I am trying to generate JSON Schema from python data classes that contain lists of other data classes as parameters. I don't really care about any functionality from pydantic other than generating json schema, since I am just using the output to auto-generate a web frontend. So solutions using other libraries are fine.
from typing import List
from dataclasses import dataclass
import pydantic
@dataclass()
class SomeParameters:
a: int = 5
@dataclass()
class SomeMoreParameters:
another: List[SomeParameters]
pydantic_cls = pydantic.dataclasses.dataclass(SomeMoreParameters)
schema = pydantic_cls.__pydantic_model__.schema()
This code errors with the following stacktrace (you can use SKIP_CYTHON=1 pip install --no-cache-dir --no-binary :all: pydantic to install pydantic so that you can debug into it):
Traceback (most recent call last):
File "/home/veith/Projects/job-ui-parametrization/debug.py", line 17, in <module>
schema = pydantic_cls.__pydantic_model__.schema()
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/main.py", line 647, in schema
s = model_schema(cls, by_alias=by_alias, ref_template=ref_template)
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/schema.py", line 185, in model_schema
m_schema, m_definitions, nested_models = model_process_schema(
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/schema.py", line 617, in model_process_schema
m_schema, m_definitions, nested_models = model_type_schema(
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/schema.py", line 658, in model_type_schema
f_schema, f_definitions, f_nested_models = field_schema(
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/schema.py", line 258, in field_schema
f_schema, f_definitions, f_nested_models = field_type_schema(
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/schema.py", line 498, in field_type_schema
items_schema, f_definitions, f_nested_models = field_singleton_schema(
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/schema.py", line 848, in field_singleton_schema
return field_singleton_sub_fields_schema(
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/schema.py", line 736, in field_singleton_sub_fields_schema
return field_type_schema(
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/schema.py", line 563, in field_type_schema
f_schema, f_definitions, f_nested_models = field_singleton_schema(
File "/home/veith/Projects/job-ui-parametrization/venv/lib/python3.9/site-packages/pydantic/schema.py", line 947, in field_singleton_schema
raise ValueError(f'Value not declarable with JSON Schema, field: {field}')
ValueError: Value not declarable with JSON Schema, field: name='_another' type=SomeParameters required=True
Process finished with exit code 1
Now according to JSON Schema documentation an array of things is valid.
The pydantic documentation uses list of integers, which also works for me.
I would expect that I can also use custom types in a list. The documentation further states Nested dataclasses are supported both in dataclasses and normal models.
Am I doing something wrong/not doing something that I should be doing? Or am I misunderstanding something, and this is actually not allowed? What are some alternatives if this is the case?
This is a bug introduced with pydantic 1.9. I have contributed a pull request fixing this issue here: https://github.com/samuelcolvin/pydantic/pull/3819