Ability to try to typecast to primitive types in record::asdict about faust

[ask]

Hey Ruby! Sorry, I missed this for so long. This sounds like a good idea,

Note though that DateTimes are already automatically converted to str by the JSON serializer, so it doesn't have to convert it at the as_dict point as when we send it to our JSON serializer that will convert it to ISO8601 strings.

I would like to support arbitrary types, similar to isodates, but I'm not sure of the best way to do that.

We could have an API more like data classes and attrs, by having users specify FieldDescriptors
with more capabilities, like how to serialize/deserialize/verify the data

class Account(Record):
    account_id: UUID = Field(to_value=uuid_to_str, from_value=str_to_uuid, required=True)
    amount: Decimal = DecimalField(max_decimal_places=1000)

Alternatively, we could have optional callbacks with reserved names like this:

class Account(Record):
  account_id: UUID

  def account_id_to_value(self, account_id: UUID) -> str:
      return str(account_id)

 def account_id_from_value(self, account_id: str) -> UUID:
    return UUID(account_id)

I guess the Field is better since you can specify the above as merely:

class Account(Record):
    account_id: UUID = Field(str, UUID, required=True)

and you could subclass it:

class UUIDField(Field):
    to_value = str
    from_value = UUID
    required = True
   

[ask]

For reference here's our custom JSONEncoder class that adds datetime and others as built-in JSON data types:
https://github.com/robinhoodmarkets/faust/blob/61cb72e13d571fdc16274140bdcd2ee3a483d2ae/faust/utils/json.py#L68-L91

It's usually faster to do it this way, as then you don't have to do yet another recursive pass over all the fields. At least that was true when I added it; I think we are at this point already doing two traversals to serialize.

[vineetgoel]

Closing this for now as this is not something we plan on adding into Faust in the near future.