When using Unions inside BaseModel ints and booleans are converted to str

[JakeSummers]

System Info:

• OS: Linux
• Python version 3.7.1
• Pydantic version 0.21

Suppose I have a class and tests like this:

from typing import Union

from pydantic import BaseModel

class Simple(BaseModel):
    value: Union[int, str, bool, None]


def test_simple_booleans_are_not_lost():

    simple = Simple(value=True)
    print(f"simple.value is of type {type(simple.value)}")
    assert simple.value == True


def test_simple_ints_are_not_lost():
    simple = Simple(value=404)
    print(f"simple.value is of type {type(simple.value)}")
    assert simple.value == 404


def test_simple_strings_are_not_lost():
    simple = Simple(value="Jake")
    print(f"simple.value is of type {type(simple.value)}")
    assert simple.value == "Jake"


def test_simple_Nones_are_not_lost():
    simple = Simple(value=None)
    print(f"simple.value is of type {type(simple.value)}")
    assert simple.value == None

This works great!

Oddly, if I change the definition of Simple from:

class Simple(BaseModel):
    value: Union[int, str, bool, None]

To:

class Simple(BaseModel):
    value: Union[str, bool, None, int]

This causes some of the tests to fail. Specifically, the test_simple_booleans_are_not_lost and test_simple_ints_are_not_lost will fail. This is because the values are converted to strings. Output from the test:

simple.value is of type <class 'str'>

.../test_simple.py:15 (test_simple_ints_are_not_lost)
404 != 404

Expected :404
Actual   :404
 <Click to see difference>

def test_simple_ints_are_not_lost():
        simple = Simple(value=404)
        print(f"simple.value is of type {type(simple.value)}")
>       assert simple.value == 404
E       AssertionError: assert '404' == 404
E        +  where '404' = <Simple value='404'>.value

test_simple.py:19: AssertionError

---

The documentation on typing.Union specifies that order of items in a Union is irrelevant.

[samuelcolvin]

What the docs say is

when comparing unions, the argument order is ignored

In the case of pydantic the order is not ignored, we could better document this, but I don't think it's worth fixing since it would require lots more checking and would slow things down.

[samuelcolvin]

This has it's own section in the new docs.