Great question! Since you can only modify the query field and can't touch the size parameter directly, here are a few practical workarounds to limit the number of returned results using just Elasticsearch query syntax:

1. Use a Range Query with an Auto-Incrementing Field #

If your documents have a globally unique auto-incrementing field (like a custom id field) or you can leverage the internal _seq_no field, you can use a range query to narrow down results to a fixed set of documents.

Example with _seq_no:

First, run a quick query to get the maximum _seq_no in your index (you only need to do this once periodically, or dynamically if your tooling supports it):

GET /your-index/_search
{
  "size": 1,
  "sort": [{"_seq_no": "desc"}],
  "_source": false
}

Suppose the max _seq_no is 1000, and you want the latest 10 results. Your query would look like this:

{
  "query": {
    "bool": {
      "filter": {
        "range": {
          "_seq_no": {
            "gt": 990
          }
        }
      }
    }
  }
}

⚠️ Note: _seq_no is per-shard, so this works best if your index has a single shard, or if you don't mind approximate results across multiple shards. For precise results, use a custom global auto-incrementing field instead.

2. Function Score + Script Filter for Precise Count #

Use function_score to assign a high score to the first N documents you want to keep, and a 0 score to everything else. Then use min_score to filter out the low-scoring documents.

Example:

{
  "query": {
    "function_score": {
      "query": { "match_all": {} }, // Replace with your actual base query
      "functions": [
        {
          "script_score": {
            "script": {
              "source": "doc['_seq_no'].value > params.max_seq_no - params.limit ? 1 : 0",
              "params": {
                "max_seq_no": 1000, // Get this from a pre-query
                "limit": 10
              }
            }
          }
        }
      ],
      "boost_mode": "replace", // Override the default score with our script result
      "min_score": 1 // Only keep documents with a score ≥1
    }
  }
}

This will strictly return only the documents that meet the script condition (in this case, the latest 10).

3. Approximate Random Sampling (For Non-Precise Needs) #

If you don't need an exact count and just want a small sample of results, use random_score to randomly select a subset of documents:

{
  "query": {
    "function_score": {
      "query": { "match_all": {} },
      "functions": [
        {
          "random_score": {}
        }
      ],
      "boost_mode": "replace",
      "min_score": 0.9 // Adjust this value to control sample size (higher = fewer results)
    }
  }
}

This returns documents with a random score above 0.9, which will be roughly 10% of your total documents (adjust the threshold to get more/less results).

内容的提问来源于stack exchange，提问作者Amin Bashiri