OpenAPI page cannot load if Schema includes lt=inf [BUG]

[vdwees]

Describe the bug

Pydantic allows inf, -inf values to be passed to a model. To exclude these, I used the lt param:

import math
from pydantic import BaseModel, Field

class MyModel(BaseModel):
    myfloat: float = Field(..., lt=math.inf)

The openapi page looks like this:

And on the server side, a runtime error:

...
File "/usr/local/lib/python3.7/site-packages/fastapi/applications.py", line 106, in openapi
return JSONResponse(self.openapi())
File "/usr/local/lib/python3.7/site-packages/starlette/responses.py", line 42, in __init__
self.body = self.render(content)
File "/usr/local/lib/python3.7/site-packages/starlette/responses.py", line 151, in render
separators=(",", ":"),
File "/usr/local/lib/python3.7/json/__init__.py", line 238, in dumps
**kw).encode(obj)
File "/usr/local/lib/python3.7/json/encoder.py", line 199, in encode
chunks = self.iterencode(o, _one_shot=True)
File "/usr/local/lib/python3.7/json/encoder.py", line 257, in iterencode
return _iterencode(o, 0)
ValueError: Out of range float values are not JSON compliant

This is a similar issue as #459, except that the invalid value is in the schema itself. The FastAPI.default_response_class is not used to serialize the openapi.json, and this IMO the correct behaviour, but it means the solution proposed in #459 does not apply here. As the model is used in other places outside of fastapi, it does not make sense to remove the lt kwarg.

[vdwees]

After further investigation, this could be a pydantic issue, as MyModel.schema() is:

{
  "properties": {
    "myfloat": {
      "exclusiveMaximum": inf,
      "title": "Myfloat",
      "type": "number"
    }
  },
  "required": [
    "myfloat"
  ],
  "title": "MyModel",
  "type": "object"
}

The serialization is failing on the inf in the schema. So the error emerges in fastapi, but might best be handled in pydantic. I raised the issue there as well: pydantic/pydantic#1417

[chris-allnutt]

@vdwees i'd agree it's a pydantic issue. JSON does allow Infinity, and -Infinity as values, I don't know if those are swagger compliant tho.

This is valid JSON

{ "to_infinity_and_beyond": Infinity}

[retnikt]

±Infinity (nor NaN) are definitely not valid JSON.

[vdwees]

@retnikt take a look at pydantic/pydantic#1417

Hope to have this issue fixed soon

[vdwees]

pydantic/pydantic#1422 was merged, should this issue be closed now, or should it wait for a pydantic release?

[tiangolo]

Thanks for the discussion here everyone! ☕

I think we can close it here as it would be a Pydantic issue more than FastAPI.

And whenever there's a new release, it will be available to be used with FastAPI as well.

So I'm gonna close it now, but feel free to add more comments or new issues.

[alex-r-bigelow]

7 months later, and this still seems to be a problem, i.e.:

@router.get('/infinity')
def test_infinity():
    return float('inf')

still throws ValueError: Out of range float values are not JSON compliant

As far as I can tell, it looks like the pydantic PR was merged and is part of the current version (1.7.3), but I'm not familiar enough with FastAPI's internals to file a sensible bug elsewhere... any hints?

[danieljfarrell]

@alex-r-bigelow I have the same problem, did you figure out a work-around?

[alex-r-bigelow]

Not yet; for now I'm making deep copies of objects and manually converting any nan, inf, and -inf values to strings

[danieljfarrell]

In my case this was because FastAPI was parsing the python object to JSON automatically.

If I set the response_class to be JSONResponse and manually convert the python object to JSON (for example my using Pydantic) then it works fine.

I guess FastAPI default parser will not allow illegal JSON values when parsing, who’d a thought it!