Generated data has circular imports?

Describe the bug

I’m trying to use datamodel-codegen to produce Pydantic models from OpenShift resources. While the command completes successfully and generates a hierarchy of Python modules, attempting to import anything results in:

>>> import generated
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/lars/projects/operate-first/acm-clusterdeployment-generator/generated/__init__.py", line 13, in <module>
    from .com.github.openshift.api.image import v1
  File "/home/lars/projects/operate-first/acm-clusterdeployment-generator/generated/com/github/openshift/api/image/v1.py", line 11, in <module>
    from ...... import Annotations9, Blobs, Images, SignedClaims
ImportError: cannot import name 'Annotations9' from partially initialized module 'generated' (most likely due to a circular import) (/home/lars/projects/operate-first/acm-clusterdeployment-generator/generated/__init__.py)

To Reproduce

I’ve started with the OpenShift schema available from https://<your_openshift_server>/openapi/v2. This produces a 17MB output file (here if you want to see it). I save this output to openshift.json and then run:

datamodel-codegen --input openshift.json --input-file-type jsonschema --output generated/

Example schema:

• https://gist.github.com/58ae7b9c912dd575b2858731e6335c50

Used commandline:

$ datamodel-codegen --input openshift.json --input-file-type jsonschema --output generated/

Expected behavior

I expect to be able to do something like:

from generated.io.k8s.api.core.v1 import Secret

Version:

• OS: Linux (fedora 33)
• Python version: 3.9.2
• datamodel-code-generator version: 0.9.4

[harshach]

Never mind! this is an issue on my schema side. Thanks

[koxudaxi]

Thank you for creating this issue. I have found a bug. This problem is related to the issue https://github.com/koxudaxi/datamodel-code-generator/issues/244 I will fix it.

deep dive into the problem

Annotations9 is referenced from generated.com.github.openshift.api.image.TagReference

class TagReference(BaseModel):
    annotations: Optional[Dict[str, Annotations9]] = Field(
        None,
        description='Optional; if specified, annotations that are applied to images retrieved via ImageStreamTags.',
    )

I search the description from the schema.

"com.github.openshift.api.image.v1.TagReference":
  {"description":"TagReference specifies optional annotations for images using this tag and an optional reference to an ImageStreamTag, ImageStreamImage, or DockerImage this tag should track.",
   "type":"object",
   "required":["name"],
   "properties":
      {"annotations":
          {"description":
            "Optional; if specified, annotations that are applied to images retrieved via ImageStreamTags.",
             "type":"object",
             "additionalProperties":{"type":"string"}
             },

The code-generator detects the model as a modular model. "com.github.openshift.api.image.v1.TagReference": And, the model will be generated in in com.github.openshift.api.image.v1 But, a model (a.k.a Annotations9) will be generated as no-modular model on top-level. So, the relation causes the ImportError Annotations9 have to be generated in com.github.openshift.api.image.v1.