Source code for weaver.wps_restapi.colander_extras

"""
This module offers multiple utility schema definitions to be employed with :mod:`colander` and :mod:`cornice_swagger`.

The :class:`colander.SchemaNode` provided here can be used in-place of :mod:`colander`
ones, but giving you extended behaviour according to provided keywords. You can therefore
do the following and all will be applied without modifying your code base.

.. code-block:: python

    # same applies for Mapping and Sequence schemas
    from colander_extras import ExtendedSchemaNode as SchemaNode
    from colander import SchemaNode     # instead of this

The schemas support extended :mod:`cornice_swagger` type converters so that you
can generate OpenAPI-3 specifications. The original package support is limited
to Swagger-2. You will also need additional in-place modifications provided
`here <https://github.com/fmigneault/cornice.ext.swagger/tree/openapi-3>`_.

Since the intended usage is to generate JSON-deserialized data structures, the
base :class:`colander.SchemaNode` have some additional patches needed to handle
JSON-native type conversion that are not directly offered by :mod:`colander`

.. seealso::
    - https://github.com/Pylons/colander/issues/80

The main classes extensions are:
 - :class:`ExtendedSchemaNode`
 - :class:`ExtendedSequenceSchema`
 - :class:`ExtendedMappingSchema`

Multiple ``<ExtensionType>SchemaNode`` variants are provided. You can use them
as building blocks, gaining each of their respective feature, to make specific
schema that meets your desired behaviour. The ``Extended``-prefixed classes
combined all available ``<ExtensionType>``. Note that it is preferable to use
the full ``Extended``-prefixed classes over individual ones as they add
complementary support of one-another features.

.. warning::
    All node extensions assume that they are used for JSON data. Deserialization
    of unknown types *could* result in invalid result. Most of the time, Python
    native iterators such as tuples and generators *could* work and be converted
    to the corresponding sequence array (ie: list), but this is not guaranteed.

.. warning::
    When defining schema nodes, **DO NOT** use the ``name`` keyword otherwise
    most mapping will fail validation as they cannot retrieve the same key-name
    from the passed dictionary for ``deserialize`` validation. Let the API
    figure out the name of the field automatically. Instead, use the keyword
    or field ``title`` for adjusting the displayed name in the Swagger UI.
    The same value will also be used to generate the ``$ref`` reference names
    of generated OpenAPI model definitions. If not explicitly provided, the
    value of ``title`` **WILL** default to the name of the schema node class.
"""
import inspect
import re
import uuid
from abc import abstractmethod
from typing import TYPE_CHECKING

import colander
from cornice_swagger.converters.exceptions import ConversionError, NoSuchConverter
from cornice_swagger.converters.schema import (
    STRING_FORMATTERS,
    NumberTypeConverter,
    ObjectTypeConverter,
    TypeConversionDispatcher,
    TypeConverter,
    ValidatorConversionDispatcher,
    convert_range_validator,
    convert_regex_validator
)

if TYPE_CHECKING:
    from typing import Any, Dict, Iterable, Optional, Sequence, Type, Union

