Create dynamic Pydantic model with typed optional values

Question

I want to create a dynamic model from my data, create JSON from it, and load it in another location.

I can't find a way to define a field in a way, that should it is optional, but when data exists for that field - it would be validated.

This is for required field: fields[col_name] = (data_type, None)

# resolve types for data
data_type = resolve_type(data)
required = is_required(data)
if required:
  fields[col_name] = (data_type, ...)
else:
  fields[col_name] = (data_type, None) <--- unclear

...

pydantic.create_model(name, **fields)

The above configuration generates JSON model that makes fields optional and typed, but then I validate by using the input data I can't pass None values - '$.inputs.0.Field', 'message': "None is not of type 'string'"

So my question - how to declare a field that would validate input, but only when it's not None. And in such a manner, that I could create JSON schema and load it in another location.

I had the same issue as well (with predefined models). At the moment there is no support for fields that are optional, as it seems that for the author either you have all the data or none. This issue could help github.com/samuelcolvin/pydantic/issues/990 — lsabi
– lsabi, Commented Sep 8, 2020 at 9:25
Sorry, I tend to face similar problems with fastAPI when I want only certain fields to be updated, I have to resubmit the whole data object and verify that some old things match... — lsabi
– lsabi, Commented Sep 8, 2020 at 13:20

Martynas Jurkus · Accepted Answer · 2020-09-08 13:20:37Z

2

Implemented workaround based on suggestions in https://github.com/samuelcolvin/pydantic/issues/990

optional_fields = list()
if required:
  fields[col_name] = (data_type, ...)
else:
  fields[col_name] = (data_type, None)
  optional_fields.append(col_name)

model = pydantic.create_model(name, **fields)

def schema_extra(schema, model):
    for column in optional_fields:
        original_type = schema["properties"][column]["type"]
        schema["properties"][column].update({"type": ["null", original_type]})

model.__config__.schema_extra = schema_extra

schema_json = model.schema_json()

answered Sep 8, 2020 at 13:20

Martynas Jurkus

9,30113 gold badges62 silver badges104 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

alex li · Accepted Answer · 2020-12-07 17:43:20Z

0

Foo = create_model("Foo", foo = (int, None))

Foo(foo=None)
# Foo(foo=None)

Foo()
# Foo(foo=None)

Foo(foo=3)
# Foo(foo=3)

Is this your desired result?

answered Dec 7, 2020 at 17:43

alex li

76410 silver badges10 bronze badges

2 Comments

Ramon Dias Over a year ago

This is not dynamic since you know that you are receiving foo. What if I don't know what key is coming in a json?

alex li Over a year ago

See Martynas Jurkus's answer, you can use a dict define the code, and use **kwargs as the arguments of create_model

Collectives™ on Stack Overflow

Create dynamic Pydantic model with typed optional values

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related