org.apache.avro.AvroTypeException: Unknown union branch

Question

I'm using this Avro schema:

prices-state.avsc

{
    "namespace": "com.company.model",
    "name": "Product",
    "type": "record",
    "fields": [
        {
            "name": "product_id",
            "type": "string"
        },
        {
            "name": "sale_prices",
            "type": {
                "name": "sale_prices",
                "type": "record",
                "fields": [
                    {
                        "name": "default",
                        "type": {
                            "name": "default",
                            "type": "record",
                            "fields": [
                                {
                                    "name": "order_by_item_price_by_item",
                                    "type": [
                                        "null",
                                        {
                                            "name": "markup_strategy",
                                            "type": "record",
                                            "fields": [
                                                {
                                                    "name": "type",
                                                    "type": {
                                                        "name": "type",
                                                        "type": "enum",
                                                        "symbols": ["margin", "sale_price"]
                                                    }
                                                }
                                            ]
                                        }
                                    ]
                                },
                                {"name": "order_by_item_price_by_weight", "type": ["null", "string"]},
                                {"name": "order_by_weight_price_by_weight", "type": ["null", "string"]}
                            ]
                        }
                    }
                ]
            }
        }
    ]
}

It validates properly on this website so I'm assuming the schema is valid.

I'm having issues building a JSON file that should then be encoded using the above schema.

I'm using this JSON for some testing:

test.json

{
    "product_id": "123",
    "sale_prices": {
        "default": {
            "order_by_item_price_by_item": {
                "markup_strategy": {
                    "type": {"enum": "margin"}
                }
            },
            "order_by_item_price_by_weight": null,
            "order_by_weight_price_by_weight": null
        }
    }
}

When running java -jar avro-tools-1.8.2.jar fromjson --schema-file prices-state.avsc test.json I get:

Exception in thread "main" org.apache.avro.AvroTypeException: Unknown union branch markup_strategy

I read here that I have to wrap things inside unions because of JSON Encoding so I tried different combinations but no one seemed to work.

Answer 1

It was a namespace resolution issue. Take this simplified schema as an example:

test.avsc

{
    "name": "Product",
    "type": "record",
    "fields": [
        {
            "name": "order_by_item_price_by_item",
            "type": [
                "null",
                {
                    "type": "record",
                    "name": "markup_strategy",
                    "fields": [{
                        "name": "type",
                        "type": {
                            "name": "type",
                            "type": "enum",
                            "symbols": ["margin", "sale_price"]
                        }
                    }]
                }
            ]
        }
    ]
}

With the following JSON it validates just fine

test.json

{
    "order_by_item_price_by_item": {
        "markup_strategy": {
            "type": "margin"
        }
    }
}

Now if you were to add a namespace on top of your schema like

test.avsc

{
    "namespace": "test",
    "name": "Product",
    "type": "record",
    "fields": [
    ...

Then you would need to change your test.json as well or you'll get

Exception in thread "main" org.apache.avro.AvroTypeException: Unknown union branch markup_strategy

final_test.json

{
    "order_by_item_price_by_item": {
        "test.markup_strategy": {
            "type": "margin"
        }
    }
}

So when inside a union type and you're JSON encoding an Avro's named type (record, fixed or enum) where the user-specified name is used then the name of that type needs to be prepended with the namespace name too for resolution.

More on namespaces and JSON encoding.

Answer 2

The answer from Francesco is great and explains the issue!
Just in case anyone(or even my future self) needs another example with optional fields and namespaces:

1. schema - StackOverflowSchema.avsc

{
  "type": "record",
  "name": "StackOverflowExampleSchema",
  "namespace": "com.schemata.stackoverflow",
  "fields": [
    {
      "name": "exampleEntity",
      "type": {
        "type": "record",
        "name": "ExampleEntity",
        "namespace": "com.stackoverflow",
        "fields": [
          {
            "name": "description",
            "type": ["null", "string"]
          },
          {
            "name": "currentAmount",
            "type": {
              "type": "record",
              "name": "MonetaryAmount",
              "namespace": "com.acme.common",
              "fields": [
                {
                  "name": "currencyCode",
                  "type": ["null", {
                    "type": "enum",
                      "name": "CurrencyCode",
                      "symbols": ["AED", "USD"],
                      "default": "USD"
                    }
                  ]
                },
                {
                  "name": "amount",
                  "type": {
                    "type": "int"
                  }
                }
              ]
            }
          },
          {
            "name": "totalAmount",
            "type": ["null", "com.acme.common.MonetaryAmount"]
          }
        ]
      }
    }
  ]
}

2. sample file - stackoverflow.json

{
  "exampleEntity": {
    "description": {
      "string": "some optional description"
    },
    "currentAmount":{
      "amount": 10",
      "currencyCode": {
        "com.acme.common.CurrencyCode": "USD"
      }
    },
    "totalAmount": {
      "com.acme.common.MonetaryAmount": {
        "amount": 20,
        "currencyCode": {
          "com.acme.common.CurrencyCode": "USD"
        }
      }
    }
  }
}

Both currentAmount and totalAmount have com.acme.common.MonetaryAmount type.
While the former is a required field the latter can be null.
Also within the MonetaryAmount we have required (amount) and optional (currencyCode).
The command below to produce the avro message:

java -jar avro-tools-1.11.0.jar fromjson --schema-file ./StackOverflowSchema.avsc stackoverflow.json > stackoverflow.avro