ElasticSerach - Statistical facets on length of the list

Question

I have the following sample mappipng:

{
    "book" : {
        "properties" : {
                        "author" : { "type" : "string" },
                        "title" : { "type" : "string" },
                        "reviews" : {
                                "properties" : {
                                        "url" : { "type" : "string" },
                                        "score" : { "type" : "integer" }
                                }
                        },
                        "chapters" : {
                                "include_in_root" : 1,
                                "type" : "nested",
                                "properties" : {
                                        "name" : { "type" : "string" }
                                }
                        }
                }
        }
}

I would like to get a facet on number of reviews - i.e. length of the "reviews" array. For instance, verbally spoken results I need are: "100 documents with 10 reviews, 20 documents with 5 reviews, ..."

I'm trying the following statistical facet:

{
    "query" : {
        "match_all" : {}
    },
    "facets" : {
        "stat1" : {
            "statistical" : {"script" : "doc['reviews.score'].values.size()"}
        }
    }
}

but it keeps failing with:

{
  "error" : "SearchPhaseExecutionException[Failed to execute phase [query_fetch], total failure; shardFailures {[mDsNfjLhRIyPObaOcxQo2w][facettest][0]: QueryPhaseExecutionException[[facettest][0]: query[ConstantScore(NotDeleted(cache(org.elasticsearch.index.search.nested.NonNestedDocsFilter@a2a5984b)))],from[0],size[10]: Query Failed [Failed to execute main query]]; nested: PropertyAccessException[[Error: could not access: reviews; in class: org.elasticsearch.search.lookup.DocLookup]
[Near : {... doc[reviews.score].values.size() ....}]
                 ^
[Line: 1, Column: 5]]; }]",
  "status" : 500
}

How can I achieve my goal?

ElasticSearch version is 0.19.9.

Here is my sample data:

{
        "author" : "Mark Twain",
        "title" : "The Adventures of Tom Sawyer",
        "reviews" : [
                {
                        "url" : "amazon.com",
                        "score" : 10
                },
                {
                        "url" : "www.barnesandnoble.com",
                        "score" : 9
                }
        ],
        "chapters" : [
                { "name" : "Chapter 1" }, { "name" : "Chapter 2" }
        ]
}

{
        "author" : "Jack London",
        "title" : "The Call of the Wild",
        "reviews" : [
                {
                        "url" : "amazon.com",
                        "score" : 8
                },
                {
                        "url" : "www.barnesandnoble.com",
                        "score" : 9
                },
                {
                        "url" : "www.books.com",
                        "score" : 5
                }
        ],
        "chapters" : [
                { "name" : "Chapter 1" }, { "name" : "Chapter 2" }
        ]
}

Answer 1

It looks like you are using curl to execute your query and this curl statement looks like this: curl localhost:9200/my-index/book -d '{....}'

The problem here is that because you are using apostrophes to wrap the body of the request, you need to escape all apostrophes that it contains. So, you script should become:

{"script" : "doc['\''reviews.score'\''].values.size()"}

or

{"script" : "doc[\"reviews.score"].values.size()"}

The second issue is that from your description it looks like your are looking for a histogram facet or a range facet but not for a statistical facet. So, I would suggest trying something like this:

curl "localhost:9200/test-idx/book/_search?search_type=count&pretty" -d '{
    "query" : {
        "match_all" : {}
    },
    "facets" : {
        "histo1" : {
            "histogram" : {
                "key_script" : "doc[\"reviews.score\"].values.size()",
                "value_script" : "doc[\"reviews.score\"].values.size()",
                "interval" : 1
            }
        }        
    }
}'

The third problem is that the script in the facet will be called for every single record in the result list and if you have a lot of results it might take really long time. So, I would suggest indexing an additional field called number_of_reviews that should be populated with the number of reviews by your client. Then your query would simply become:

curl "localhost:9200/test-idx/book/_search?search_type=count&pretty" -d '{
    "query" : {
        "match_all" : {}
    },
    "facets" : {
        "histo1" : {
            "histogram" : {
                "field" : "number_of_reviews"
                "interval" : 1
            }
        }        
    }
}'