# pylint: disable=C0209,consider-using-f-string


[docs]LITERAL_SCHEMA_TYPES = frozenset([ colander.Boolean, colander.Number, # int, float, etc. colander.String, colander.Time, colander.Date, colander.DateTime,
# colander.Enum, # not supported but could be (literal int/str inferred from Python Enum object) ])
[docs]def _make_node_instance(schema_node_or_class): # type: (Union[colander.SchemaNode, Type[colander.SchemaNode]]) -> colander.SchemaNode """ Obtains a schema node instance in case it was specified only by type reference. This helps being more permissive of provided definitions while handling situations like presented in the example below: .. code-block:: python class Map(OneOfMappingSchema): # uses types instead of instances like 'SubMap1([...])' and 'SubMap2([...])' _one_of = (SubMap1, SubMap2) """ if isinstance(schema_node_or_class, colander._SchemaMeta): # noqa: W0212 schema_node_or_class = schema_node_or_class() if not isinstance(schema_node_or_class, colander.SchemaNode): # refer to original class to support non-extended raise ConversionTypeError( f"Invalid item should be a SchemaNode, got: {type(schema_node_or_class)!s}") return schema_node_or_class
[docs]def _get_schema_type(schema_node, check=False): # type: (Union[colander.SchemaNode, Type[colander.SchemaNode]], bool) -> Optional[colander.SchemaType] """ Obtains the schema-type from the provided node, supporting various initialization methods. - ``typ`` is set by an instantiated node from specific schema (e.g.: ``colander.SchemaNode(colander.String())``) - ``schema_type`` can also be provided, either by type or instance if using class definition with property :param schema_node: item to analyse :param check: only attempt to retrieve the schema type, and if failing return ``None`` :returns: found schema type :raises ConversionTypeError: if no ``check`` requested and schema type cannot be found (invalid schema node) """ schema_node = _make_node_instance(schema_node) schema_type = getattr(schema_node, "typ", getattr(schema_node, "schema_type")) if isinstance(schema_type, type): schema_type = schema_type() # only type instead of object, instantiate with default since no parameters anyway if not isinstance(schema_type, colander.SchemaType): if check: return None raise ConversionTypeError(f"Invalid schema type could not be detected: {type(schema_type)!s}") return schema_type
[docs]def _get_node_name(schema_node, schema_name=False): # type: (colander.SchemaNode, bool) -> str """ Obtains the name of the node with best available value. :param schema_node: node for which to retrieve the name. :param schema_name: - If ``True``, prefer the schema definition (class) name over the instance or field name. - Otherwise, return the field name, the title or as last result the class name. :returns: node name """ title = getattr(schema_node, "title", None) if title in ["", colander.required]: title = None if schema_name: return title or type(schema_node).__name__ return getattr(schema_node, "name", None) or title or type(schema_node).__name__
[docs]class SchemaNodeTypeError(TypeError): """ Generic error indicating that the definition of a SchemaNode is invalid. This usually means the user forgot to specify a required element for schema creation, or that a provided combination of keywords, sub-nodes and/or schema type don't make any sense together, that they are erroneous, or that they cannot be resolved because of some kind of ambiguous definitions leading to multiple conflicting choices. """
[docs]class ConversionTypeError(ConversionError, TypeError): """ Conversion error due to invalid type. """
[docs]class ConversionValueError(ConversionError, ValueError): """ Conversion error due to invalid value. """
[docs]class OneOfCaseInsensitive(colander.OneOf): """ Validator that ensures the given value matches one of the available choices, but allowing case insensitive values. """ def __init__(self, choices, *args, **kwargs): # type: (Iterable[str], Any, Any) -> None insensitive_choices = {} # set with kept order for choice in choices: insensitive_choices.setdefault(choice, None) if isinstance(choice, str): # add common combinations (not technically all possible ones) insensitive_choices.setdefault(choice.lower(), None) insensitive_choices.setdefault(choice.upper(), None) insensitive_choices = list(insensitive_choices) super(OneOfCaseInsensitive, self).__init__(insensitive_choices, *args, **kwargs) def __call__(self, node, value): # type: (colander.SchemaNode, Any) -> None if str(value).lower() not in (choice.lower() for choice in self.choices): return super(OneOfCaseInsensitive, self).__call__(node, value)
[docs]class BoundedRange(colander.Range): """ Validator of value within range with added ``exclusive`` bounds support. """ def __init__(self, min=None, max=None, exclusive_min=False, exclusive_max=False, **kwargs): # type: (Optional[Union[float, int]], Optional[Union[float, int]], bool, bool, Any) -> None self.min_excl = exclusive_min self.max_excl = exclusive_max super(BoundedRange, self).__init__(min=min, max=max, **kwargs) def __call__(self, node, value): super(BoundedRange, self).__call__(node, value) if self.min_excl and self.min is not None: if value <= self.min: min_err = colander._( self.min_err, mapping={"val": value, "min": self.min, "exclusive": True} ) raise colander.Invalid(node, min_err) if self.max_excl and self.max is not None: if value >= self.max: max_err = colander._( self.max_err, mapping={"val": value, "max": self.max, "exclusive": True} ) raise colander.Invalid(node, max_err)
[docs]class StringRange(BoundedRange): """ Validator that provides the same functionalities as :class:`colander.Range` for a numerical string value. """ def __init__(self, min=None, max=None, exclusive_min=False, exclusive_max=False, **kwargs): # type: (Optional[Union[float, int, str]], Optional[Union[float, int, str]], bool, bool, Any) -> None try: if isinstance(min, str): min = float(min) if "." in min or "e" in min else int(min) if isinstance(max, str): max = float(max) if "." in max or "e" in max else int(max) except (TypeError, ValueError): raise SchemaNodeTypeError("StringRange validator created with invalid min/max non-numeric string.") super(StringRange, self).__init__( min=min, max=max, exclusive_min=exclusive_min, exclusive_max=exclusive_max, **kwargs ) def __call__(self, node, value): # type: (colander.SchemaNode, str) -> Union[float, int] if not isinstance(value, str): raise colander.Invalid(node=node, value=value, msg="Value is not a string.") if not str.isnumeric(value): raise colander.Invalid(node=node, value=value, msg="Value is not a numeric string.") return super(StringRange, self).__call__(node, float(value) if "." in value or "e" in value else int(value))
[docs]class SchemeURL(colander.Regex): """ String representation of an URL with extended set of allowed URI schemes. .. seealso:: :class:`colander.url` [remote http(s)/ftp(s)] :class:`colander.file_uri` [local file://] """ def __init__(self, schemes=None, msg=None, flags=re.IGNORECASE): # type: (Optional[Iterable[str]], Optional[str], Optional[re.RegexFlag]) -> None if not schemes: schemes = [""] if not msg: msg = colander._(f"Must be a URL matching one of schemes {schemes}") # noqa regex_schemes = r"(?:" + "|".join(schemes) + r")" regex = colander.URL_REGEX.replace(r"(?:http|ftp)s?", regex_schemes) super(SchemeURL, self).__init__(regex, msg=msg, flags=flags)
[docs]class SemanticVersion(colander.Regex): """ String representation that is valid against Semantic Versioning specification. .. seealso:: https://semver.org/ """ def __init__(self, *args, v_prefix=False, rc_suffix=True, **kwargs): # type: (Any, bool, bool, Any) -> None if "regex" in kwargs: self.pattern = kwargs.pop("regex") else: v_prefix = "v" if v_prefix else "" rc_suffix = r"(\.[a-zA-Z0-9\-_]+)*" if rc_suffix else "" self.pattern = ( r"^" + v_prefix + r"\d+" # major r"(\.\d+" # minor r"(\.\d+" # patch + rc_suffix + r")*)*$" ) super(SemanticVersion, self).__init__(regex=self.pattern, *args, **kwargs)
[docs]class ExtendedBoolean(colander.Boolean): def __init__(self, *args, true_choices=None, false_choices=None, allow_string=False, **kwargs): # type: (Any, Optional[Iterable[str]], Optional[Iterable[str]], bool, Any) -> None """ Initializes the extended boolean schema node. When arguments :paramref:`true_choices` or :paramref:`false_choices` are provided, the corresponding string values are respectively considered as valid `truthy`/`falsy` values. Otherwise (default), ``strict`` values only of explicit type :class:`bool` will be considered valid. When values are specified :mod:`colander` converts them to string lowercase to compare against `truthy`/`falsy` values it should accept. For real `OpenAPI` typing validation, do **NOT** add other values like ``"1"`` to avoid conflict with :class:`ExtendedInteger` type for schemas that support both variants. If an `OpenAPI` field is expected to support `truthy`/`falsy` values, it is recommended to explicitly define its schema using a ``oneOf`` keyword of all relevant schemas it supports, an any applicable validators for explicit values. This is the safest way to ensure the generated `OpenAPI` schema corresponds to expected type validation. """ if not allow_string and true_choices is None and false_choices is None: # use strict variant self.true_choices = () self.false_choices = () self.true_val = "true" self.false_val = "false" self.false_reprs = [str(False)] self.true_reprs = [str(True)] else: # use normal variant (bool-like values) true_choices = true_choices if true_choices else ("true", ) false_choices = false_choices if false_choices else ("false", ) super(ExtendedBoolean, self).__init__( *args, true_choices=true_choices, false_choices=false_choices, **kwargs )
[docs] def deserialize(self, node, cstruct): # type: (colander.SchemaNode, Any) -> Union[Type[colander.null, bool]] if cstruct is colander.null: return cstruct # strict type variant if not self.true_choices and not self.false_choices: # note: cannot compare with literal 'True' and 'False' since '0' and '1' are equivalent (implicit convert) if isinstance(cstruct, bool): return cstruct raise colander.Invalid(node, colander._("\"${val}\" is neither True or False.", mapping={"val": cstruct})) # normal type variant return super(ExtendedBoolean, self).deserialize(node, cstruct)
[docs]class ExtendedNumber(colander.Number): """ Definition of a numeric value, either explicitly or implicit with permissive :class:`str` representation. Behaviour in each case: - ``strict=True`` and ``allow_string=False``: Value can only be explicit numeric type that matches exactly the base ``num`` type (default). All implicit conversion between :class:`float`, :class:`int` or :class:`str` are disallowed. - ``strict=True`` and ``allow_string=True``: Value can be the explicit numeric type (:class:`int` or :class:`float`) or a numeric :class:`str` value representing the corresponding base numeric type. Implicit conversion between :class:`float` and :class:`int` is still disallowed. - ``strict=False`` (``allow_string`` doesn't matter): Value can be anything as long as it can be converted to the expected numeric type (:class:`int` or :class:`float`). Recommended usage: - When making `OpenAPI` schema definitions for JSON body elements within a request or response object, default parameters ``strict=True`` and ``allow_string=False`` should be used to ensure the numeric type is respected. As for other literal data `Extended` schema types, keyword `oneOf` should be used when multiple similar value types are permitted for a field in order to document in `OpenAPI` the specific type definitions of expected data, which is automatically converted by ``json`` properties of request and response classes. - When defining `OpenAPI` query parameters, ``strict=True`` and ``allow_string=True`` should be used. This ensures that documented schemas still indicate only the numeric type as expected data format, although technically the ``path`` of the request will contain a :class:`str` representing the number. Queries are not automatically converted by request objects, but will be converted and validated as the explicit number following deserialization when using those configuration parameters. """ def __init__(self, *_, allow_string=False, strict=True, **__): # default applied based on 'strict' in kwargs # then, make even stricter number validation if requested super(ExtendedNumber, self).__init__(*_, **__) if strict: self.num = self.number if allow_string else self.strict @staticmethod @abstractmethod
