Schema-based check assignment for compiled release and dataset checks

[jpmckinney]

Field-level checks are explicitly assigned to each field. This means that, if the OCDS schema changes, or if extensions are used, then definitions.py will need to be updated. Instead, it's possible to determine which checks to perform on which fields, by inspecting the OCDS release schema.

There is also hard-coding of paths to period objects in period_duration_in_days.py and period.py, and of paths to value objects in value_realistic.py (and a bunch more).

• coherent.milestone_status

[jpmckinney]

For rules like "For any field (or array item) with "$ref": "#/definitions/Identifier"":

In other OCDS code, we run jsonref.JsonRef.replace_refs(schema), loop through the schema and, if a path matches a condition, we yield it as a tuple. Then, we loop through the data. If a path in the data matches a path from the schema (we need to remove array indices from the data’s path in order to match the schema’s path, e.g. (‘contracts’, 0, ‘id’) becomes (‘contracts’, ‘id’)), then we perform an operation.

[jpmckinney]

For field-level definitions, I compared my calculated definitions (x) to the existing definitions with:

print(set(x) - set(definitions))
print(set(definitions) - set(x))

for k in set(x) & set(definitions):
    if x[k] != definitions[k] and (
        # Empty
        not x[k]
        # Non-partial
        or not any(hasattr(v[0], "func") for v in x[k])
        # Partial: functools.partial(foo, "bar") != functools.partial(foo, "bar")
        or any(
            x[k][i][1] != definitions[k][i][1]
            or x[k][i][0].func != definitions[k][i][0].func
            or x[k][i][0].args != definitions[k][i][0].args
            or x[k][i][0].keywords != definitions[k][i][0].keywords
            for i in range(len(x[k]))
        )
    ):
        print(k)
        print(x[k])
        print(definitions[k])