FastAPI OpenAPI doesn't handle well small float input parameters

[davidefiocco]

Hi, if you run code along the lines of:

from fastapi import FastAPI
import uvicorn


app = FastAPI(title="Sum one",
              description="Get a number, add one to it",
              version="0.1.0",
              )

@app.post("/compute")
def compute(input: float = 0.0000001): # problematic float default value
    return {'result': input + 1}

if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=8000)

and then

• Open the browser at http://localhost:8000/docs and click and try out the /compute endpoint, then click on the Execute button
• Execution fails because validation of the parameter field (a small float in scientific notation) is not passed, and the libraries automatically format the small float in scientific notation:

To solve, the user needs to manually override the scientific notation of the float in order to execute correctly, as done below:

• Intended behavior would be flawless execution with a float value.

Environment

• OS: Windows (but this should reproduce on Linux as well)
• FastAPI Version: 0.61.0

[SirTelemak]

Try to use Decimal type as it is better working with floating point values

[Kludex]

I don't think that's a FastAPI issue. After @SirTelemak comment I've tried:

from decimal import Decimal
from fastapi import FastAPI

app = FastAPI()

@app.get("/")
def test(num: Decimal):
    return str(type(num))

It doesn't work on swagger:

But the endpoint works as it's supposed to:

(base) ➜  ~ curl "http://localhost:8000?num=1e-7"
"<class 'decimal.Decimal'>"%     

Maybe this can help to understand the problem: https://swagger.io/docs/specification/data-models/data-types/

[tiangolo]

Thanks for the help here @SirTelemak and @Kludex ! 👏 🙇

If that solves the original problem, then you can close this issue @davidefiocco ✔️

Otherwise, if you need to accept scientific notation, as that is not a "standard" number, at least in the terms of OpenAPI, JSON, etc. you could annotate it as a str and then use a dependency to parse it and validate it. Or you could create a custom type with a class that inherits from str and adds custom Pydantic validatdors: https://pydantic-docs.helpmanual.io/usage/types/#custom-data-types

But in general, to accept that notation and work with Swagger UI it would have to be annotated as a string.

Edit 2022-10-19 reading the JSON spec, scientific notation seems to be actually valid numbers.

[davidefiocco]

Thanks @tiangolo and all! Maybe I am missing something, but in my example I am not "proactively" using scientific notation, just a small float:

@app.post("/compute")
def compute(input: float = 0.0000001): # problematic float default value
    return {'result': input + 1}

This however results in swagger/openapi (last time I tried) in a number reformatted in a failing scientific notation.
Is there a way to "inhibit" such reformatting?

[tiangolo]

Interesting! Indeed, I just tried your example and OpenAPI shows the default in scientific notation. Funny enough, I was just checking the JSON spec, and my previous comment is incorrect. That would be a valid number in JSON.

So, the problem is actually that Swagger UI is not accepting something that is indeed a valid number. Maybe you could report it there: https://github.com/swagger-api/swagger-ui/issues/new/choose

I tried searching and I didn't find an issue mentioning scientific notation. And I would think that's a UI bug there.

Sorry for the long delay! 🙈 I wanted to personally address each issue and they piled up through time, but now I'm checking each one in order.

[github-actions]

Assuming the original need was handled, this will be automatically closed now. But feel free to add more comments or create new issues or PRs.