在多值字段上被Solr查询阻止

Question

I have a problem on a logical solr query. Here is my index structure :

"docs":[
        {
    "id":"doc1",
     "type.id": [
              "1",
              "2",
              "3"
            ],
            "type.demande": [
              "non",
              "non",
              "oui"
            ],
            "type.reponse": [
              "oui",
              "oui",
              ""
            ],
            "type.traitement": [
              "oui",
              "oui",
              "non"
            ]

    },
    {
    "id":"doc2",
     "type.id": [
              "1",
              "2",
              "3"
            ],
            "type.demande": [
              "non",
              "non",
              "non"
            ],
            "type.reponse": [
              "oui",
              "oui",
              "non"
            ],
            "type.traitement": [
              "oui",
              "oui",
              ""
            ]
    }
    ]

Each field type.id, demande, reponse and traitement are multivalued. Each type is a group of these 4 fields. Ex. : type id 1 from doc 2 has demande "non", réponse "oui", traitement "oui". I want to do a query of the last value of these fields to filter on values of the last inserted type. For example of solr query :

q=(type.demande:"oui" AND (type.reponse:"" OR type.reponse:"non"))

When I try this query, it seems that solr will search any value matching on the fields. I searched on solr docs, but I didn't find a proposal of "last" or "lastValueOf" which can permit me to do this.

Thanks.

Answer 1

You can't with flat documents, really - but you can do it with child documents . There's an easier solution if this is the only scenario you're querying, and that is to create a secondary value that you filter against - a merged form of these fields that you can query as a single value.

When indexing you'd merge the values into a single field with a separator, for example | , then query it as a single field:

merged_demande_reponse:("oui|non" OR "oui|")

This field should not be processed further in any way. The good thing about this solution is that each document is kept plain and all functionality works out of the box without having to think about child documents each time.

Another solution is to index each combination as a document by itself (which is what child documents does in the background). Each combination of values becomes its own document with a common id:

{
  "id": <generated id>,
  "docid": "doc1",
  "type.id": "1",
  "type.demande": "non"
  "type.reponse": "oui"
  "type.traitement": "oui"
},
{
  "id": <generated id>,
  "docid": "doc1",
  "type.id": "2",
  "type.demande": "non"
  "type.reponse": "oui"
  "type.traitement": "oui"
},
... etc.