Incorrect type annotations for untyped JSON schema arrays

[ncoghlan]

Describe the bug

When a JSON schema field is declared as "type": "array" with no further details, the generated data model code fails typechecking:

src/_models/__init__.py:1990: error: Missing type parameters for generic type "List"  [type-arg]
src/_models/__init__.py:1991: error: Missing type parameters for generic type "List"  [type-arg]

To Reproduce

Example schema snippet:

    "system/rpc/listDownloadedModels/returns": {
      "type": "array"
    },

Used commandline:

$ datamodel-codegen --input /path/to/schema.json --input-file-type jsonschema --output /path/to/src/_models/__init__.py \
    --output-model-type pydantic_v2.BaseModel --use-annotated --use-union-operator

The generated model that fails typechecking:

class SystemRpcListDownloadedModelsReturns(RootModel[List]):
    root: List

Expected behavior

Untyped arrays should be effectively typed as List[Any] rather than List

Version:

• OS: Linux
• Python version: 3.12
• datamodel-code-generator version: 0.26.3

Additional context

--use-generic-container-types just shifted the error to complaining about Sequence instead.

--use-standard-collections crashed mypy outright, so I didn't investigate that any further.

The offending schema field not having any type information at all regarding the permitted elements is probably a bug in its own right, but not one that has anything to do with datamodel-code-generator.

[ncoghlan]

I've since tried the msgspec output format on this problem, and it also does the wrong thing:

SystemRpcListDownloadedModelsReturns = List

So presumably the problem is occurring earlier, when the JSON schema entry is converted into a Python type hint (a bare array should be mapping to List[Any] rather than List)

[apinkney97]

I'm having the same problem, on v0.31.2, with Python 3.13.

Unlike the above, MyPy doesn't crash when I use --use-standard-collections, but either way I get bare list or List, and no [Any].

I'm using the swagger spec output by a FastAPI app that itself uses Pydantic models to generate the schema. These problematic fields are defined as list[Any] in the original models.