Issue #57 - Introducing DataClasses and Pydantic (Static Type) Validation

[robons]

N.B. This PR relies on 3 separate bug fixes to the pydantic library which are currently stored in the forked repo at https://github.com/robons/pydantic/
N.B. - This does not allow us to start using specific validators on a per-attribute basis. This is something we might want to look into in future.

Dataclasses

This PR makes quite a change to our code-base. It enables us to easily validate that all of a class' attributes match the associated static type annotations. To support this, and to make a bit easier, it switches the majority of csvqb's models over to being dataclasses.

Dataclasses give us a simpler way of defining classes and makes our lives easier by implementing common functions you might want on your class including init, repr, str, eq, etc.

Dataclasses also provide an easy way for converting your model to a dictionary and specifying what the default values are for your parameters.

And all you have to do is annotate the model with the @dataclass attribute...

@dataclass
class NewQbDimension(QbDimension):
    label: str
    description: Optional[str] = field(default=None, repr=False)
    code_list: Optional[QbCodeList] = field(default=None, repr=False)
    parent_dimension_uri: Optional[str] = field(default=None, repr=False)
    source_uri: Optional[str] = field(default=None, repr=False)
    range_uri: Optional[str] = field(default=None, repr=False)
    uri_safe_identifier_override: Optional[str] = field(default=None, repr=False)

Note that I haven't had to write the __init__ function, or any other the other functions that the dataclasses will implement for me.

Examples

Example of validation process:

from csvqb.models.cube import *
cube = Cube(CatalogMetadata(None), None, [
    QbColumn("Some Column", NewQbDimension(None)),
    QbColumn(None, ExistingQbAttribute(None))
])
[error.message for error in cube.validate()]

[
    "Cube(metadata=CatalogMetadata(title=None)) - ('metadata', 'title') - none is not an allowed value",
    "Cube(metadata=CatalogMetadata(title=None)) - ('columns', 0, 'component', 'label') - none is not an allowed value",
    "Cube(metadata=CatalogMetadata(title=None)) - ('columns', 1, 'csv_column_title') - none is not an allowed value",
    "Cube(metadata=CatalogMetadata(title=None)) - ('columns', 1, 'component', 'attribute_uri') - none is not an allowed value"
]

Example where user has incorrectly defined columns:

from csvqb.models.cube import *

cube = Cube(CatalogMetadata("Some Title"), None, [
    NewQbUnit("Shouldn't be here")
])
[error.message for error in cube.validate()]

[
    "Cube(metadata=CatalogMetadata(title='Some Title')) - ('columns', 0) - instance of CsvColumn, tuple or dict expected",
    "'NewQbUnit' object has no attribute 'csv_column_title'"
]

It's also possible to directly call the pydantic_validation() method on any model which extends PydanticModel.

This satisfies Issue #57

[canwaf]

Thus concludes my review. Happy merging.