@itamarst FYSA, this PR introduces some API improvements per #105 so you can now do something like

let cddl_schema = CDDL::from_slice(
  r#"
    foo = {
      bar: tstr
    }
  "#.as_bytes(),
)
.unwrap();

let documents = [r#"{ "bar": "foo" }"#, r#"{ "bar": "foo2" }"#];

documents
  .iter()
  .all(|doc| cddl_schema.validate(doc.as_bytes(), None).is_ok());

Let me know what you think

I can probably create a branch off of your branch, will try that.

That is, create a branch of pycddl that uses branch of cddl, to try the new API.

I tested CDDL::from_slice with my current approach of parsing the schema for every validation. It works, simplified the code slightly.

(The lifetime requirements is still making interop with PyO3 hard, though, since it's explicitly not supported: PyO3/pyo3#502. So #104 is still blocker on keeping pre-parsed CDDL around.)

For reference, here's the branch using this branch: https://gitlab.com/tahoe-lafs/pycddl/-/blob/12-persistent-schema-instead-of-parsing-schema-on-each-validation/src/lib.rs

Thanks for testing! Yea, I'll be working through #104 once I get this PR finished up.