[BUG] Zero results returned for nested text-keyword field blends

[Swiddis]

What is the bug?
Field merging causes zero matches to happen when there's a script pushdown involving text/keyword mixed-type docs on nested fields.

This was fixed in #5358 for non-nested fields, but not for nested ones, as DeepMergeRule follows different rules. The pushed down script attempts to fetch doc_values for the text, which doesn't exist for this field type.

How can one reproduce the bug?
Here's a repro script. The tldr is that we create two documents matching {"log":{"user_agent": "requests/2.32.3"}} with a keyword/text mismatch and attempt to search for both with where log.user_agent = "requests/2.32.3". Output at the end.

#!/bin/bash

set -e

OS_URL="http://localhost:9200"

echo "=== OpenSearch Version ==="
curl -s "$OS_URL" | jq -r '.version.number'

echo -e "\n=== Cleanup ==="
curl -s -X DELETE "$OS_URL/test_*" > /dev/null 2>&1 || true

echo -e "\n=== Test 1: Top-level field conflict ==="
echo "Creating test_flat_1 with text+keyword subfield..."
curl -s -X PUT "$OS_URL/test_flat_1" -H 'Content-Type: application/json' -d'
{
  "mappings": {
    "properties": {
      "user_agent": {
        "type": "text",
        "norms": false,
        "fields": {
          "keyword": {
            "type": "keyword",
            "ignore_above": 256
          }
        }
      }
    }
  }
}' | jq -r '.acknowledged'

echo "Creating test_flat_2 with keyword only..."
curl -s -X PUT "$OS_URL/test_flat_2" -H 'Content-Type: application/json' -d'
{
  "mappings": {
    "properties": {
      "user_agent": {
        "type": "keyword"
      }
    }
  }
}' | jq -r '.acknowledged'

echo "Indexing documents..."
curl -s -X POST "$OS_URL/test_flat_1/_doc" -H 'Content-Type: application/json' -d'
{"user_agent": "requests/2.32.3"}
' > /dev/null
curl -s -X POST "$OS_URL/test_flat_2/_doc" -H 'Content-Type: application/json' -d'
{"user_agent": "requests/2.32.3"}
' > /dev/null
curl -s -X POST "$OS_URL/_refresh" > /dev/null

echo "Query: source=test_flat_* | where user_agent = \"requests/2.32.3\""
RESULT=$(curl -s -X POST "$OS_URL/_plugins/_ppl" -H 'Content-Type: application/json' -d'
{
  "query": "source=test_flat_* | where user_agent = \"requests/2.32.3\""
}
' | jq -r '.total')
echo "Result: $RESULT rows (expected: 2)"
if [ "$RESULT" -eq 2 ]; then
  echo "PASS: Top-level field conflict handled correctly"
else
  echo "FAIL: Expected 2 rows, got $RESULT"
fi

echo -e "\n=== Test 2: Nested field conflict ==="
echo "Creating test_nested_1 with nested text+keyword subfield..."
curl -s -X PUT "$OS_URL/test_nested_1" -H 'Content-Type: application/json' -d'
{
  "mappings": {
    "properties": {
      "log": {
        "properties": {
          "user_agent": {
            "type": "text",
            "norms": false,
            "fields": {
              "keyword": {
                "type": "keyword",
                "ignore_above": 256
              }
            }
          }
        }
      }
    }
  }
}' | jq -r '.acknowledged'

echo "Creating test_nested_2 with nested keyword only..."
curl -s -X PUT "$OS_URL/test_nested_2" -H 'Content-Type: application/json' -d'
{
  "mappings": {
    "properties": {
      "log": {
        "properties": {
          "user_agent": {
            "type": "keyword"
          }
        }
      }
    }
  }
}' | jq -r '.acknowledged'

echo "Indexing documents..."
curl -s -X POST "$OS_URL/test_nested_1/_doc" -H 'Content-Type: application/json' -d'
{"log": {"user_agent": "requests/2.32.3"}}
' > /dev/null
curl -s -X POST "$OS_URL/test_nested_2/_doc" -H 'Content-Type: application/json' -d'
{"log": {"user_agent": "requests/2.32.3"}}
' > /dev/null
curl -s -X POST "$OS_URL/_refresh" > /dev/null

echo "Query: source=test_nested_* | where log.user_agent = \"requests/2.32.3\""
RESULT=$(curl -s -X POST "$OS_URL/_plugins/_ppl" -H 'Content-Type: application/json' -d'
{
  "query": "source=test_nested_* | where log.user_agent = \"requests/2.32.3\""
}
' | jq -r '.total')
echo "Result: $RESULT rows (expected: 2)"
if [ "$RESULT" -eq 2 ]; then
  echo "PASS: Nested field conflict handled correctly"
else
  echo "FAIL: Expected 2 rows, got $RESULT"
fi

echo -e "\n=== Field Capabilities ==="
echo "Top-level field:"
curl -s "$OS_URL/test_flat_*/_field_caps?fields=user_agent" | jq '.fields.user_agent'
echo -e "\nNested field:"
curl -s "$OS_URL/test_nested_*/_field_caps?fields=log.user_agent" | jq '.fields["log.user_agent"]'

echo -e "\n=== Cleanup ==="
curl -s -X DELETE "$OS_URL/test_*" > /dev/null 2>&1
echo "Test indices deleted"

What is the expected behavior?
Should correctly resolve the field & return 2 results.

What is your host/environment?

• mainline

Do you have any screenshots?
N/A

Do you have any additional context?

Related: #4659

=== OpenSearch Version ===
3.9.0-SNAPSHOT

=== Cleanup ===

=== Test 1: Top-level field conflict ===
Creating test_flat_1 with text+keyword subfield...
true
Creating test_flat_2 with keyword only...
true
Indexing documents...
Query: source=test_flat_* | where user_agent = "requests/2.32.3"
Result: 2 rows (expected: 2)
PASS: Top-level field conflict handled correctly

=== Test 2: Nested field conflict ===
Creating test_nested_1 with nested text+keyword subfield...
true
Creating test_nested_2 with nested keyword only...
true
Indexing documents...
Query: source=test_nested_* | where log.user_agent = "requests/2.32.3"
Result: 0 rows (expected: 2)
FAIL: Expected 2 rows, got 0

=== Field Capabilities ===
Top-level field:
{
  "text": {
    "type": "text",
    "searchable": true,
    "aggregatable": false,
    "indices": [
      "test_flat_1"
    ]
  },
  "keyword": {
    "type": "keyword",
    "searchable": true,
    "aggregatable": true,
    "indices": [
      "test_flat_2"
    ]
  }
}

Nested field:
{
  "text": {
    "type": "text",
    "searchable": true,
    "aggregatable": false,
    "indices": [
      "test_nested_1"
    ]
  },
  "keyword": {
    "type": "keyword",
    "searchable": true,
    "aggregatable": true,
    "indices": [
      "test_nested_2"
    ]
  }
}

=== Cleanup ===
Test indices deleted