How to create a field mapping in Azure Search with a complex targetField

Question

I use the Azure Search indexer to index documents from a MongoDB CosmosDB which contains objects with fields named _id. As Azure Search does not allow underscores at the beginning of a field name in the index, I want to create a field mapping.

JSON structure in Cosmos --> structure in index

{
  "id": "test"
  "name": "test",
  "productLine": {
     "_id": "123",       --> "id": "123"
     "name": "test"
  }
}

The documentation has exactly this scenario as an example but only for a top level field.

"fieldMappings" : [ { "sourceFieldName" : "_id", "targetFieldName" : "id" } ]}

I tried the following:

"fieldMappings" : [ { "sourceFieldName" : "productLine/_id", "targetFieldName" : "productLine/id" } ] }

that results in an error stating:

Value is not accepted. Valid values: "doc_id", "name", "productName".

What is the correct way to create a mapping for a target field that is a subfield?

Answer 1

It's not possible to directly map subfields. You can get around this by adding a Skillset with a Shaper cognitive skill to the indexer, and an output field mapping.

You will also want to attach a Cognitive Services resource to the skillset. The shaper skill doesn't get billed, but attaching a Cognitive Services resource allows you to process more than 20 documents per day.

Shaper skill

{
  "@odata.type": "#Microsoft.Skills.Util.ShaperSkill",
  "context": "/document",
  "inputs": [
    {
      "name": "id",
      "source": "/document/productLine/_id"
    },
    {
      "name": "name",
      "source": "/document/productLine/name"
    }
  ],
  "outputs": [
    {
      "name": "output",
      "targetName": "renamedProductLine"
    }
  ]
}

Indexer skillset and output field mapping

"skillsetName": <skillsetName>,
"outputFieldMappings": [
    {
        "sourceFieldName": "/document/renamedProductLine",
        "targetFieldName": "productLine"
    }
]