Semantic_text match_all with Highlighter

[Samiul-TheSoccerFan]

This PR addresses the missing highlighting for semantic_text fields when a match_all query is used. We guide users to use semantic highlighting to view their chunks, but highlighting does not currently return any results with match_all. Highlighting is usually unnecessary for non-inference fields with match_all, but it's essential for semantic_text fields to return the generated chunks. This change ensures those highlighting results are returned as expected.

Test cases:

Only text fields

PUT my-index/
{
  "mappings": {
    "properties": {
      "semantic1": {
        "type": "text"
      },
      "semantic2": {
        "type": "text"
      }
    }
  }
}

POST my-index/_doc/1
{
  "semantic1": "Puggles are pugs and beagles",
  "semantic2": "Chiweenies are chihuahuas and dachshunds"
}


GET my-index/_search
{
  "query": {
    "match_all": {}
  }
}

only inference fields

PUT my-semantic-index/
{
  "mappings": {
    "properties": {
      "semantic1": {
        "type": "semantic_text"
      },
      "semantic2": {
        "type": "semantic_text"
      }
    }
  }
}

POST my-semantic-index/_doc/1
{
  "semantic1": "Puggles are pugs and beagles",
  "semantic2": "Chiweenies are chihuahuas and dachshunds"
}

GET my-semantic-index/_search
{
  "query": {
    "match_all": {}
  },
  "highlight": {
    "fields": {
      "semantic1": {
        "type": "semantic",
        "number_of_fragments": 1
      },
      "semantic2": {
        "type": "semantic",
        "number_of_fragments": 1
      }
    }
  }
}

Both infer and non-inference fields

PUT test-index
{
  "mappings": {
    "properties": {
      "test_field": {
        "type": "semantic_text"
      },
      "non_infer_field": {
        "type": "text"
      }
    }
  }
}

PUT test-index/_doc/doc1
{
  "test_field": "these are not the droids you're looking for. He's free to go around",
  "non_infer_field": "this is a non-inference field"
}

GET test-index/_search
{
  "query": {
    "match_all": {}
  },
  "highlight": {
    "fields": {
      "test_field": {
        "type": "semantic",
        "number_of_fragments": 1
      }
    }
  }
}

Both dense vector fields

PUT _inference/text_embedding/my-e5-model
{
  "service": "elasticsearch",
  "service_settings": {
    "num_allocations": 1,
    "num_threads": 1,
    "model_id": ".multilingual-e5-small"
  }
}

PUT my-dense-semantic-index/
{
  "mappings": {
    "properties": {
      "semantic1": {
        "type": "semantic_text",
        "inference_id": "my-e5-model"
      },
      "semantic2": {
        "type": "semantic_text",
        "inference_id": "my-e5-model"
      }
    }
  }
}

POST my-dense-semantic-index/_doc/1
{
  "semantic1": ["Puggles are pugs and beagles", "    ", "now with chunks"],
  "semantic2": ["Chiweenies are chihuahuas and dachshunds"  , "    ", "now with chunks"]
}

GET my-dense-semantic-index/_search
{
  "query": {
    "match_all": {}
  },
  "highlight": {
    "fields": {
      "semantic1": {
        "type": "semantic",
        "number_of_fragments": 3
      },
      "semantic2": {
        "type": "semantic",
        "number_of_fragments": 3
      }
    }
  }
}

Some additional cases

GET my-*/_search
{
  "query": {
    "match_all": {}
  },
  "highlight": {
    "fields": {
      "semantic1": {
        "number_of_fragments": 1
      },
      "semantic2": {
        "number_of_fragments": 1
      }
    }
  }
}

GET my-*/_search
{
  "query": {
    "match_all": {}
  },
  "highlight": {
    "fields": {
      "semantic1": {
        "type": "semantic",
        "number_of_fragments": 1
      },
      "semantic2": {
        "type": "semantic",
        "number_of_fragments": 1
      }
    }
  }
}

// throws error as fields are semantic_text field
GET my-*/_search
{
  "query": {
    "match_all": {}
  },
  "highlight": {
    "fields": {
      "semantic1": {
        "type": "score",
        "number_of_fragments": 1
      },
      "semantic2": {
        "type": "score",
        "number_of_fragments": 1
      }
    }
  }
}

[kderusso]

Nice start, I've added some high level comments on overall organization. We can do a deeper review when it's closer to ready.

[Mikep86]

There's a much simpler way to fix this by updating the semantic highlighter. Take a look at how you can change QueryVisitor in extractDenseVectorQueries and extractSparseVectorQueries:  https://github.com/elastic/elasticsearch/blob/main/x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/highlight/SemanticTextHighlighter.java#L250

[Samiul-TheSoccerFan]

@elasticmachine update branch

[elasticsearchmachine]

Pinging @elastic/search-eng (Team:SearchOrg)

[elasticsearchmachine]

Pinging @elastic/search-relevance (Team:Search - Relevance)

[elasticsearchmachine]

Hi @Samiul-TheSoccerFan, I've created a changelog YAML for you.

[kderusso]

Thanks for these changes, looks great @Samiul-TheSoccerFan !

[kderusso]

Nitpick: Since this is a bug, perhaps we call the changelog summary something like `Fix issue where highlighting was not returned with match_all queries"

[kderusso]

😍 Nice optimization from the first solution!

[Samiul-TheSoccerFan]

@elasticmachine update branch

[Mikep86]

Approach looks good! Caught a copy/paste error in the tests that we should fix though.

[Mikep86]

LGTM! Good clean solution and tests :)

[Samiul-TheSoccerFan]

@elasticmachine update branch

[Samiul-TheSoccerFan]

@elasticmachine update branch

[elasticsearchmachine]

💔 Backport failed

Status	Branch	Result
❌	8.19	Commit could not be cherrypicked due to conflicts

You can use sqren/backport to manually backport by running backport --upstream elastic/elasticsearch --pr 128702

[Samiul-TheSoccerFan]

💚 All backports created successfully

Status	Branch	Result
✅	8.19

Questions ?

Please refer to the Backport tool documentation

[Samiul-TheSoccerFan]

💚 All backports created successfully

Status	Branch	Result
✅	9.0
✅	8.18

Questions ?

Please refer to the Backport tool documentation