Hey there! Let's work through this together.

First off, to clear up your first question: Not setting a manual unique ID isn't causing that inflated count. Elasticsearch auto-generates a unique ID for every document if you don't provide one, and the standard count operation counts documents, not the individual elements inside array fields. If you're seeing a bigger number than expected from a count query, it's almost certainly because that query is tallying up documents, not the items in your items arrays.

Wait, quick side note: In your example, the total number of items across all documents should be 6 (2 from a, 2 from b, 2 from c)—I assume that "3" was a typo? Either way, here's how to get the correct total of array elements:

Option 1: Use value_count Aggregation (Easiest) #

If your items field is mapped as a keyword (or if your text field has a keyword sub-field, which is common), the value_count aggregation will count every occurrence of a value in the field—including all elements in each array:

{
  "size": 0,  // We don't need the actual documents, just the aggregation result
  "aggs": {
    "total_items_count": {
      "value_count": {
        "field": "items"  // Use "items.keyword" if your field is text with a keyword sub-field
      }
    }
  }
}

Option 2: Sum Array Lengths with a Script #

If fielddata is disabled for your text field (and you don't want to enable it), you can calculate the length of each items array per document and sum those lengths up:

{
  "size": 0,
  "aggs": {
    "total_items_count": {
      "sum": {
        "script": "doc['items'].length"
      }
    }
  }
}

Why Your Original Count Was Off #

The standard _count API (like GET /your_index/_count) or a count query returns how many documents match your filters. In your case, that's 3 documents—but you're trying to count the elements inside those documents' arrays, which requires an aggregation instead of a simple document count.

If you were seeing an even larger number than 3, double-check that your query isn't matching more documents than you intended (e.g., no accidental filters missing). But the main issue is using a document count when you need an array element tally.

内容的提问来源于stack exchange，提问作者skyfail