[docs] def number(num): raise NotImplementedError
@staticmethod @abstractmethod
[docs] def strict(num): raise NotImplementedError
[docs]class ExtendedFloat(ExtendedNumber, colander.Float): """ Float definition with strict typing validation by default. This is to distinguish it from explicit definitions of ``float``-like numbers or strings. By default, values such as ``"1"``, ``1.0``, ``True`` will not be automatically converted to equivalent ``1.0``. """ def __init__(self, *_, allow_string=False, strict=True, **__): # type: (Any, bool, bool, Any) -> None colander.Float.__init__(self) ExtendedNumber.__init__(self, *_, strict=strict, allow_string=allow_string, **__) @staticmethod
[docs] def number(num): if (isinstance(num, str) and "." in num) or isinstance(num, float): return float(num) raise ValueError("Value is not a Floating point number (Integer not allowed).")
@staticmethod
[docs] def strict(num): if not isinstance(num, float): raise ValueError("Value is not a Floating point number (Boolean, Integer and String not allowed).") return num
[docs] def serialize(self, node, appstruct): result = super(ExtendedFloat, self).serialize(node, appstruct) if result is not colander.null: result = float(result) return result
[docs]class ExtendedInteger(ExtendedNumber, colander.Integer): """ Integer definition with strict typing validation by default. This is to distinguish it from explicit definitions of ``integer``-like numbers or strings. By default, values such as ``"1"``, ``1.0``, ``True`` will not be automatically converted to equivalent ``1``. """ def __init__(self, *_, allow_string=False, strict=True, **__): # type: (Any, bool, bool, Any) -> None colander.Integer.__init__(self) ExtendedNumber.__init__(self, *_, strict=strict, allow_string=allow_string, **__) @staticmethod
[docs] def number(num): if not float(num).is_integer() or isinstance(num, bool): raise ValueError("Value is not an Integer number (Float not allowed).") return int(num)
@staticmethod
[docs] def strict(num): # note: # - original colander function does not handle all cases # (e.g.: float("1.23").is_integer() -> False, but still not a float) # - furthermore, True/False are considered 'int', so must double check for 'bool' if not isinstance(num, int) or isinstance(num, bool): raise ValueError("Value is not a Integer number (Boolean, Float and String not allowed).") return num
[docs] def serialize(self, node, appstruct): result = super(ExtendedInteger, self).serialize(node, appstruct) if result is not colander.null: result = int(result) return result
[docs]class ExtendedString(colander.String): """ String with auto-conversion for known OpenAPI ``format`` field where no direct :mod:`colander` type exist. Converts :class:`uuid.UUID` to corresponding string when detected in the node if it defined ``format="uuid"``. For ``format="date"`` and ``format="date-time"``, consider instead using :class:`colander.Date` and :class:`colander.DateTime` respectively since more advanced support and features are provided with them. """
[docs] def deserialize(self, node, cstruct): # type: (colander.SchemaNode, Any) -> str try: if str(getattr(node, "format", "")).lower() == "uuid": if isinstance(cstruct, str): return str(uuid.UUID(cstruct)) if isinstance(cstruct, uuid.UUID): return str(cstruct) except ValueError: raise colander.Invalid(node, msg="Not a valid UUID string.", value=str(cstruct)) return super(ExtendedString, self).deserialize(node, cstruct)
[docs]class XMLObject(object): """ Object that provides mapping to known XML extensions for OpenAPI schema definition. Name of the schema definition in the OpenAPI will use :attr:`prefix` and the schema class name. Prefix can be omitted from the schema definition name by setting it to :class:`colander.drop`. The value of ``title`` provided as option or .. seealso:: - https://github.com/OAI/OpenAPI-Specification/blob/master/versions/3.0.3.md#user-content-xml-object - https://swagger.io/docs/specification/data-models/representing-xml/ """
[docs] attribute = None # define the corresponding node object as attribute instead of field
[docs] name = None # name of the attribute, or default to the class name of the object
[docs] namespace = None # location of "xmlns:<prefix> <location>" specification
[docs] prefix = None # prefix of the namespace
[docs] wrapped = None # used to wrap array elements called "<name>" within a block called "<name>s"
@property
[docs] def xml(self): spec = {} if isinstance(self.attribute, bool): spec["attribute"] = self.attribute if isinstance(self.name, str): spec["name"] = self.name if isinstance(self.namespace, str): spec["namespace"] = self.namespace if isinstance(self.prefix, str): spec["prefix"] = self.prefix if self.wrapped: # only add if True to avoid over-populate spec, default is False spec["wrapped"] = self.wrapped return spec or None
[docs]class ExtendedNodeInterface(object):
[docs] _extension = None # type: str
[docs] def _deserialize_impl(self, cstruct): raise NotImplementedError("ExtendedNodeInterface deserialize implementation missing")
[docs]class ExtendedSchemaMeta(colander._SchemaMeta): pass
[docs]class ExtendedSchemaBase(colander.SchemaNode, metaclass=ExtendedSchemaMeta): """ Utility base node definition that initializes additional parameters at creation time of any other extended schema. When no explicit ``title`` is specified by either keyword argument or field definition within container class, default it to the literal name of the class defining the schema node. This title can then be employed by other extended schema implementations to define *cleaner* schema references, notably in the case of :class:`KeywordMapper` derived classes that do not necessarily have any explicit target ``name`` field. When the schema node is a simple field within a container schema (mapping, sequence, etc.), operation is skipped to avoid applying the generic ``SchemaNode`` or ``ExtendedSchemaNode`` name of the basic node class. In this case, converters already employ the target ``name`` of the class attribute of the container schema under which that node gets created. When the schema node is a generic :class:`colander.String` without explicit ``validator``, but that one can be inferred from either ``pattern`` or ``format`` OpenAPI definition, the corresponding ``validator`` gets automatically generated. """ @staticmethod
[docs] def schema_type(): raise NotImplementedError("Using SchemaNode for a field requires 'schema_type' definition.")
def __init__(self, *args, **kwargs): schema_name = _get_node_name(self, schema_name=True) schema_type = _get_schema_type(self, check=True) if isinstance(self, XMLObject): if isinstance(self.title, str): title = self.title else: title = kwargs.get("title", schema_name) # pylint: disable=no-member # prefix added by XMLObject if self.prefix is not colander.drop: title = f"{self.prefix or 'xml'}:{title}" kwargs["title"] = title elif isinstance(schema_type, (colander.Mapping, colander.Sequence)): if self.title in ["", colander.required]: title = kwargs.get("title", schema_name) kwargs["title"] = title self.title = title if self.validator is None and isinstance(schema_type, colander.String): _format = kwargs.pop("format", getattr(self, "format", None)) pattern = kwargs.pop("pattern", getattr(self, "pattern", None)) if isinstance(pattern, str): self.validator = colander.Regex(pattern) elif isinstance(pattern, colander.Regex): self.validator = pattern elif _format in STRING_FORMATTERS: self.validator = STRING_FORMATTERS[_format]["validator"] default = kwargs.get("default", colander.null) if self.default is colander.null and default is not colander.null: self.default = default if self.default is not colander.null and self.missing is not colander.drop: self.missing = self.default # setting value makes 'self.required' return False, but doesn't drop it try: super(ExtendedSchemaBase, self).__init__(*args, **kwargs) ExtendedSchemaBase._validate(self) except Exception as exc: raise SchemaNodeTypeError(f"Invalid schema definition for [{schema_name}]") from exc @staticmethod
[docs] def _validate(node): if node.default and node.validator not in [colander.null, None]: try: node.validator(node, node.default) except (colander.Invalid, TypeError): if node.default is not colander.drop: raise SchemaNodeTypeError( "Default value [{!s}] of [{!s}] is not valid against its own validator.".format( node.default, _get_node_name(node, schema_name=True))
)
[docs]class DropableSchemaNode(ExtendedNodeInterface, ExtendedSchemaBase): """ Schema that can be dropped if the value is missing. Drops the underlying schema node if ``missing=drop`` was specified and that the value representing it represents an *empty* value instead of raising a invalid schema error. In the case of nodes corresponding to literal schema type (i.e.: Integer, String, etc.), the *empty* value looked for is ``None``. This is to make sure that ``0`` or ``""`` are preserved unless explicitly representing *no-data*. In the case of container schema types (i.e.: list, dict, etc.), it is simply considered *empty* if there are no element in it, without any more explicit verification. Original behaviour of schema classes that can have children nodes such as :class:`colander.MappingSchema` and :class:`colander.SequenceSchema` are to drop the sub-node only if its value is resolved as :class:`colander.null` or :class:`colander.drop`. This results in *optional* field definitions replaced by ``None`` in many implementations to raise :py:exc:`colander.Invalid` during deserialization. Inheriting this class in a schema definition will handle this situation automatically. Required schemas (without ``missing=drop``, defaulting to :class:`colander.required`) will still raise for undefined nodes. The following snippet shows the result that can be achieved using this schema class: .. code-block:: python class SchemaA(DropableSchemaNode, MappingSchema): field = SchemaNode(String()) class SchemaB(MappingSchema): s1 = SchemaA(missing=drop) # optional s2 = SchemaA() # required SchemaB().deserialize({"s1": None, "s2": {"field": "ok"}}) # results: {'s2': {'field': 'ok'}} .. seealso: - https://github.com/Pylons/colander/issues/276 - https://github.com/Pylons/colander/issues/299 .. seealso:: - :class:`DropableMappingSchema` - :class:`DropableSequenceSchema` """
[docs] _extension = "_ext_dropable"
def __init__(self, *args, **kwargs): super(DropableSchemaNode, self).__init__(*args, **kwargs) setattr(self, DropableSchemaNode._extension, True) @staticmethod
[docs] def schema_type(): raise NotImplementedError("Using SchemaNode for a field requires 'schema_type' definition.")
# pylint: disable=W0222,signature-differs
[docs] def deserialize(self, cstruct): return ExtendedSchemaNode.deserialize(self, cstruct) # noqa
[docs] def _deserialize_impl(self, cstruct): if not getattr(self, DropableSchemaNode._extension, False): return cstruct if self.default is colander.null and self.missing is colander.drop: if cstruct is colander.drop: return colander.drop containers = (colander.SequenceSchema.schema_type, colander.MappingSchema.schema_type, colander.TupleSchema.schema_type) if self.schema_type in containers and not cstruct: return colander.drop elif cstruct in (None, colander.null): return colander.drop return cstruct
[docs]class DefaultSchemaNode(ExtendedNodeInterface, ExtendedSchemaBase): """ Schema that will return the provided default value when the corresponding value is missing or invalid. If ``default`` keyword is provided during :class:`colander.SchemaNode` creation, overrides the returned value by this default if missing from the structure during :meth:`deserialize` call. Original behaviour was to drop the missing value instead of replacing by ``default``. Executes all other :class:`colander.SchemaNode` operations normally. .. seealso:: - :class:`DefaultMappingSchema` - :class:`DefaultSequenceSchema` """
[docs] _extension = "_ext_default"
