I am messing around with the newly added search-as-you-type field type.
Having documents with following structure
doc_1
...
"nickname": ["Super", "Terrible", "Mediocre"]
...
doc_2
...
"nickname": ["Grande", "Perfect", "Superior"]
....
For a search-as-you-type usecase I am trying to aggregate the results using terms aggregation and order the buckets by maximum _score for each bucket. This allows me to get the unique values that are most relevant for a single valuedfield. However sometimes the field is multivalued like "nickname": ["Super", "Terrible", "Mediocre"] , when searching for Super like in the query below, doing a terms aggregation sorted by max score now includes buckets Super, Terrible,Grande, Perfect and Mediocre with high _score, because the _score is the same for the whole document and all elements in the multivalued field are aggregated. I would like the aggregation to be limited only on the Super and Superior buckets which are the inner_hits. Is it it possible to limit the aggregation only to the hits inside the multivalued fields, excluding other buckets from the aggregation? It seems like nested datatype and inner_hits should help, but i fail to achieve terms aggregation over the inner_hits.
This issue (https://github.com/elastic/elasticsearch/issues/17376) says that it is not possible to aggregate inner_hits , what is the intended solution to this, should I just index the nested entities separetely in another index where I can aggregate as usual?
What I am trying to achieve is getting most relevant results which actually exist in the index so I can do some further filtering of the documents in another query.
The query targeting search-as-you-type field and its subfields. I would like to have a similar results with but only with inner_hits buckets present in the results. Is it possible?
POST /target_index/_search
{
"timeout": "5000s",
"_source": "nickname",
"query": {
"multi_match": {
"type": "phrase_prefix",
"query": "Super",
"fields": [
"nickname.autocomplete",
"nickname.autocomplete._2gram",
"nickname.autocomplete._3gram"
]
}
},
"aggs": {
"termAgg": {
"terms": {
"field": "nickname.keyword",
"size": 10,
"min_doc_count": 1,
"shard_min_doc_count": 0,
"show_term_doc_count_error": false,
"order": {
"max_score": "desc"
}
},
"aggs": {
"max_score": {
"max": {
"script": "_score"
}
}
}
}
}
}