def unique_ids(validator, ui, instance, schema, id_name="id"):
if ui and validator.is_type(instance, "array"):
non_unique_ids = set()
all_ids = set()
for item in instance:
try:
item_id = item.get(id_name)
except AttributeError:
# if item is not a dict
item_id = None
if (item_id and not isinstance(item_id, list)
and not isinstance(item_id, dict)):
if item_id in all_ids:
non_unique_ids.add(item_id)
all_ids.add(item_id)
else:
if not uniq(instance):
msg = "Array has non-unique elements"
err = ValidationError(msg, instance=instance)
err.error_id = "uniqueItems_no_ids"
yield err
return
for non_unique_id in sorted(non_unique_ids):
msg = "Non-unique {} values".format(id_name)
err = ValidationError(msg, instance=non_unique_id)
err.error_id = "uniqueItems_with_{}".format(id_name)
yield err
def oneOf_draft4(validator, oneOf, instance, schema):
"""
oneOf_draft4 validator from
https://github.com/Julian/jsonschema/blob/d16713a4296663f3d62c50b9f9a2893cb380b7af/jsonschema/_validators.py#L337
Modified to:
- sort the instance JSON, so we get a reproducible output that we
can can test more easily
- Yield all the individual errors for linked or embedded releases within a
record.
- Return more information on the ValidationError object, to allow us to
replace the translation with a message in cove-ocds
"""
subschemas = enumerate(oneOf)
all_errors = []
for index, subschema in subschemas:
errs = list(validator.descend(instance, subschema, schema_path=index))
if not errs:
first_valid = subschema
break
# We check the title, because we don't have access to the field name,
# as it lives in the parent.
# It will not match the releases array in a release package, because
# there is no oneOf.
if (schema.get("title") == "Releases" or schema.get("description")
== "An array of linking identifiers or releases"):
# If instance is not a list, or is a list of zero length, then
# validating against either subschema will work.
# Assume instance is an array of Linked releases, if there are no
# "id"s in any of the releases.
if type(instance) is not list or all("id" not in release
for release in instance):
if "properties" in subschema.get(
"items",
{}) and "id" not in subschema["items"]["properties"]:
for err in errs:
err.assumption = "linked_releases"
yield err
return
# Assume instance is an array of Embedded releases, if there is an
# "id" in each of the releases
elif all("id" in release for release in instance):
if "id" in subschema.get("items", {}).get(
"properties", {}) or subschema.get("items", {}).get(
"$ref", "").endswith("release-schema.json"):
for err in errs:
err.assumption = "embedded_releases"
yield err
return
else:
err = ValidationError(
"This array should contain either entirely embedded releases or "
"linked releases. Embedded releases contain an 'id' whereas linked "
"releases do not. Your releases contain a mixture.")
err.error_id = "releases_both_embedded_and_linked"
yield err
return
all_errors.extend(errs)
else:
err = ValidationError(
f"{json.dumps(instance, sort_keys=True, default=decimal_default)} "
"is not valid under any of the given schemas",
context=all_errors,
)
err.error_id = "oneOf_any"
yield err
more_valid = [s for i, s in subschemas if validator.is_valid(instance, s)]
if more_valid:
more_valid.append(first_valid)
reprs = ", ".join(repr(schema) for schema in more_valid)
err = ValidationError(f"{instance!r} is valid under each of {reprs}")
err.error_id = "oneOf_each"
err.reprs = reprs
yield err
