Enforce header entity length in grammar

[Ghesselink]

No description provided.

[Ghesselink]

Doesn't work so simple, have to take another look

self = <__init__.file object at 0x7f45d99f5bd0>

    @property
    def header(self):
        HEADER_FIELDS = {
            "file_description": namedtuple('file_description', ['description', 'implementation_level']),
            "file_name": namedtuple('file_name', ['name', 'time_stamp', 'author', 'organization', 'preprocessor_version', 'originating_system', 'authorization']),
            "file_schema":  namedtuple('file_schema', ['schema_identifiers']),
        }
        header = {}
    
        for field_name, namedtuple_class in HEADER_FIELDS.items():
            field_data = self.header_.get(field_name.upper(), [])
>           header[field_name.lower()] = namedtuple_class(*field_data)
E           TypeError: file_name.__new__() takes 8 positional arguments but 20 were given

for

HEADER;
FILE_DESCRIPTION(('ViewDefinition [Alignment-basedView]'),'2;1');
FILE_NAME('Header example2.ifc', '2022-09-16T10:35:07', ('Evandro Alfieri'), ('buildingSMART Int.'), 'IFC Motor 1.0', 'Company - Application - 26.0.0.0', 'none');
FILE_SCHEMA(('IFC4X3_ADD2'));
ENDSEC;

[aothms]

Some things are easier when you do them not in the grammar, but on the parsed ast tree. e.g assert len(f.header.file_description) == len(HEADER_FIELDS['file_description']) (in pseudo code)

[aothms]

Do we want to really raise an exception and terminate or wait for potentially additional errors?

[aothms]

Really nice, just would prefer without the getattr() if possible, but understand where it's coming from: exceptions are for exceptional control flow, but we are accumulating them.

[aothms]

Can we actually catch both cases separately CollectedValidationErrors, ValidationError instead, or refactor ValidationError to always point to a list of errors.

[Ghesselink]

I designed it in a way that ValidationErrors is never raised directly; the the issues are represented in subclasses (e.g. SyntaxError, DuplicateNameError, etc) and collected into a single CollectedValidationErrors instance, which is then raised. As a result, the getattr was already redundant and we can rely on exc.errors directly
json.dump([e.asdict() for e in exc.errors], sys.stdout, indent=2)
ValidationError is now only used for testing as it covers all sub classes with pytest.raises(_ValidationError).

I've also renamed it to _ValidationErrors and added documentation, clarifying that we only use it internally. The parser itself also doesn't import it directly.

[aothms]

Sounds great!