Deploying pydantic-docs with    Cloudflare Pages

View logs

The one testing issue that I'm running into currently is that nowHttpUrl, etc aren't direct subclasses ofUrl, so isinstance checks are failing validation...

Maybe, this can be fixed with a union against an isinstance schema, but I hate using unions for internally defined schemas.
We could also subclass fromUrl andMultiHostUrl, but that seems quite messy?

CodSpeed Performance Report

Merging#10766 willnot alter performance

_{Comparingurl-fix-i-think (b7c8485) withmain (678ec30)}

Summary

✅ 44 untouched benchmarks

Coverage report

This PR does not seem to contain any modification to coverable code.

I've fixed the above concerns with a wrap validator, though this admittedly slows validation significantly (see regression above).

My suggestion would be a lazy type adapter (via functools.cache) which is called ininit.

Does this mean we can rip out validation logic on__init__ inpydantic-core?@samuelcolvin@davidhewitt

Happy to fix validation on init in this pr, hopefully that moves this into the "don't need to deal with for a while" state.

We might be able to pull this off with just an after validator, which could be good...

My suggestion would be a lazy type adapter (via functools.cache) which is called ininit.

Does this mean we can rip out validation logic on__init__ inpydantic-core?@samuelcolvin@davidhewitt

Probably not, because it currently needs to assume valid state internally. Maybe in the future, but I wouldn't rush to do it.

I agree with@davidhewitt, and I don't see any particular advantage of ripping out that logic right now.

Probably not, because it currently needs to assume valid state internally. Maybe in the future, but I wouldn't rush to do it.

I think this effectively results in double validation for some cases (see perf degradation above), but I'm ok with this as an intermediate state for now.

we should definitely avoid double validation, I think that's doable.

Here's where I'm having trouble with double validation:

frompydanticimportTypeAdapter,HttpUrlta=TypeAdapter(HttpUrl)http_url=HttpUrl('http://example.com/something')assertstr(http_url)=='http://example.com/something'assertrepr(http_url)=="HttpUrl('http://example.com/something')"asserthttp_url.__class__==HttpUrlasserthttp_url.host=='example.com'asserthttp_url.path=='/something'asserthttp_url.usernameisNoneasserthttp_url.passwordisNonehttp_url_validated=ta.validate_python(http_url)"""pydantic_core._pydantic_core.ValidationError: 1 validation error for function-wrap[wrap_val()]  URL input should be a string or URL [type=url_type, input_value=HttpUrl('http://example.com/something'), input_type=HttpUrl]    For further information visit https://errors.pydantic.dev/2.10/v/url_type"""

So, I think we need to have some sort of isinstnace check in the wrap validator. This is unfortunate though, as we'll no longer respectconfig.revalidate_instances...

I think this is where having acls passed topydantic-core'surl_schema could be helpful to help with this isinstance check that's failing, but I'd really prefer to keep that out of the scope of this fix, and make that change in v2.11. We already tried that once, and things went awry with duplicated validation.

Thanks for the abundance of iterations here@samuelcolvin@davidhewitt.

I think we're moving in the right direction here, next steps include improving performance and optimizing / standardizing how we validate types like this on__init__.