Check with regex.match() evaluated before "required attribute missing"

[excalq]

Bug Report

When a schema contains a check, using regex.match(attribute ...), this appears to be evaluated before the existence of a non-optional attribute in the instantiated object.

If the required attribute is absent, a compiler error is thrown regarding the schema's check regex.match(attribute) itself, rather than notification that the user's object is lacking a required attribute. The more obvious error would be raising that the attribute is missing (attribute .. is required and can't be None or Undefined).

As it were, the current error causes a user to debug a problem with the check. In my case the missing field was in an inherited schema's object, compounding my confusion.

The reproduction will easily illustrate this.

1. Minimal reproduce step (Required)

Playground Link

import regex

schema CrossplaneResource:
  name: str
  description: str
  check:
    regex.match(description, "[a-zA-Z0-9+=,.@-_]"), "Description must match the regex [a-zA-Z0-9+=,.@-_]"

ice: CrossplaneResource = {
  name: 'Vanilla Ice'
  # description: 'Ice Ice Baby' # Works with this field uncommented
}

2. What did you expect to see? (Required)

The missing field should fail with an error before the check dies on the missing description variable.

EvaluationError
  --> /tmp/sandbox1241309759/prog.k:15:1
   |
15 |   name: 'Vanilla Ice'
   |  attribute 'description' of CrossplaneResource is required and can't be None or Undefined
   |

3. What did you see instead (Required)

EvaluationError
  --> /tmp/sandbox1225202080/prog.k:11:1
   |
11 |     regex.match(description, "[a-zA-Z0-9+=,.@-_]"), "Description must match the regex [a-zA-Z0-9+=,.@-_]"
   |  match() missing 2 required positional arguments: 'string' and 'pattern'
   |

4. What is your KCL components version? (Required)

0.8.7-darwin-arm64

[excalq]

Another aspect to this is being unsure if it's possible to do validation on optional fields (if set). I assumed I could do:

check:
    description == None or regex.match(description, ...), 'The error message'

But this produces the same error reported above. My workaround is having the field be non-optional, and setting a default description value in the schema.

[Peefy]

You can write the following code for the optional attribute using the if filter, and I will change the optional attribute check time in some PR.

check:
     regex.match(description, ...) if description, 'The error message'

[excalq]

Well that makes a lot of sense, thinking Pythonically! Thanks a ton.