Children aggregation #

A single-bucket aggregation that scopes its sub-aggregations to the child documents of a relation field (parent-child joins). In a corpus where parents and children share an index, it is the lens that looks at the child side while the outer query talks about parents.

How it works #

Engine note: the current implementation runs sub-aggregations over the scoped document set keyed by the relation; it does not resolve per-parent child sets.

Concretely, the engine returns ONE bucket, keyed by the type name, whose doc_count is the size of the current candidate set, with the sub-aggregations executed over that same set. It does not walk the relation to isolate actual child documents, and per-parent grouping (one bucket per parent) is not available — full parent-child join semantics need join support at the storage layer. Treat it as a labelled scope for relation-oriented analysis, and pair it with the traverse query when the graph edges themselves matter.

Examples #

Articles carry a related_to relation; look at the relation’s child side:

The response is a one-element buckets array carrying the relation name as its key, with the sub-aggregations flattened inside:

{
  ...
  "aggregations": {
    "related": {
      "buckets": [
        {
          "key": "related_to",
          "doc_count": 44,
          "top_tags": {
            "buckets": [
              { "key": "optimizer", "doc_count": 2 },
              { "key": "search", "doc_count": 2 },
              ...
            ]
          }
        }
      ]
    }
  }
}

Use cases #

• Child-side rollups. Keep the outer query about the parent scope while a nested terms/sum reports child-shaped statistics under a clearly labelled bucket.
• Relation health checks. A value_count under the bucket plus a traverse walk of the edges shows whether declared relations are actually populated.
• Graph neighbourhood stats. On an edge-shaped corpus like articles (related_to edges between 20 nodes), the bucket labels a scope whose sub-aggregations describe the documents carrying edges.
• Migration scaffolding. Queries written against the ES shape parse and run, so porting an existing parent-child dashboard degrades to a scoped aggregate instead of failing outright.

Parameters for children #

• type
  (Required, string) Name of the child relation to aggregate.
• aggs
  (Optional, object) Sub-aggregations over the child documents.

Related #

• nested — the nested-object counterpart, with full engine support
• Relational field parameters — declaring references / relation fields in the schema
• Traverse query — walking relation edges as a graph
• Join query — cross-collection semi-joins