def __init__(self, *args, **kwargs): super(DefaultSchemaNode, self).__init__(*args, **kwargs) setattr(self, DefaultSequenceSchema._extension, True) @staticmethod
[docs] def schema_type(): raise NotImplementedError("Using SchemaNode for a field requires 'schema_type' definition.")
# pylint: disable=W0222,signature-differs
[docs] def deserialize(self, cstruct): return ExtendedSchemaNode.deserialize(self, cstruct) # noqa
[docs] def _deserialize_impl(self, cstruct): if not getattr(self, DefaultSchemaNode._extension, False): return cstruct if cstruct is colander.null and self.required and self.default in (colander.null, colander.drop): raise colander.Invalid(node=self, msg="Missing value for required field without any default.") # if nothing to process in structure, ask to remove (unless picked by default) result = colander.drop if cstruct is not colander.null: result = cstruct if not isinstance(self.default, type(colander.null)) and result is colander.drop: result = self.default return result
[docs]class VariableSchemaNode(ExtendedNodeInterface, ExtendedSchemaBase): """ Object schema that allows defining a field key as *variable* by name supporting deserialization validation. This definition is useful for defining a dictionary where the key name can be any string value but contains an underlying schema that has a very specific structure to be validated, such as in the following example. .. code-block:: json { "<any-key-id>": { "name": "required", "value": "something" }, "<another-key-id>": { "name": "other required", "value": "same schema" } } This is accomplished using the following definition:: class RequiredDict(ExtendedMappingSchema): name = ExtendedSchemaNode(String()) value = ExtendedSchemaNode(String()) class ContainerAnyKey(ExtendedMappingSchema): var = RequiredDict(variable="{id}") In the above example, the ``var`` node name that would normally be automatically generated and used to define the dictionary key will be replaced by ``{id}`` (any string provided by ``variable`` keyword). The actual attribute name ``var`` could be replaced by anything. .. warning:: Since ``variable`` tells the deserialization converter to try matching any valid children node schema with the provided structure regardless of key name, you should ensure that the variable child node has at least one :class:`colander.required` field (either directly or much deeper in the structure) to ensure it has something to validate against. Otherwise, anything will be matched (ie: drop all for empty structure would be considered as valid). The above statement also applies in case you provide more than one variable schema node under a mapping where both underlying schema are different. Without any required child node to distinguish between them, the sub-structure under each variable node *could* end up interchanged. .. note:: It is recommended to use variable names that include invalid characters for class/attribute names (e.g.: ``{}`` or ``<>``) in order to ensure that any substitution when attempting to find the schema matching the variable doesn't result in overrides. As presented in the following example, ``const`` *could* get overridden if there was a structure to parse that contained both the ``const`` sub-structure and another one with an arbitrary key matched against ``<var>``. .. note:: In order to avoid invalid rendering of ``<>`` strings (interpreted as HTML tags by Swagger-UI), it is recommended to employ ``{}`` notation for representing variable names. .. code-block:: python class ContainerAnyKeyWithName(ExtendedMappingSchema): var = RequiredDict(variable="const") # 'const' could clash with other below const = RequiredDict(String()) Using ``<const>`` instead would ensure that no override occurs as it is a syntax error to write ``<const> = RequiredDict(String())`` in the class definition, but this value can still be used to create the internal mapping to evaluate sub-schemas without name clashes. As a plus, it also helps giving an indication that *any key* is accepted. .. seealso:: - :class:`ExtendedSchemaNode` - :class:`ExtendedMappingSchema` """
[docs] _extension = "_ext_variable"
[docs] _variable = "variable" # name of property containing variable name
[docs] _variable_map = "variable_map" # name of property containing variable => real node/key matched
@classmethod
[docs] def is_variable(cls, node): # type: (colander.SchemaNode) -> bool """ If current node is the variable field definition. """ return getattr(node, cls._variable, None) is not None
[docs] def has_variables(self): """ If the current container schema node has sub-node variables. """ if isinstance(_get_schema_type(self), colander.Mapping): return any(VariableSchemaNode.is_variable(node) for node in self.children) return False
@staticmethod
[docs] def schema_type(): raise NotImplementedError("Using SchemaNode for a field requires 'schema_type' definition.")
def __init__(self, *args, **kwargs): # define node with provided variable by keyword or within a SchemaNode class definition var = kwargs.pop(self._variable, getattr(self, self._variable, None)) super(VariableSchemaNode, self).__init__(*args, **kwargs) if var: # note: literal type allowed only for shorthand notation, normally not allowed schema_type = _get_schema_type(self) if not isinstance(schema_type, tuple(list(LITERAL_SCHEMA_TYPES) + [colander.Mapping])): # noqa raise SchemaNodeTypeError( "Keyword 'variable' can only be applied to Mapping and literal-type schema nodes. " "Got: {!s} ({!s})".format(type(self), schema_type)) self.name = kwargs.get("name", var) if not self.title: self.title = var self.raw_title = var setattr(self, self._variable, var) self._mark_variable_children() setattr(self, VariableSchemaNode._extension, True)
[docs] def _mark_variable_children(self): """ Ensures that any immediate children schema with variable key are detected. Verifies if a :class:`colander.MappingSchema` (or any of its extensions) contains children :class:`VariableSchemaNode` schema nodes for adequate :meth:`deserialize` result later on. If children nodes are detected as *variable*, this schema is marked for special processing of the children nodes so they don't get removed from the result (in case of optional field specified by ``missing=drop``) nor raise :class:`colander.Invalid` because they are supposed to be :class:`colander.required`. .. note:: Because mapping schema deserialization is normally processed by validating the content of a sub-node according to its ``name``, it is not possible to use the normal approach (i.e.: get dictionary value under matching key-name and validate it against the sub-schema of same name). We must verify against every *variable* node available in the mapping (also ignoring constants nodes with explicitly named keys), then guess a valid match and finally return it with modified name corresponding to the expected ``variable`` value in the parent mapping schema. Returning this modified ``name`` with ``variable`` makes the value/sub-schema correspondence transparent to the parent mapping when dictionary get-by-key is called during the mapping validation. .. warning:: Because of the above reversed-processing method, all *mapping* nodes must derive from :class:`VariableSchemaNode` to ensure they pre-process potential *variable* candidates. """ if self.schema_type is colander.Mapping: var_children = self._get_sub_variable(self.children) var_full_search = [var_children] # TODO: see if this will still be necessary... # When a Keyword schema contains a mapping with itself containing a variable, the mapping should directly # handle this child-variable detection mechanism. There is no need to go search for vars "2 level" lower. # if isinstance(_make_node_instance(self), KeywordMapper): # keyword_objects = KeywordMapper.get_keyword_items(self) # noqa # var_full_search.extend( # [self._get_sub_variable(_make_node_instance(var_obj).children) # for var_obj in keyword_objects if var_obj.schema_type is colander.Mapping] # ) for var_subnodes in var_full_search: if len(var_subnodes): var_names = [child.name for child in var_subnodes] for var in var_names: if len([v for v in var_names if v == var]) == 1: continue raise SchemaNodeTypeError("Invalid node '{}' defines multiple schema nodes " "with name 'variable={}'.".format(type(self), var)) var_names = [getattr(child, self._variable, None) for child in var_subnodes] setattr(self, self._variable_map, {var: [] for var in var_names})
[docs] def _get_sub_variable(self, subnodes): return [child for child in subnodes if getattr(child, self._variable, None)]
# pylint: disable=W0222,signature-differs
[docs] def deserialize(self, cstruct): return ExtendedSchemaNode.deserialize(self, cstruct) # noqa
@staticmethod
[docs] def _check_deserialize(node, cstruct): if not getattr(node, VariableSchemaNode._extension, False): return False if cstruct in (colander.drop, colander.null): return False # skip step in case operation was called as sub-node from another schema but doesn't # correspond to a valid variable map container (e.g.: SequenceSchema) if not isinstance(node, VariableSchemaNode): return False var_map = getattr(node, node._variable_map, {}) if not isinstance(var_map, dict) or not len(var_map): return False # value must be a dictionary map object to allow variable key if not isinstance(cstruct, dict): msg = f"Variable key not allowed for non-mapping data: {type(cstruct).__name__}" raise colander.Invalid(node=node, msg=msg) return True
@staticmethod
[docs] def _deserialize_remap(node, cstruct, var_map, var_name, has_const_child): invalid_var = colander.Invalid(node, value=var_map) try: # Substitute real keys with matched variables to run full deserialize so # that mapping can find nodes name against attribute names, then re-apply originals. # We must do this as non-variable sub-schemas could be present and we must also # validate them against full schema. if not has_const_child: result = node.default or {} else: for mapped in var_map.values(): # if multiple objects corresponding to a variable sub-schema where provided, # we only give one as this is what is expected for normal-mapping deserialize cstruct[mapped[0]["node"]] = cstruct.pop(mapped[0]["name"]) result = super(VariableSchemaNode, node).deserialize(cstruct) # noqa for mapped in var_map.values(): # invalid if no variable match was found, unless optional if mapped is None and node.missing is colander.required: raise KeyError for var_mapped in mapped: result[var_mapped["name"]] = var_mapped["cstruct"] except colander.Invalid as invalid: invalid_var.msg = f"Tried matching variable '{var_name}' sub-schemas but no match found." invalid_var.add(invalid) raise invalid_var except KeyError: invalid_var.msg = f"Tried matching variable '{var_name}' sub-schemas but mapping failed." raise invalid_var if not isinstance(result, dict): raise TypeError("Variable result must be of mapping schema type. Got [{}] value {}".format( _get_node_name(node, schema_name=True), result )) return result
[docs] def _deserialize_impl(self, cstruct): if not VariableSchemaNode._check_deserialize(self, cstruct): return cstruct var_map = getattr(self, self._variable_map, {}) var_children = self._get_sub_variable(self.children) const_child_keys = [child.name for child in self.children if child not in var_children] var = None for var_child in var_children: var = getattr(var_child, self._variable, None) var_map[var] = [] var_msg = f"Requirement not met under variable: {var}." var_invalid = colander.Invalid(node=self, msg=var_msg, value=cstruct) # attempt to find any sub-node matching the sub-schema under variable for child_key, child_cstruct in cstruct.items(): # skip explicit nodes as well as other variables already matched # cannot match the same child-cstruct again with another variable if child_key in const_child_keys or child_key in var_map: continue schema_class = _make_node_instance(var_child) try: var_cstruct = schema_class.deserialize(child_cstruct) if var_cstruct in [colander.drop, colander.null]: if var_child.missing is colander.drop: continue if var_child.default not in [colander.drop, colander.null]: var_cstruct = var_child.default # not reached if previous raised invalid var_map[var].append({ "node": schema_class.name, "name": child_key, "cstruct": var_cstruct }) except colander.Invalid as invalid: if var_child.missing is colander.drop: continue if var_child.missing is not colander.required: var_map[var].append({ "node": schema_class.name, "name": child_key, "cstruct": var_child.default }) else: var_invalid.add(invalid) var_val = var_map.get(var, colander.null) if var_val is colander.null: # allow unmatched variable item under mapping if it is not required if var_child.missing is colander.drop: continue raise var_invalid # invalid if no variable match was found, unless optional for mapped in var_map.values(): if len(mapped) < 1 and var_child.missing is colander.required: raise var_invalid return VariableSchemaNode._deserialize_remap(self, cstruct, var_map, var, bool(const_child_keys))
