How to use value of nested documents in script scoring - elasticsearch

Question

Schema looks like this:
"mappings": {
"_doc": {
"_all": {
"enabled": false
},
"properties": {
"category_boost": {
"type": "nested",
"properties" : {
"category": {
"type": "text",
"index": false
},
"boost": {
"type": "integer",
"index": false
}
}
}
}
}
}
The document in elastic does have data:
"category_boost": [
{
"category": "A",
"boost": 98
},
{
"category": "B",
"boost": 96
},
{
"category": "C",
"boost": 94
},
],
Inside scoring function:
for (int i=0; i<doc['"'category_boost.boost'"'].size(); ++i) {
if (doc['"'category_boost.category'"'][i].value.equals(params.category)) {
boost = doc['"'category_boost.boost'"'][i].value;
}
}
Also tried length to get size of the array, but did help. Since it does not affect results, I tried to divide by size() and it throws division by zero error, so I conclude the size is 0.
Overall problem: have a map of category->boost which is dynamic and I cannot hardcode into schema. I tried type object with json object, but it turned out you cannot access those objects in scoring functions, therefore I went with arrays with defined types.

Answer 1

nested datatype create sub-documents for representing the items of your collections. So access their doc values in a script is possible but you need to be inside a nested query.
Here is one way of doing it, I hope it fulfills your requirements. This example only returns the document with a score depending on the chosen category.
NB : I used elasticsearch 7 in my local, so your will have to modify the mapping to add your "_doc" entry etc....
Here is the modified mapping, I removed the index: false in nested properties since we now use them in queries
PUT test-score_nested
{
"mappings": {
"properties": {
"category_boost": {
"type": "nested",
"properties": {
"category": {
"type": "keyword"
},
"boost": {
"type": "integer"
}
}
}
}
}
}
Then I add your sample data :
POST test-score_nested/_doc
{
"category_boost": [
{
"category": "A",
"boost": 98
},
{
"category": "B",
"boost": 96
},
{
"category": "C",
"boost": 94
}
]
}
And then the query.
We go one level deep in the nested collection
Inside the collection we use a function score query with the replace mode
Inside the function score, we use a filter query to "select" the good category and use its boost for the scoring
POST test-score_nested/_search
{
"query": {
"nested": {
"path": "category_boost",
"query": {
"function_score": {
"boost_mode": "replace",
"query": {
"term": {
"category_boost.category": {
"value": "A"
}
}
},
"functions": [
{
"field_value_factor": {
"field": "category_boost.boost"
}
}
]
}
}
}
}
}
returns
{
"took" : 0,
"timed_out" : false,
"_shards" : {
"total" : 1,
"successful" : 1,
"skipped" : 0,
"failed" : 0
},
"hits" : {
"total" : {
"value" : 1,
"relation" : "eq"
},
"max_score" : 98.0,
"hits" : [
{
"_index" : "test-score_nested",
"_type" : "_doc",
"_id" : "v3Smqm0BZ7nyeX7PPevA",
"_score" : 98.0,
"_source" : {
"category_boost" : [
{
"category" : "A",
"boost" : 98
},
{
"category" : "B",
"boost" : 96
},
{
"category" : "C",
"boost" : 94
}
]
}
}
]
}
}
I hope it will help you!