Excellent pydantic support for sized containers

[antonagestam]

Currently there are two issues:

• Pydantic doesn't expose the "sub-field" type to __modify_schema__ so we can't add e.g. {items: {"type": "integer"}} for a NonEmpty[int]. See discussion. Fixed in Add optional field argument to __modify_schema__ pydantic/pydantic#3434.
• The PhantomSized type is "too" flexible. Consider adding an intermediary BoundedSize or similar that takes class arguments min/max and reflects that in minItems/maxItems.

[flaeppe]

Not sure if I'm running in to the mentioned Caveat regarding pydantic and NonEmpty below? Or is it a separate issue I'm running in to? (Feels a little bit like I'm running in to some problem with sub-field type not being available)

from phantom.sized import NonEmpty
from pydantic import BaseModel

class M(BaseModel):
    x: NonEmpty[list[str]]

M(x=["abc"])

This is the error I get:

pydantic.error_wrappers.ValidationError: 1 validation error for M
x
  Could not parse phantom.sized.NonEmpty from ['abc'] (type=type_error)

[antonagestam]

@flaeppe The documented caveat is only regarding JSON schema validation, e.g. minItems, maxItems and items.type aren't automatically set.

I think you will want to use NonEmpty[str] and not NonEmpty[list[str]], there's related discussion in this issue: #160