[docs]class SortableMappingSchema(ExtendedNodeInterface, ExtendedSchemaBase): """ Adds sorting capabilities to mapping schema. Extended schema nodes that inherit from :class:`colander.Mapping` schema-type such that they can request ordering of resulting fields by overriding properties :attr:`_sort_first` and :attr:`_sort_after` within the schema definition with lists of fields names to sort. .. seealso:: - :func:`_order_deserialize` """
[docs] _extension = "_ext_sortable"
[docs] _sort_first = [] # type: Sequence[str]
[docs] _sort_after = [] # type: Sequence[str]
def __init__(self, *args, **kwargs): super(SortableMappingSchema, self).__init__(*args, **kwargs) setattr(self, SortableMappingSchema._extension, True) # pylint: disable=W0222,signature-differs
[docs] def deserialize(self, cstruct): return ExtendedSchemaNode.deserialize(self, cstruct) # noqa
@staticmethod
[docs] def schema_type(): raise NotImplementedError("Using SchemaNode for a field requires 'schema_type' definition.")
[docs] def _deserialize_impl(self, cstruct): if not getattr(self, SortableMappingSchema._extension, False): return cstruct if not isinstance(cstruct, dict): return cstruct sort_first = getattr(self, "_sort_first", []) sort_after = getattr(self, "_sort_after", []) if sort_first or sort_after: return self._order_deserialize(cstruct, sort_first, sort_after) return cstruct
@staticmethod
[docs] def _order_deserialize(cstruct, sort_first=None, sort_after=None): # type: (Dict[str, Any], Optional[Sequence[str]], Optional[Sequence[str]]) -> Dict[str, Any] """ Enforces ordering of expected fields in deserialized result, regardless of specified children/inherited schema. This function takes care of moving back items in a consistent order for better readability from API responses against different loaded definitions field order from remote servers, local database, pre-defined objects, etc. This way, any field insertion order from both the input ``cstruct`` following deserialization operation, the internal mechanics that :mod:`colander` (and extended OpenAPI schema definitions) employ to process this deserialization, and the ``result`` dictionary fields order obtained from it all don't matter. Using this, the order of inheritance of schema children classes also doesn't matter, removing the need to worry about placing classes in any specific order when editing and joining the already complicated structures of inherited schemas. :param cstruct: JSON structure to be sorted that has already been processed by a schema's ``deserialize`` call. :param sort_first: ordered list of fields to place first in the result. :param sort_after: ordered list of fields to place last in the result. :returns: results formed from cstruct following order: (<fields_firsts> + <other_fields> + <fields_after>) """ sort_first = sort_first if sort_first else [] sort_after = sort_after if sort_after else [] result = {field: cstruct.pop(field, None) for field in sort_first if field in cstruct} remain = {field: cstruct.pop(field, None) for field in sort_after if field in cstruct} result.update(cstruct) result.update(remain) return result
[docs]class ExtendedSchemaNode(DefaultSchemaNode, DropableSchemaNode, VariableSchemaNode, ExtendedSchemaBase): """ Base schema node with support of extended functionalities. Combines all :class:`colander.SchemaNode` extensions so that ``default`` keyword is used first to resolve a missing field value during :meth:`deserialize` call, and then removes the node completely if no ``default`` was provided, and evaluate variables as needed. .. seealso:: - :class:`ExtendedMappingSchema` - :class:`ExtendedSequenceSchema` - :class:`DefaultSchemaNode` - :class:`DropableSchemaNode` - :class:`VariableSchemaNode` """
[docs] _extension = "_ext_combined"
@staticmethod
[docs] def schema_type(): raise NotImplementedError("Using SchemaNode for a field requires 'schema_type' definition.")
[docs] def _deserialize_extensions(self, cstruct): result = cstruct # process extensions to infer alternative parameter/property values # node extensions order is important as they can impact the following ones for node in [DropableSchemaNode, DefaultSchemaNode, VariableSchemaNode]: # type: Type[ExtendedNodeInterface] # important not to break if result is 'colander.null' since Dropable and Default # schema node implementations can substitute it with their appropriate value if result is colander.drop: # if result is to drop though, we are sure that nothing else must be done break result = node._deserialize_impl(self, result) return result
[docs] def deserialize(self, cstruct): schema_type = _get_schema_type(self) result = ExtendedSchemaNode._deserialize_extensions(self, cstruct) try: # process usual base operation with extended result if result not in (colander.drop, colander.null): # when processing mapping/sequence, if the result is an empty container, return the default instead # this is to avoid returning many empty containers in case upper level keywords (oneOf, anyOf, etc.) # need to discriminate between them # empty container means that none of the sub-schemas/fields where matched against input structure if isinstance(schema_type, colander.Mapping): # skip already preprocessed variable mapping from above VariableSchemaNode deserialize # otherwise, following 'normal' schema deserialize could convert valid structure into null if self.has_variables(): return result result = colander.MappingSchema.deserialize(self, result) elif isinstance(schema_type, colander.Sequence): result = colander.SequenceSchema.deserialize(self, result) else: # special cases for JSON conversion and string dump, serialize parsable string timestamps # deserialize causes Date/DateTime/Time to become Python datetime, and result raises if not string # employ serialize instead which provides the desired conversion from datetime objects to string if isinstance(schema_type, (colander.Date, colander.DateTime, colander.Time)): if not isinstance(result, str): result = colander.SchemaNode.serialize(self, result) else: result = colander.SchemaNode.deserialize(self, result) result = self.default if result is colander.null else result except colander.Invalid: # if children schema raised invalid but parent specifically requested # to be dropped by default and is not required, silently discard the whole structure if self.default is colander.drop: return colander.drop raise if result is colander.null and self.missing is colander.required: raise colander.Invalid(node=self, msg=self.missing_msg) return SortableMappingSchema._deserialize_impl(self, result)
[docs]class DropableSequenceSchema(DropableSchemaNode, colander.SequenceSchema): """ Sequence schema that supports the dropable functionality. Extends :class:`colander.SequenceSchema` to auto-handle dropping missing entry definitions when its value is either ``None``, :class:`colander.null` or :class:`colander.drop`. """
[docs] schema_type = colander.SequenceSchema.schema_type
[docs]class DefaultSequenceSchema(DefaultSchemaNode, colander.SequenceSchema): """ Sequence schema that supports the default value functionality. Extends :class:`colander.SequenceSchema` to auto-handle replacing the result using the provided ``default`` value when the deserialization results into a sequence that should normally be dropped. """
[docs] schema_type = colander.SequenceSchema.schema_type
[docs]class ExtendedSequenceSchema(DefaultSchemaNode, DropableSchemaNode, colander.SequenceSchema): """ Sequence schema that supports all applicable extended schema node functionalities. Combines :class:`DefaultSequenceSchema` and :class:`DefaultSequenceSchema` extensions so that ``default`` keyword is used first to resolve a missing sequence during :meth:`deserialize` call, and then removes the node completely if no ``default`` was provided. .. seealso:: - :class:`ExtendedSchemaNode` - :class:`ExtendedMappingSchema` """
[docs] schema_type = colander.SequenceSchema.schema_type
def __init__(self, *args, **kwargs): super(ExtendedSequenceSchema, self).__init__(*args, **kwargs) self._validate()
[docs] def _validate(self): # pylint: disable=arguments-differ ExtendedSchemaBase._validate(self.children[0])
[docs]class DropableMappingSchema(DropableSchemaNode, SortableMappingSchema, colander.MappingSchema): """ Mapping schema that supports the dropable functionality. Override the default :class:`colander.MappingSchema` to auto-handle dropping missing field definitions when the corresponding value is either ``None``, :class:`colander.null` or :class:`colander.drop`. """
[docs] schema_type = colander.MappingSchema.schema_type
[docs]class DefaultMappingSchema(DefaultSchemaNode, SortableMappingSchema, colander.MappingSchema): """ Mapping schema that supports the default value functionality. Override the default :class:`colander.MappingSchema` to auto-handle replacing missing entries by their specified ``default`` during deserialization. """
[docs] schema_type = colander.MappingSchema.schema_type
[docs]class VariableMappingSchema(VariableSchemaNode, colander.MappingSchema): """ Mapping schema that supports the variable functionality. Override the default :class:`colander.MappingSchema` to auto-handle replacing missing entries by their specified ``variable`` during deserialization. """
[docs] schema_type = colander.MappingSchema.schema_type
[docs]class ExtendedMappingSchema( DefaultSchemaNode, DropableSchemaNode, VariableSchemaNode, SortableMappingSchema, colander.MappingSchema ): """ Combines multiple extensions of :class:`colander.MappingSchema` handle their corresponding keywords. Resolution is done so that ``default`` keyword is used first to resolve a missing object during :meth:`deserialize` call, and then removes the node completely if no ``default`` was provided. .. seealso:: - :class:`DefaultSchemaNode` - :class:`DropableSchemaNode` - :class:`VariableSchemaNode` - :class:`ExtendedSchemaNode` - :class:`ExtendedSequenceSchema` - :class:`SortableMappingSchema` - :class:`PermissiveMappingSchema` """
[docs] schema_type = colander.MappingSchema.schema_type
def __init__(self, *args, **kwargs): super(ExtendedMappingSchema, self).__init__(*args, **kwargs) self._validate_nodes()
[docs] def _validate_nodes(self): for node in self.children: ExtendedSchemaBase._validate(node)
[docs]class StrictMappingSchema(ExtendedMappingSchema): """ Object schema that will ``raise`` any unknown field not represented by children schema. This is equivalent to `OpenAPI` object mapping with ``additionalProperties: false``. This type is useful for defining a dictionary that matches *exactly* a specific set of values and children schema. ..note:: When doing schema deserialization to validate it, unknown keys would normally be removed without this class (default behaviour is to ``ignore`` them). With this schema, content under an unknown key is fails validation. .. seealso:: :class:`PermissiveMappingSchema` """ def __init__(self, *args, **kwargs): kwargs["unknown"] = "raise" super(StrictMappingSchema, self).__init__(*args, **kwargs) # sub-type mapping itself must also have 'raise' such that its own 'deserialize' copies the fields over self.typ.unknown = "raise"
[docs]class EmptyMappingSchema(StrictMappingSchema): """ Mapping that guarantees it is completely empty for validation during deserialization. Any children added to this schema are removed automatically. """ def __init__(self, *args, **kwargs): # type: (Any, Any) -> None super(EmptyMappingSchema, self).__init__(*args, **kwargs) self.children = []
