Samuel Colvin

Host is required for `MultiHostUrl`

8

In theory we could have a `MultiHostUrl`, but with host optional. Example from a postgresql dsn: `postgresql://derp@/databasename?host=/var/run/postgresql` See https://github.com/pydantic/pydantic/issues/3900

Should we change how we decode Base64?

1

Based on https://github.com/marshallpierce/rust-base64/issues/105 I imagine our current use of the base64 crate could annoy some people.

~~See https://github.com/pydantic/pydantic/discussions/6893#discussioncomment-6561902~~ ~~(EDIT by @davidhewitt - the above related to validators, and is fixed after #844)~~ It would be nice if errors within functional serializers showed a traceback with both...

`EmailValidator` type

13

Using [mail_parser](https://docs.rs/mail-parser/latest/mail_parser/) I guess. Rationale: * remove the `email-validator` optional dependency from pydantic, and its descendant dependencies - `dnspython` and `idna` * should be faster * the `NameEmail` `"My Name...

remove calls to `__instancecheck__` on `ModelValidator`

4

As of #492 `ModelValidator` is checking if the input is an instance of `BaseModel` before doing anything else. Because `BaseModel` is an abc, the `isinstance` rust call is eventually calling...

name for constrained string validators should chanage

1

To just `str` if the constraints are from config, then if they're local maybe to `str[max_len=3]` etc.?

Replace `SchemaFilter` on `TypedDictSerializer`

2

We currently store a `SchemaFilter` on `TypedDictSerializer` and use that to exclude fields, we should instead use an `exclude` bool on `TypedDictField` to avoid unnecessary hashing. https://github.com/pydantic/pydantic-core/blob/cd5cd657cc8a57f6310b3fbb19d9657136087eec/src/serializers/type_serializers/typed_dict.rs#L70

`Arguments` serializer

1

I guess by default should take a dict and operate like a `TypeDict` serializer.

new-class validator discussion

3

@samuelcolvin It seems to me that the approach you are describing (especially given point 2) is along the lines of what I'm thinking of as the "schema-based" approach (i.e., where...