Remove duplicate validation, replace assumed validation on AsdfFile.__init__ with AsdfFile.validate

[braingram]

This PR removes duplicate validation during to_asdf:

roman_datamodels/src/roman_datamodels/datamodels/_core.py

Lines 219 to 223 in 70f05e8

	def to_asdf(self, init, *args, **kwargs):
	with validate.nuke_validation(), _temporary_update_filename(self, Path(init).name):
	asdf_file = self.open_asdf(**kwargs)
	asdf_file.tree = {"roman": self._instance}
	asdf_file.write_to(init, *args, **kwargs)

The assignment to tree triggers the first validation and the write_to triggers a second.

This PR also updates several uses of AsdfFile.__init__ to validate tree contents (replacing it with an explicit call to AsdfFile.validate).

These changes are to prepare roman_datamodels for an upcoming deprecation in asdf (see asdf-format/asdf#1691).

Regression tests running with no errors: https://plwishmaster.stsci.edu:8081/job/RT/job/Roman-Developers-Pull-Requests/515/

Checklist

• Added entry in CHANGES.rst under the corresponding subsection
• updated relevant tests
• updated relevant documentation
• Passed romancal regression testing on Jenkins / PLWishMaster. Link: https://plwishmaster.stsci.edu:8081/job/RT/job/romancal/XXX/

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (ee11c53) 97.44% compared to head (8bfd39a) 97.50%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #301      +/-   ##
==========================================
+ Coverage   97.44%   97.50%   +0.06%     
==========================================
  Files          28       29       +1     
  Lines        2698     2726      +28     
==========================================
+ Hits         2629     2658      +29     
+ Misses         69       68       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[WilliamJamieson]

This seems a bit verbose. Is there something wrong with:

self._asdf = asdf.AsdfFile({"roman": self._instance})
self._asdf.validate()

[braingram]

Thanks for taking a look and good question. Unfortunately, the code you proposed would validate the instance twice. Once during AsdfFile.__init__ and again during AsdfFile.validate. Also, the reason for this PR is that asdf is deprecating validation during AsdfFile.__init__. The code you proposed would result in a DeprecationWarning and test failure when the validation fails. The code proposed in this PR avoids the possibility of the DeprecationWarning by not providing the _instance during AsdfFile.__init__..

[WilliamJamieson]

Is asdf deprecating passing a dictionary directly to the AsdfFile initializer? If so, then that is a huge change of behavior!

If it's simply raising DepreciationWarning because of the validation change then a more thoughtful change should be considered as initializing with a dictionary is an extremely common usage of AsdfFile.

I would request that a validate=True option should be added to the initializer which preserves the current behavior without deprecation. You can set it so it defaults to None and raises the warning of impending behavior changes in that case.

[braingram]

asdf is not deprecating initializing with a dictionary. However if that dictionary is invalid one will see a DeprecationWarning noting that validation on AsdfFile.__init__ is deprecated and a ValidationError. If the dictionary is valid no warning or error will be seen. This was done to allow test suites (like the one in roman_datamodels) to detect the change.

[braingram]

@WilliamJamieson I'm not sure I addressed your comments here. asdf is deprecating (for removal) validation on AsdfFile.__init__ to reduce the duplication of functionality in asdf. Adding a validate keyword that triggers validation overlaps with functionality that is already provided by AsdfFile.validate.

[braingram]

@WilliamJamieson does this PR look good to you? I'd like to resolve this so I can merge the PR in asdf adding the deprecation warning without breaking the CI here.

[PaulHuwe]

LGTM modulo William's comments.