[docs]class PermissiveMappingSchema(ExtendedMappingSchema): """ Object schema that will ``preserve`` any unknown field to remain present in the resulting deserialization. This type is useful for defining a dictionary where some field names are not known in advance, or when more optional keys that don't need to all be exhaustively provided in the schema are acceptable. ..note:: When doing schema deserialization to validate it, unknown keys would normally be removed without this class (default behaviour is to ``ignore`` them). With this schema, content under an unknown key using ``preserve`` are passed down without any validation. Other fields that are explicitly specified with sub-schema nodes will still be validated as per usual behaviour. .. seealso:: :class:`StrictMappingSchema` Example:: class AnyKeyObject(PermissiveMappingSchema): known_key = SchemaNode(String()) AnyKeyObject().deserialize({"unknown": "kept", "known_key": "requirement"})) # result: dictionary returned as is instead of removing 'unknown' entry # 'known_key' is still validated with its string schema .. note:: This class is only a shorthand definition of ``unknown`` keyword for convenience. All :class:`colander.MappingSchema` support this natively. """ def __init__(self, *args, **kwargs): # type: (Any, Any) -> None kwargs["unknown"] = "preserve" super(PermissiveMappingSchema, self).__init__(*args, **kwargs) # sub-type mapping itself must also have 'preserve' such that its own 'deserialize' copies the fields over self.typ.unknown = "preserve"
[docs]class PermissiveSequenceSchema(ExtendedMappingSchema):
[docs] item = PermissiveMappingSchema()
[docs]class KeywordMapper(ExtendedMappingSchema): """ Generic keyword mapper for any sub-implementers. Allows specifying multiple combinations of schemas variants for an underlying schema definition. Each implementer must provide the corresponding ``keyword`` it defines amongst `OpenAPI` specification keywords. """
[docs] schema_type = colander.MappingSchema.schema_type
[docs] _keyword_schemas_only_object = False # override validation as needed
[docs] _keyword_schemas_same_struct = False # override validation as needed
[docs] _keywords = frozenset(["_one_of", "_all_of", "_any_of", "_not"])
[docs] _keyword_map = {_kw: _kw.replace("_of", "Of").replace("_", "") for _kw in _keywords} # kw->name
[docs] _keyword_inv = {_kn: _kw for _kw, _kn in _keyword_map.items()} # name->kw
[docs] _keyword = None # type: str
def __init__(self, *args, **kwargs): super(KeywordMapper, self).__init__(*args, **kwargs) if not hasattr(self, self._keyword): # try retrieving from a kwarg definition (either as literal keyword or OpenAPI name) if kwargs: maybe_kwargs = [_kw for _kw in kwargs if _kw in self._keyword_map or _kw in self._keyword_inv] if len(maybe_kwargs) == 1: self._keyword = self._keyword_inv.get(maybe_kwargs[0], maybe_kwargs[0]) setattr(self, self._keyword, kwargs.get(maybe_kwargs[0])) if not self._keyword: raise SchemaNodeTypeError(f"Type '{self}' must define a keyword element.") self._validate_keyword_unique() self._validate_keyword_schemas() @classmethod
[docs] def get_keyword_name(cls): return cls._keyword_map[cls._keyword]
[docs] def get_keyword_items(self): return getattr(self, self._keyword, [])
[docs] def _validate_keyword_unique(self): kw_items = self.get_keyword_items() if not hasattr(kw_items, "__iter__") or not len(kw_items): # noqa raise ConversionValueError( "Element '{}' of '{!s}' must be iterable with at least 1 value. " "Instead it was '{!s}'".format(self._keyword, _get_node_name(self, schema_name=True), kw_items) ) total = 0 for kw in self._keywords: if hasattr(self, kw): total += 1 if total > 1: raise SchemaNodeTypeError(f"Multiple keywords '{list(self._keywords)}' not permitted for '{self!s}'") if not total == 1: raise SchemaNodeTypeError(f"Missing one of keywords '{list(self._keywords)}' for '{self!s}'")
[docs] def _validate_keyword_schemas(self): """ Validation of children schemas under keyword. Validation of keyword sub-nodes to be only defined as schema *objects* if property ``_keyword_schemas_only_object = True`` (i.e.: any node that defines its schema type as :class:`Mapping`). Validation of keyword sub-nodes to all have matching structure of container if ``_keyword_schemas_same_struct = True`` (i.e.: all :class:`colander.Mapping`, all literal schema-types, etc.). """ children = self.get_keyword_items() keyword = self.get_keyword_name() schema_name = _get_node_name(self, schema_name=True) if getattr(self, "_keyword_schemas_only_object", False): for child in children: if child.schema_type is not colander.Mapping: raise SchemaNodeTypeError( "Keyword schema '{}' of type '{}' can only have object children, " "but '{}' is '{}'.".format(schema_name, keyword, type(child), child.schema_type) ) if getattr(self, "_keyword_schemas_same_struct", False): node_types = {child.name: _get_schema_type(child) for child in children} if not ( all(isinstance(typ, colander.Mapping) for _, typ in node_types.items()) or all(isinstance(typ, colander.Sequence) for _, typ in node_types.items()) or all(isinstance(typ, tuple(LITERAL_SCHEMA_TYPES)) for _, typ in node_types.items()) ): raise SchemaNodeTypeError( "Keyword schema '{}' of type '{}' can only have children of same schemas-type structure, " "but different ones were found '{}'.".format(schema_name, keyword, node_types) ) ExtendedMappingSchema._validate_nodes(self) for node in children: ExtendedSchemaBase._validate(node)
@abstractmethod
[docs] def _deserialize_keyword(self, cstruct): """ Deserialization and validation of a keyword-based schema definition. This method must be implemented by the specific keyword to handle invalid subnodes according to the behaviour it offers. .. seealso:: - :meth:`_deserialize_subnode` """ raise NotImplementedError
[docs] def _deserialize_subnode(self, node, cstruct, index): """ Deserialization and validation of sub-nodes under a keyword-based schema definition. This method must be called by keyword deserialization implementers for deserialization of every sub-node in order to apply extended behaviour operations accordingly. The original ``deserialize`` method of :mod:`colander` schema nodes should not be called directly, otherwise extensions will not be handled. This method will call it after resolving any applicable extension. .. note:: Because sub-nodes are within a non-schema node iterable, the SchemaMeta will not have extracted the destination name for us (ie: map key to compare against). Furthermore, the destination is not directly in the KeywordMapper class, but in its parent where its instance will be dumped according to the keyword resolution. Therefore, regardless of the child, they all have the same parent destination. .. seealso:: - :meth:`_deserialize_keyword` - :class:`ExtendedSchemaNode` """ if not node.name: # pass down the parent name for reference, but with an index to distinguish from it # distinction is also important such that generated schema definitions in OpenAPI don't override each other sub_name = _get_node_name(node, schema_name=True) or str(index) node.name = _get_node_name(self, schema_name=True) + "." + sub_name if isinstance(node, KeywordMapper): return KeywordMapper.deserialize(node, cstruct) return ExtendedSchemaNode.deserialize(node, cstruct)
# pylint: disable=W0222,signature-differs
[docs] def deserialize(self, cstruct): if cstruct is colander.null: if self.required and not VariableSchemaNode.is_variable(self): raise colander.Invalid(self, "Missing required field.") # keyword schema has additional members other than nested in keyword, mapping is required # process any deserialization as mapping schema if self.children: return ExtendedSchemaNode.deserialize(self, colander.null) # otherwise, only null/drop/default are to be processed # since nested keyword schemas are not necessarily mappings, deserialize only extended features # using 'ExtendedSchemaNode.deserialize' would raise "not a mapping" if nested schemas is something else return ExtendedSchemaNode._deserialize_extensions(self, cstruct) # first process the keyword subnodes result = self._deserialize_keyword(cstruct) # if further fields where explicitly added next to the keyword schemas, # validate and apply them as well if isinstance(result, dict) and self.children: mapping_data = super(KeywordMapper, self).deserialize(cstruct) result.update(mapping_data) result = SortableMappingSchema._deserialize_impl(self, result) return result
[docs]class OneOfKeywordSchema(KeywordMapper): """ Allows specifying multiple supported mapping schemas variants for an underlying schema definition. Corresponds to the ``oneOf`` specifier of `OpenAPI` specification. Example:: class Variant1(MappingSchema): [...fields of Variant1...] class Variant2(MappingSchema): [...fields of Variant2...] class RequiredByBoth(MappingSchema): [...fields required by both Variant1 and Variant2...] class OneOfWithRequiredFields(OneOfKeywordSchema, RequiredByBoth): _one_of = (Variant1, Variant2) [...alternatively, field required by all variants here...] In the above example, the validation (ie: ``deserialize``) process will succeed if only one of the ``_one_of`` variants' validator completely succeed, and will fail if every variant fails validation execution. The operation will also fail if more than one validation succeeds. .. note:: Class ``OneOfWithRequiredFields`` in the example is a shortcut variant to generate a specification that represents the pseudo-code ``oneOf([<list-of-objects-with-same-base>])``. The real OpenAPI method to implement the above very commonly occurring situation is as presented by the following pseudo-code:: oneOf[allOf[RequiredByBoth, Variant1], allOf[RequiredByBoth, Variant2]] This is both painful to read and is a lot of extra code to write when you actually expand it all into classes (each ``oneOf/allOf`` is another class). Class :class:`OneOfKeywordSchema` will actually simplify this by automatically making the ``allOf`` definitions for you if it detects other schema nodes than ``oneOf`` specified in the class bases. You can still do the full ``oneOf/allOf`` classes expansion manually though, it will result into the same specification. .. warning:: When ``oneOf/allOf`` automatic expansion occurs during schema generation .. warning:: When calling :meth:`deserialize`, because the validation process requires **exactly one** of the variants to succeed to consider the whole object to evaluate as valid, it is important to insert *more permissive* validators later in the ``_one_of`` iterator (or ``validator`` keyword). For example, having a variant with all fields defined as optional (ie: with ``missing=drop``) inserted as first item in ``_one_of`` will make it always succeed regardless of following variants (since an empty schema with everything dropped is valid for optional-only elements). This would have as side effect of potentially failing the validation if other object are also valid depending on the received schema because the schema cannot be discriminated between many. If this occurs, it means the your schemas are too permissive. In the event that you have very similar schemas that can sometime match except one identifier (e.g.: field ``type`` defining the type of object), consider adding a ``validator`` to each sub-node with explicit values to solve the discrimination problem. As a shortcut, the OpenAPI keyword ``discriminator`` can be provided to try matching as a last resort. For example:: class Animal(ExtendedMappingSchema): name = ExtendedSchemaNode(String()) type = ExtendedSchemaNode(String()) # with explicit definition, this shouldn't be here ## With explicit definitions, each below 'Animal' class should be defined as follows ## type = ExtendedSchemaNode(String(), validator=colander.OneOf(['<animal>'])) class Cat(Animal): [...] # many **OPTIONAL** fields class Dog(Animal): [...] # many **OPTIONAL** fields # With the discriminator keyword, following is possible class SomeAnimal(OneOfMappingSchema): discriminator = { "propertyName": "type", # correspond to 'type' of 'Animal' "mapping": { "cat": Cat, "dog": Dog # map expected values to target schemas } } _one_of = [ Cat(), Dog(), ] .. note:: Keyword ``discriminator`` supports a map of key-string to schemas-type as presented in the example, and the key must be located at the top level of the mapping. If only ``discriminator = "<field>"`` is provided, the definition will be created automatically using the ``example`` (which should be only the matching value) of the corresponding field of each node within the ``_one_of`` mapping. When multiple valid schemas are matched against the input data, the error will be raised and returned with corresponding erroneous elements for each sub-schema (fully listed). .. seealso:: - :class:`AllOfKeywordSchema` - :class:`AnyOfKeywordSchema` - :class:`NotKeywordSchema` """
