Retrieve both parent and child documents

[Fang_Yan]

I have a data model with parent-child relationship.

Parent type subject mapping:
{
"properties": {
"MRN": {
"type": "keyword"
},
"display_name": {
"type": "keyword"
},
"protocols": {
"type": "keyword"
},
"gender": {
"type": "keyword"
},
"dob": {
"type": "date"
}
....
}

Child type lab mapping:
{
"_parent": {
"type": "subject",
"eager_global_ordinals": false
},
"properties": {
"GUID": {
"type": "keyword"
},
"blinded": {
"type": "boolean"
},
"collect_time": {
"type": "date"
},
.....
}

Each lab document has only 1 parent, while a subject can have many lab documents.

I want to retrieve both parent and child document so the subject's gender, dob etc can be returned as search result.

Sample query:
{"bool": {
"must": [
{
"has_parent": {
"parent_type": "subject",
"inner_hits": {
"_source": {
"includes": ["MRN", "display_name", "gender", "race", "ethnic_group", "dob"]
}
},
"query": {
"bool": {
"must": [
{
"terms": {
"protocols": ["18785-BTRIS-TEST-01"]
}
}
]
}
}
}
},
{
"term": {
"_type": "lab"
}
},
{
"query_string": {
"default_field": "observation.name",
"query": "glucose"
}
}
]
}}

This query gives the expected results. However, it's quite slow. (Takes about 5 seconds on my test environment for size 1000). If I remove the "inner_hits" part, the same query run much faster (100~ milliseconds).

My question: how to retrieve both parent and child documents in a more efficient manner?

[Matthew_Isett]

The most efficient retrieval is de-normalize the data needed into a single object. Which you built by placing the needed ["MRN", "display_name", "gender", "race", "ethnic_group", "dob"] in the child document.

If you upgrade to 5.2 - you can use the Search Profiler in Dev Tools to look at the time breakouts.

[Matthew_Isett]

Can you test out this reverse- nested agg. See if it provides a performance increase over inner hits

https://www.elastic.co/guide/en/elasticsearch/reference/current/search-aggregations-bucket-reverse-nested-aggregation.html