[WIP] Add test to validate examples in the main PALS repository

[EZoni]

Overview

This PR adds to the main CI workflow a new step that tests the examples in the main PALS repository.

The test goes as follows:

• Clone the examples/ directory from the main PALS repository.
• Loop over all *.pals.yaml files in the examples/ directory and read, parse, and validate the data from each file.

This makes sure that the Python implementation is actually consistent with the examples in the main PALS repository.

To do

• Fix the the Python implementation and/or the examples in the main PALS repository.

[EZoni]

Sparse checkout here to avoid cloning the entire repository.

[EZoni]

Loading the data into a BeamLine object triggers automatic validation.

[EZoni]

Also note that this, BeamLine(**data), is how we currently load data in our unit tests, e.g.,

pals-python/tests/test_serialization.py

Lines 25 to 26 in 6998a31

	# Parse the YAML data back into a BeamLine object
	loaded_line = pals.BeamLine(**yaml_data)

which is why I would expect it to work for the examples in the main PALS repository as well.

[EZoni]

The test shows indeed that the Python implementation and the examples in the main PALS repository are not consistent (which, in turn, shows that this test is indeed useful):

Traceback (most recent call last):
Parsing data from pals_temp/examples/fodo.pals.yaml...
  File "/home/runner/work/pals-python/pals-python/examples/test_external_examples.py", line 26, in <module>
    main()
  File "/home/runner/work/pals-python/pals-python/examples/test_external_examples.py", line 22, in main
    BeamLine(**data)
TypeError: pals.kinds.BeamLine.BeamLine() argument after ** must be a mapping, not list

I think the Python implementation does not like that the example in the main PALS repository is structured as a list (each item starting with -):

- drift1:
    kind: Drift
    length: 0.25

- quad1:
    kind: Quadrupole
    MagneticMultipoleP:
      Bn1: 1.0
    length: 1.0

- fodo_cell:
    kind: BeamLine
    line:
    - drift1
    - quad1
    - drift2:
        kind: Drift
        length: 0.5
    - quad2:
        inherit: quad1
        MagneticMultipoleP:
          Bn1: -1.0
    - drift1

- fodo_channel:
    kind: BeamLine
    line:
    - fodo_cell:
        repeat: 3

Possibly related to #32? Should we fix the Python implementation or the syntax in the main repository in this case?

P.S. Note that once this is fixed, the test will still fail because the Python implementation does not support inheritance yet.

[EZoni]

I think if we call BeamLine(data) instead of BeamLine(**data) (this is wrong in Python regardless...) and add a custom __init__ similar to the following one (vibe-coded) to the BeamLine class,

+    def __init__(self, data=None, **kwargs):
+        """Custom init to accept data as a positional argument"""
+        if data is not None and not kwargs:
+            # If data is passed as a positional argument without kwargs,
+            # use model_validate to process it through validators
+            instance = self.__class__.model_validate(data)
+            # Copy the validated attributes to this instance
+            for key, value in instance.__dict__.items():
+                object.__setattr__(self, key, value)
+            if hasattr(instance, '__pydantic_fields_set__'):
+                object.__setattr__(self, '__pydantic_fields_set__', instance.__pydantic_fields_set__)
+        else:
+            # Otherwise use the normal Pydantic initialization
+            super().__init__(**kwargs)
+

the test does pass on a simpler FODO example (without the features that haven't been implemented yet), where the data is structured as a list.

However, I'm not sure if this is the right way to go. I think we need to make a decision at the API level first.