[docs] _keyword_schemas_only_object = False
[docs] _keyword = "_one_of"
[docs] _discriminator = "discriminator"
[docs] discriminator = None
@classmethod @abstractmethod
[docs] def _one_of(cls): # type: () -> Iterable[Union[colander.SchemaNode, Type[colander.SchemaNode]]] # noqa: W0212 """ Sequence of applicable schema nested under the ``oneOf`` keyword. Must be overridden in the schema definition using it. """ raise SchemaNodeTypeError(f"Missing '{cls._keyword}' keyword for schema '{cls}'.")
def __init__(self, *args, **kwargs): discriminator = getattr(self, self._discriminator, None) discriminator = kwargs.pop(self._discriminator, discriminator) super(OneOfKeywordSchema, self).__init__(*args, **kwargs) discriminator_spec = None if discriminator: schema_name = _get_node_name(self, schema_name=True) keyword = self.get_keyword_name() discriminator_spec = discriminator if isinstance(discriminator, str): discriminator_spec = {"propertyName": discriminator} if isinstance(discriminator_spec, dict) and "mapping" not in discriminator_spec: mapping = {} for node in self.get_keyword_items(): node_fields = [getattr(field, "example", colander.null) for field in node.children if field.name == discriminator_spec["propertyName"]] if len(node_fields) != 1: continue example = node_fields[0] if example is not colander.null: if example in mapping: raise SchemaNodeTypeError( "Keyword schema '{}' of type '{}' specification with 'discriminator' attempts " "to refer to duplicate example values '{}' between '{}' and '{}'".format( schema_name, keyword, example, _get_node_name(mapping[example], schema_name=True), _get_node_name(node, schema_name=True), ) ) mapping[example] = node discriminator_spec["mapping"] = mapping if not ( isinstance(discriminator_spec, dict) and all(prop in discriminator_spec for prop in ["propertyName", "mapping"]) and isinstance(discriminator_spec["propertyName"], str) and isinstance(discriminator_spec["mapping"], dict) and len(discriminator_spec["mapping"]) and all(isinstance(node, colander.SchemaNode) and node.schema_type is colander.Mapping for name, node in discriminator_spec["mapping"].items()) ): raise SchemaNodeTypeError( "Keyword schema '{}' of type '{}' specification with 'discriminator' must be a string " "or dictionary with 'propertyName' and 'mapping' of target value to schema nodes, " "but was specified as '{!s}".format(schema_name, keyword, discriminator)) setattr(self, self._discriminator, discriminator_spec) @property
[docs] def discriminator_spec(self): return getattr(self, self._discriminator, None)
[docs] def _deserialize_keyword(self, cstruct): """ Test each possible case, return all corresponding errors if not exactly one of the possibilities is valid. """ invalid_one_of = {} valid_one_of = [] valid_nodes = [] for index, schema_class in enumerate(self._one_of): # noqa try: schema_class = _make_node_instance(schema_class) result = self._deserialize_subnode(schema_class, cstruct, index) if result is colander.drop: continue valid_one_of.append(result) valid_nodes.append(schema_class) except colander.Invalid as invalid: invalid_one_of.update({_get_node_name(invalid.node, schema_name=True): invalid.asdict()}) message = ( f"Incorrect type must be one of: {list(invalid_one_of)}. Errors for each case: {invalid_one_of}" ) if valid_one_of: # if found only one, return it, otherwise try to discriminate if len(valid_one_of) == 1: return valid_one_of[0] message = ( "Incorrect type cannot distinguish between multiple valid schemas. " "Must be only one of: {}.".format([_get_node_name(node, schema_name=True) for node in valid_nodes]) ) discriminator = self.discriminator_spec if discriminator: # try last resort solve valid_discriminated = [] error_discriminated = {} for i, obj in enumerate(valid_one_of): node = valid_nodes[i] node_name = _get_node_name(node) node_field = [child for child in node.children if child.name == discriminator["propertyName"]] if len(node_field) == 1: error_discriminated.update({node_name: obj}) node_value = node_field[0].example node_discriminated = discriminator["mapping"].get(node_value) if node_discriminated and isinstance(obj, type(node_discriminated)): valid_discriminated.append(obj) if len(valid_discriminated) == 1: return valid_discriminated[0] elif len(valid_discriminated) > 1: invalid_one_of = error_discriminated message = ( "Incorrect type cannot discriminate between multiple valid schemas. " "Must be only one of: {}.".format(list(invalid_one_of)) ) raise colander.Invalid(node=self, msg=message, value=discriminator) # because some schema nodes will convert types during deserialize without error, # attempt to discriminate base-type values compared with tested value # (e.g.: discriminate between float vs numerical string allowed schema variations) if not isinstance(cstruct, (dict, set, list, tuple)): # pylint: disable=C0123 valid_values = list(filter(lambda c: c == cstruct and type(c) == type(cstruct), valid_one_of)) if len(valid_values) == 1: return valid_values[0] message = ( "Incorrect type cannot differentiate between multiple base-type valid schemas. " "Must be only one of: {}.".format(valid_values) ) # not a single valid sub-node was found if self.missing is colander.drop: return colander.drop raise colander.Invalid(node=self, msg=message, value=cstruct)
[docs]class AllOfKeywordSchema(KeywordMapper): """ Allows specifying all the required partial mapping schemas for an underlying complete schema definition. Corresponds to the ``allOf`` specifier of `OpenAPI` specification. Example:: class RequiredItem(ExtendedMappingSchema): item = ExtendedSchemaNode(String()) class RequiredType(ExtendedMappingSchema): type = ExtendedSchemaNode(String()) class AllRequired(AnyKeywordSchema): _all_of = [RequiredItem(), RequiredType()] Value parsed with schema this definition will be valid only when every since one of the sub-schemas is valid. Any sub-schema raising an invalid error for any reason with make the whole schema validation fail. .. seealso:: - :class:`OneOfKeywordSchema` - :class:`AnyOfKeywordSchema` - :class:`NotKeywordSchema` """
[docs] _keyword_schemas_only_object = True
[docs] _keyword_schemas_same_struct = True
[docs] _keyword = "_all_of"
@classmethod @abstractmethod
[docs] def _all_of(cls): # type: () -> Iterable[Union[colander.SchemaNode, Type[colander.SchemaNode]]] # noqa: W0212 """ Sequence of applicable schema nested under the ``allOf`` keyword. Must be overridden in the schema definition using it. """ raise SchemaNodeTypeError(f"Missing '{cls._keyword}' keyword for schema '{cls}'.")
[docs] def _deserialize_keyword(self, cstruct): """ Test each possible case, return all corresponding errors if any of the possibilities is invalid. """ required_all_of = {} missing_all_of = {} merged_all_of = {} for index, schema_class in enumerate(self._all_of): # noqa try: schema_class = _make_node_instance(schema_class) # update items with new ones required_all_of.update({_get_node_name(schema_class, schema_name=True): str(schema_class)}) result = self._deserialize_subnode(schema_class, cstruct, index) if result is colander.drop: if schema_class.missing is colander.drop: continue if isinstance(schema_class.default, dict): result = schema_class.default else: raise colander.Invalid(node=schema_class, msg="Schema is missing when required.", value=result) merged_all_of.update(result) except colander.Invalid as invalid: missing_all_of.update({_get_node_name(invalid.node, schema_name=True): str(invalid)}) if missing_all_of: # if anything failed, the whole definition is invalid in this case message = ( "Incorrect type must represent all of: {}. Missing following cases: {}" .format(list(required_all_of), list(missing_all_of)) ) raise colander.Invalid(node=self, msg=message) return merged_all_of
[docs]class AnyOfKeywordSchema(KeywordMapper): """ Allows specifying all mapping schemas that can be matched for an underlying schema definition. Corresponds to the ``anyOf`` specifier of `OpenAPI` specification. Contrary to :class:`OneOfKeywordSchema` that MUST be validated with exactly one schema, this definition will continue parsing all possibilities and apply validated sub-schemas on top of each other. Not all schemas have to be valid like in the case of :class:`AllOfKeywordSchema` to succeed, as long as at least one of them is valid. Example:: class RequiredItem(ExtendedMappingSchema): item = ExtendedSchemaNode(String()) class RequiredType(ExtendedMappingSchema): type = ExtendedSchemaNode(String()) class RequiredFields(ExtendedMappingSchema): field_str = ExtendedSchemaNode(String()) field_int = ExtendedSchemaNode(Integer()) class AnyRequired(AnyKeywordSchema): _any_of = [RequiredItem(), RequiredType(), RequiredFields()] # following is valid because their individual parts have all required sub-fields, result is their composition AnyRequired().deserialize({"type": "test", "item": "valid"}) # result: {"type": "test", "item": "valid"} # following is also valid because even though 'item' is missing, the 'type' is present AnyRequired().deserialize({"type": "test"}) # result: {"type": "test"} # following is invalid because every one of the sub-field of individual parts are missing AnyRequired().deserialize({"type": "test"}) # following is invalid because fields of 'RequiredFields' are only partially fulfilled AnyRequired().deserialize({"field_str": "str"}) # following is valid because although fields of 'RequiredFields' are not all fulfilled, 'RequiredType' is valid AnyRequired().deserialize({"field_str": "str", "type": "str"}) # result: {"type": "test"} # following is invalid because 'RequiredFields' field 'field_int' is incorrect schema type AnyRequired().deserialize({"field_str": "str", "field_int": "str"}) # following is valid, but result omits 'type' because its schema-type is incorrect, while others are valid AnyRequired().deserialize({"field_str": "str", "field_int": 1, "items": "fields", "type": 1}) # result: {"field_str": "str", "field_int": 1, "items": "fields"} .. warning:: Because valid items are applied on top of each other by merging fields during combinations, conflicting field names of any valid schema will contain only the final valid parsing during deserialization. .. seealso:: - :class:`OneOfKeywordSchema` - :class:`AllOfKeywordSchema` - :class:`NotKeywordSchema` """
[docs] _keyword_schemas_only_object = False
[docs] _keyword_schemas_same_struct = True
[docs] _keyword = "_any_of"
@classmethod @abstractmethod
[docs] def _any_of(cls): # type: () -> Iterable[Union[colander.SchemaNode, Type[colander.SchemaNode]]] # noqa: W0212 """ Sequence of applicable schema nested under the ``anyOf`` keyword. Must be overridden in the schema definition using it. """ raise SchemaNodeTypeError(f"Missing '{cls._keyword}' keyword for schema '{cls}'.")
[docs] def _deserialize_keyword(self, cstruct): """ Test each possible case, return if no corresponding schema was found. """ option_any_of = {} merged_any_of = colander.null invalid_any_of = colander.Invalid(node=self) for index, schema_class in enumerate(self._any_of): # noqa try: schema_class = _make_node_instance(schema_class) # update items with new ones option_any_of.update({_get_node_name(schema_class, schema_name=True): str(schema_class)}) result = self._deserialize_subnode(schema_class, cstruct, index) if result not in (colander.drop, colander.null): # technically not supposed to have 'Sequence' type since they can only have one child # only possibility is all similar objects or all literals because of '_keyword_schemas_same_struct' if schema_class.schema_type is colander.Mapping: if merged_any_of is colander.null: merged_any_of = {} merged_any_of.update(result) else: # schema nodes override one another if valid for multiple schemas merged_any_of = result except colander.Invalid as invalid: invalid_any_of.add(invalid) # if nothing could be resolved, verify for a default value if merged_any_of is colander.null: merged_any_of = self.default # nothing succeeded, the whole definition is invalid in this case if merged_any_of is colander.null: invalid_any_of.msg = ( f"Incorrect type must represent any of: {list(option_any_of)}. All missing from: {cstruct}" ) raise invalid_any_of return merged_any_of
