Support parameter type typing.Dict[str, typing.Any]

[tekumara]

My CLI app can use a profile key, which is used to load the appropriate config:

def load_profile(profile: str) -> Dict[str, Any]:
    ...

@app.command()
def describe(
    config: Dict[str, Any] = typer.Option(None, "--profile", help="Profile in the config file to use", callback=load_profile),
) -> List[Dict[str, Any]]:

However, this errors with:

RuntimeError: Type not yet supported: typing.Dict[str, typing.Any]

The solution you would like

The ability to support Dict as a type, when returned from a callback function.

[tekumara]

The workaround, for now, is to remove the type hint from the config param (which also disables type checking).

[pavan-uppari]

Handling Dict type.

Issue - If we pass dict directly from terminal, it will be coming as string to us.

Fix: We can use pydantic Json with given dict type and use pydantic TypeAdapter to validate data and get the actual value

Usage

import json
from pydantic import TypeAdapter
from pydantic.types import Json

dict_type = dict[str, int]  #type from method annotations
data = '{"key": 1, "key2": 2}'  #input from terminal - json data

type_adapter = TypeAdapter(Json[dict_type])
t2 = type_adapter.validate_python(data, strict=True)
print(t2) #pass this value to function

Usage with typer

def main(data: dict[str, int]):
    print(f"data is {data}")

$ typer main.py run {"key": 1, "key2": 2}
data is {"key": 1, "key2": 2}

NOTE: User should provide json data as input

@tiangolo If this is ok, we can follow the same for classes as types, let user provide data as dict and pydantic will handle the validation

Inspired from pydantic/pydantic#8446 (comment)