Request: Rename fields in AvroBaseModel

[lzulauf]

Is your feature request related to a problem? Please describe.
I would like a way to set the name of fields in AvroBaseModels separate from the field name defined in python.

I'm trying to use AvroBaseModel to represent payloads like the following:

{
    "schema": {
        "value": "my_schema_id"
    }
}

The desired avro and jsonschemas for this look like:

desired_avroschema = {
  "type": "record",
  "name": "MyModel",
  "fields": [
    {
      "name": "schema",
      "type": {
        "type": "record",
        "name": "Schema",
        "fields": [
          {
            "name": "value",
            "type": "string"
          }
        ]
      }
    }
  ]
}

desired_jsonschema = {
  "title": "MyModel",
  "type": "object",
  "properties": {
    "schema": {
      "title": "Schema",
      "type": "object",
      "properties": {
        "value": {
          "title": "Value",
          "type": "string"
        }
      },
      "required": [
        "value"
      ]
    }
  },
  "required": [
    "schema"
  ]
}

(note that the jsonschema can use a definition reference for the section with "title": "Schema")

In order to model this, I might write code like this:

class Schema(AvroBaseModel):
    value: str
    
class MyModel(AvroBaseModel):
    schema: Schema

but that would result in the following error from pydantic:

NameError: Field name "schema" shadows a BaseModel attribute; use a different field name with "alias='schema'".

We can fix this by giving the field a different name and then using a pydantic Config fields dictionary to set the name that the schema should use:

class Schema(AvroBaseModel):
    value: str
    
class MyModel(AvroBaseModel):
    schema_: Schema
    
    class Config:
        fields = {"schema_": "schema"}

This fixes the jsonschema that we get with MyModel.json_schema(), but the avro schema we get with MyModel.avro_schema() still uses the name of the python field (schema_), resulting in the following:

{
  "type": "record",
  "name": "MyModel",
  "fields": [
    {
      "name": "schema_",
      "type": {
        "type": "record",
        "name": "schema",
        "fields": [
          {
            "name": "value",
            "type": "string"
          }
        ]
      }
    }
  ]
}

Describe the solution you'd like
I would like a mechanism that allows renaming of fields, similar to what can be done for the jsonschema via pydantic Config classes. Something like:

class MyModel(AvroBaseModel):
    schema_: Schema
    
    class Config:
        fields = {"schema_": "schema"}

    class Meta:
        fields = {"schema_": "schema"}

(or ideally, a single config class that drives both systems so that they always produce equivalent schemas)

Describe alternatives you've considered
I've tried setting schema_name and alias_nested_items, but these set the name of the record, not the name of the field referencing the record.

Additional context
Apologies if there is a mechanism to do this already, I could not find it. Thank you!

[marcosschroh]

Hi @lzulauf

I will try to fix this asap

[kevinjacobs-delfi]

FYI, Pydantic allows fields to be renamed using alias or serialization_alias field metadata.