[docs]class NotKeywordSchema(KeywordMapper): """ Allows specifying specific schema conditions that fails underlying schema definition validation if present. Corresponds to the ``not`` specifier of `OpenAPI` specification. Example:: class RequiredItem(ExtendedMappingSchema): item = ExtendedSchemaNode(String()) class MappingWithType(ExtendedMappingSchema): type = ExtendedSchemaNode(String()) class MappingWithoutType(NotKeywordSchema, RequiredItem): _not = [MappingWithType()] class MappingOnlyNotType(NotKeywordSchema): _not = [MappingWithType()] # following will raise invalid error even if 'item' is valid because 'type' is also present MappingWithoutType().deserialize({"type": "invalid", "item": "valid"}) # following will return successfully with only 'item' because 'type' was not present MappingWithoutType().deserialize({"item": "valid", "value": "ignore"}) # result: {"item": "valid"} # following will return an empty mapping dropping 'item' since it only needs to ensure 'type' was not present, # but did not provide any additional fields requirement from other class inheritances MappingOnlyNotType().deserialize({"item": "valid"}) # result: {} .. seealso:: - :class:`OneOfKeywordSchema` - :class:`AllOfKeywordSchema` - :class:`AnyOfKeywordSchema` """
[docs] _keyword_schemas_only_object = True
[docs] _keyword_schemas_same_struct = True
[docs] _keyword = "_not"
@classmethod @abstractmethod
[docs] def _not(cls): # type: () -> Iterable[Union[colander.SchemaNode, Type[colander.SchemaNode]]] # noqa: W0212 """ Sequence of applicable schema nested under the ``not`` keyword. Must be overridden in the schema definition using it. """ raise SchemaNodeTypeError(f"Missing '{cls._keyword}' keyword for schema '{cls}'.")
[docs] def _deserialize_keyword(self, cstruct): """ Raise if any sub-node schema that should NOT be present was successfully validated. """ invalid_not = {} for index, schema_class in enumerate(self._not): # noqa try: schema_class = _make_node_instance(schema_class) result = self._deserialize_subnode(schema_class, cstruct, index) if isinstance(result, dict) and not len(result): continue # allow empty result meaning every item was missing and dropped invalid_names = [node.name for node in schema_class.children] invalid_not.update({_get_node_name(schema_class, schema_name=True): invalid_names}) except colander.Invalid: pass # error raised as intended when missing field is not present if invalid_not: message = f"Value contains not allowed fields from schema conditions: {invalid_not}" raise colander.Invalid(node=self, msg=message, value=cstruct) # If schema was a plain NotKeywordSchema, the result will be empty as it serves only to validate # that the subnodes are not present. Otherwise, if it derives from other mapping classes, apply them. # If deserialization was not applied here, everything in the original cstruct would bubble up. return ExtendedMappingSchema.deserialize(self, cstruct)
[docs]class KeywordTypeConverter(TypeConverter): """ Generic keyword converter that builds schema with a list of sub-schemas under the keyword. """
[docs] def convert_type(self, schema_node): keyword = schema_node.get_keyword_name() keyword_schema = { keyword: [] } for item_schema in schema_node.get_keyword_items(): obj_instance = _make_node_instance(item_schema) obj_converted = self.dispatcher(obj_instance) keyword_schema[keyword].append(obj_converted) return keyword_schema
[docs]class OneOfKeywordTypeConverter(KeywordTypeConverter): """ Object converter that generates the ``oneOf`` keyword definition. This object does a bit more work than other :class:`KeywordTypeConverter` as it handles the shorthand definition as described in :class:`OneOfKeywordSchema` .. seealso:: - :class:`OneOfKeywordSchema` """
[docs] def convert_type(self, schema_node): # type: (OneOfKeywordSchema) -> Dict keyword = schema_node.get_keyword_name() one_of_obj = { keyword: [] } for item_schema in schema_node.get_keyword_items(): item_obj = _make_node_instance(item_schema) # shortcut definition of oneOf[allOf[],allOf[]] mix, see OneOfKeywordSchema docstring # (eg: schema fields always needed regardless of other fields supplied by each oneOf schema) if len(getattr(schema_node, "children", [])): if not isinstance(schema_node, colander.MappingSchema): raise ConversionTypeError( f"Unknown base type to convert oneOf schema item is no a mapping: {type(schema_node)}" ) # specific oneOf sub-item, will be processed by itself during dispatch of sub-item of allOf # rewrite the title of that new sub-item schema from the original title to avoid conflict schema_title = _get_node_name(schema_node, schema_name=True) # generate the new nested definition of keywords using schema node bases and children item_title = _get_node_name(item_obj, schema_name=True) # NOTE: to avoid potential conflict of schema reference definitions with other existing ones, # use an invalid character that cannot exist in Python class name defining the schema titles one_of_title = schema_title + "." + item_title shared_title = schema_title + ".Shared" obj_req_title = item_title + ".AllOf" # fields that are shared across all the oneOf sub-items # pass down the original title of that object to refer to that schema reference obj_shared = ExtendedMappingSchema(title=shared_title) obj_shared.children = schema_node.children obj_one_of = item_obj.clone() obj_one_of.title = one_of_title all_of = AllOfKeywordSchema(title=obj_req_title, _all_of=[obj_shared, obj_one_of]) obj_converted = self.dispatcher(all_of) else: obj_converted = self.dispatcher(item_obj) one_of_obj[keyword].append(obj_converted) discriminator = schema_node.discriminator_spec if schema_node.discriminator: discriminator_spec = {"propertyName": discriminator["propertyName"], "mapping": {}} for value, node in discriminator["mapping"].items(): discriminator_spec["mapping"][value] = node.title one_of_obj["discriminator"] = discriminator_spec return one_of_obj
[docs]class AllOfKeywordTypeConverter(KeywordTypeConverter): """ Object converter that generates the ``allOf`` keyword definition. """
[docs]class AnyOfKeywordTypeConverter(KeywordTypeConverter): """ Object converter that generates the ``anyOf`` keyword definition. """
[docs]class NotKeywordTypeConverter(KeywordTypeConverter): """ Object converter that generates the ``not`` keyword definition. """
[docs] def convert_type(self, schema_node): result = ObjectTypeConverter(self.dispatcher).convert_type(schema_node) result["additionalProperties"] = False return result
[docs]class VariableObjectTypeConverter(ObjectTypeConverter): """ Object convertor with ``additionalProperties`` for each ``properties`` marked as :class:`VariableSchemaNode`. """
[docs] def convert_type(self, schema_node): converted = super(VariableObjectTypeConverter, self).convert_type(schema_node) converted.setdefault("additionalProperties", {}) if self.dispatcher.openapi_spec == 3: for sub_node in schema_node.children: if VariableSchemaNode.is_variable(sub_node): if isinstance(sub_node, KeywordMapper): # keyword mapping properties = converted["properties"].pop(sub_node.name) keyword = sub_node.get_keyword_name() kw_props = properties.pop(keyword) converted["additionalProperties"].update({keyword: kw_props}) for prop in properties: # re-add other fields like title if not already defined in keyword container converted.setdefault(prop, properties[prop]) else: # normal mapping converted["additionalProperties"].update( {sub_node.name: converted["properties"].pop(sub_node.name)} ) if sub_node.name in converted.get("required", []): converted["required"].remove(sub_node.name) return converted
[docs]class DecimalTypeConverter(NumberTypeConverter):
[docs] format = "decimal"
[docs]class MoneyTypeConverter(DecimalTypeConverter):
[docs] pattern = "^[0-9]+.[0-9]{2}$"
[docs] convert_validator = ValidatorConversionDispatcher( convert_range_validator(colander.Range(min=0)), convert_regex_validator(colander.Regex(pattern, msg="Number must be formatted as currency decimal."))
) # TODO: replace directly in original cornice_swagger # (see: https://github.com/Cornices/cornice.ext.swagger/issues/133)
[docs]class OAS3TypeConversionDispatcher(TypeConversionDispatcher):
[docs] openapi_spec = 3
def __init__(self, custom_converters=None, default_converter=None): # type: (Optional[Dict[colander.SchemaType, TypeConverter]], Optional[TypeConverter]) -> None self.keyword_converters = { OneOfKeywordSchema: OneOfKeywordTypeConverter, AllOfKeywordSchema: AllOfKeywordTypeConverter, AnyOfKeywordSchema: AnyOfKeywordTypeConverter, NotKeywordSchema: NotKeywordTypeConverter, } self.keyword_validators = { colander.OneOf: OneOfKeywordSchema, colander.Any: AnyOfKeywordSchema, colander.All: AllOfKeywordSchema, colander.NoneOf: NotKeywordSchema, } # NOTE: define items to inherit variable/extended functionalities, # when merging in original cornice_swagger, it should be the 'default' behaviour of object converter # user custom converters can override everything, but they must use extended classes to use extra features extended_converters = { colander.Mapping: VariableObjectTypeConverter, colander.Decimal: DecimalTypeConverter, colander.Money: MoneyTypeConverter, } extended_converters.update(self.keyword_converters) if custom_converters: extended_converters.update(custom_converters) super(OAS3TypeConversionDispatcher, self).__init__(extended_converters, default_converter) def __call__(self, schema_node): # type: (colander.SchemaNode) -> Dict[str, Any] schema_type = schema_node.typ schema_type = type(schema_type) # dispatch direct reference to keyword schemas converter_class = None for base_class in self.keyword_converters: if base_class in inspect.getmro(type(schema_node)): converter_class = self.keyword_converters.get(base_class) break if converter_class is None and self.openapi_spec == 3: # dispatch indirect conversions specified by MappingSchema/SequenceSchema # using a colander validator as argument matching keyword schemas # (eg: MappingSchema(validator=colander.OneOf([Obj1, Obj2])) ) if isinstance(schema_node, (colander.MappingSchema, colander.SequenceSchema)): if isinstance(schema_node.validator, tuple(self.keyword_validators)): keyword_class = self.keyword_validators[type(schema_node.validator)] keyword_items = getattr(schema_node.validator, "choices", getattr(schema_node.validator, "validators")) keyword_kwargs = {keyword_class.get_keyword_name(): keyword_items} keyword_schema = keyword_class(**keyword_kwargs) converted = self(keyword_schema) # noqa return converted if converter_class is None: converter_class = self.converters.get(schema_type) if converter_class is None: # find derived extended schema type if any(issubclass(schema_type, s_type) for s_type in self.converters): for sub_type in inspect.getmro(schema_type)[1:]: converter_class = self.converters.get(sub_type) if converter_class is not None: break elif self.default_converter: converter_class = self.default_converter else: raise NoSuchConverter(f"schema_type: {schema_type}") converter = converter_class(self) converted = converter(schema_node) # noqa # TODO: only this part is actually 'custom' for our use # rest on top is to be integrated in original 'cornice_swagger' if schema_node.title: # ignore the names key use as useful keywords to define location if schema_node.name in ["header", "body", "querystring", "path"]: converted["title"] = schema_node.title else: # otherwise use either the explicitly provided title or name # colander capitalizes the title, which makes it wrong most of the time # when using CamelCase or camelBack schema definitions if isinstance(schema_node.raw_title, str): converted["title"] = schema_node.title else: converted["title"] = schema_node.name if converted.get("default") is colander.drop: converted.pop("default") xml = getattr(schema_node, "xml", None) if isinstance(xml, dict): converted["xml"] = xml return converted