
"match" : { "name" : "xyz" }
},
{
"match" : { "_type" : "emps" }
}
],
"must_not": [
{
"has_child": {
"type": "emp",
"query": {
"exists":
{
"field": "name"
}
}
}
}
]
}
}
]
}
}
}

[diegoprd]

@gitemconte I got into the same issue... and yes, I guess we should create a PR for this soon... but for now, as a workaround, I am manually doing it by using the getQuery() function

So, you could do something like this:

const body = bodybuilder()
    .query(...)
    .agg(...)

const innerBoolQueryForTheMustClause = bodybuilder()
   .query(...)

body.getQuery().bool.must.push(innerBoolQueryForTheMustClause.getQuery())

I know is not nice... but as a workaround, for now, it let me do the query I needed...

NOTE: There is a little edge case in the case of not having a must clause yet... so you can easily add the must close by adding a couple of .query('match_all') to the main body...

const body = bodybuilder()
    .query('match_all')
    .query('match_all')

const innerBoolQueryForTheMustClause = bodybuilder()
   .query(...)

body.getQuery().bool.must.push(innerBoolQueryForTheMustClause.getQuery())

Again, I know is not a nice approach but it will do the its job for now...
hope that helps.

[sudo-suhas]

I have been working on a library for building elasticsearch DSL heavily inspired by elastic.js which unfortunately was abandoned.

The requirement, SELECT * FROM test where status = 'active' and (city = 'New York' or city = 'Toronto') can be achieved like so:

const bob = require('elastic-builder');

const boolQry = bob
    .boolQuery()
    .must(bob.matchQuery('status', 'active'))
    .must(
        bob.boolQuery()
            .should(bob.matchQuery('city', 'New York'))
            .should(bob.matchQuery('city', 'Toronto'))
            .minimumShouldMatch(1)
    )
    .toJSON();

Link: http://github.com/sudo-suhas/elastic-builder

Hope this helps 😄

[gophry]

@gitemconte

I tried on a similar requirement: works fine.. on ElasticSearch 2.4 and 5.2

bodybuilder()
    .query('match','status','active')
	.orQuery('match','city','New York')
    .orQuery('match','city','Toronto')
.build()

[caioflores]

@gitemconte I made a approach similar to @diegoprd one, I needed something like this in my filters:

  "filter": {
    "bool": {
      "must": [
        {
          "bool": {
            "should": [
              {
                "term": {"Field": "Value"}
              },
              {
                "term": {"Field": "Value2"}
              }
            ]
          }
        },
        {
          "bool": {
            "should": [
              {
                "term": {"Field2": "Value3"}
              },
              {
                "term": {"Field2": "Value4"}
              }
            ]
          }
        }
      ]
    }
  }

To get this result dinamically with bodybuilder I did:

   const query = bodybuilder().query('match_all');

   // Did this to create the must
   query.filter('term', 'Field', 'temporary');
   query.filter('term', 'Field', 'temporary');
   query.getFilter().bool.must.length = 0;
   
  // For each one of my 'shoulds' I created a filter of bodybuilder type
  // and applied the 'orFilter', than I inserted it in the main query
  Object.keys(filters).forEach((key) => {
      const filter = bodybuilder();
       for (let i = 0; i < filters[key].length; i += 1) {
         const value = filters[key][i];
         filter.orFilter('term', key, value);
       }
       query.getFilter().bool.must.push(filter.getFilter());
   });

Obviously this is not a good solution but it works, and I hope it helps someone, if you have any question just ask.

I started using bodybuilder and my project needs many nested queries, @danpaz we can create a PR for this? How can we help?

[danpaz]

I'd be happy to look at a PR to address this as I see it's caused confusion for folks. I just haven't had time myself to dig into it, and as @gophry points out sometimes the additional nesting isn't exactly needed. Your query may be different, though, not sure. In any case there seems to be a need for this feature.

Btw you may be able to use the undocumented getFilter method to improve your workaround. Take a look at the source code for that, let me know if you have any questions about it.

[caioflores]

@danpaz thanks for the answer. I'll take a look at getFilter and try @gophry approach to see if I can handle it in a better way.

[johannes-scharlach]

As of #134 it is possible to do nested filters, so I'm closing this issue.

As described in #142, this is not possible for nested queries right now. A possible solution that breaks compatability to elasticsearch version 1 is presented in #144.

[dongyunqi2018]

it works in this way:

curl -X GET localhost:9200/transfer_trance/_search?pretty -H "Content-Type:application/json" -d'
{
    "query": {
        "bool": {
            "must": [
                {
                    "term": {
                        "is_del": 1
                    }
                },
                {
                    "bool": {
                        "should": [
                            {
                                "term": {
                                    "monitor_url.keyword": "http://tx-g.xxxxxx.com/d/INzdsdYnu?idfa=__IDFA__&os=__OS__&timestamp=__TS__&callback=__CALLBACK_URL__"
                                }
                            },
                            {
                                "term": {
                                    "download_url.keyword": "https://count.xxxxxxx.com/d.php?id=738896&urlos=android&from_type=web&time=1733"
                                }
                            }
                        ]
                    }
                }
            ]
        }
    },
    "size": 1,
    "sort": [
        {
            "created_at": "desc"
        }
    ]
}
'; echo

[prasher-ayce]

it works in this way:

curl -X GET localhost:9200/transfer_trance/_search?pretty -H "Content-Type:application/json" -d'
{
    "query": {
        "bool": {
            "must": [
                {
                    "term": {
                        "is_del": 1
                    }
                },
                {
                    "bool": {
                        "should": [
                            {
                                "term": {
                                    "monitor_url.keyword": "http://tx-g.xxxxxx.com/d/INzdsdYnu?idfa=__IDFA__&os=__OS__&timestamp=__TS__&callback=__CALLBACK_URL__"
                                }
                            },
                            {
                                "term": {
                                    "download_url.keyword": "https://count.xxxxxxx.com/d.php?id=738896&urlos=android&from_type=web&time=1733"
                                }
                            }
                        ]
                    }
                }
            ]
        }
    },
    "size": 1,
    "sort": [
        {
            "created_at": "desc"
        }
    ]
}
'; echo

@dongyunqi2018 I am facing a similar issue your solution is not working for me,
My required results should work like this
where com_ref= "something" AND( fr_id = 123 OR fr_id =456).
Here is what i am trying but not getting any results
{ "query": { "bool": { "filter": { "bool": { "must": [ { "match_phrase": { "com_ref": "ddfdfdfdf" } }, { "bool": { "should": [ { "term": { "fr_id ": 511168 } }, { "term": { "fr_id ": 511258 } } ] } } ] } } } } }