howler-api 3.0.0.dev374__py3-none-any.whl

howler/odm/base.py

@@ -0,0 +1,1543 @@
+"""HOWLER's built in Object Document Model tool.
+
+The classes in this module can be composed to build database
+independent data models in python. This gives us:
+- single source of truth for our data schemas
+- database independent serialization
+- type checking
+
+
+"""
+
+from __future__ import annotations
+
+import copy
+import json
+import re
+import typing
+from datetime import datetime
+from enum import Enum as PyEnum
+from enum import EnumMeta
+from typing import Any as _Any
+from typing import Dict, Tuple, Union
+from venv import logger
+
+import arrow
+import validators
+from dateutil.tz import tzutc
+
+from howler.common import loader
+from howler.common.exceptions import HowlerKeyError, HowlerNotImplementedError, HowlerTypeError, HowlerValueError
+from howler.common.net import is_valid_domain, is_valid_ip
+from howler.utils.dict_utils import flatten, recursive_update
+from howler.utils.isotime import now_as_iso
+from howler.utils.uid import get_random_id
+
+BANNED_FIELDS = {
+    "_id",
+    "__access_grp1__",
+    "__access_lvl__",
+    "__access_req__",
+    "__access_grp2__",
+}
+DATEFORMAT = "%Y-%m-%dT%H:%M:%S.%fZ"
+FIELD_SANITIZER = re.compile("^[a-z][a-z0-9_-]*$")
+FLATTENED_OBJECT_SANITIZER = re.compile("^[a-z][a-z0-9_.]*$")
+NOT_INDEXED_SANITIZER = re.compile("^[A-Za-z0-9_ -]*$")
+UTC_TZ = tzutc()
+
+DOMAIN_REGEX = (
+    r"(?:(?:[A-Za-z0-9\u00a1-\uffff][A-Za-z0-9\u00a1-\uffff_-]{0,62})?[A-Za-z0-9\u00a1-\uffff]\.)+"
+    r"(?:xn--)?(?:[A-Za-z0-9\u00a1-\uffff]{2,}\.?)"
+)
+DOMAIN_ONLY_REGEX = f"^{DOMAIN_REGEX}$"
+EMAIL_REGEX = f"^[a-zA-Z0-9!#$%&'*+/=?^_‘{{|}}~-]+(?:\\.[a-zA-Z0-9!#$%&'*+/=?^_‘{{|}}~-]+)*@({DOMAIN_REGEX})$"
+IPV4_REGEX = r"(?:(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)\.){3}(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)"
+IPV6_REGEX = (
+    r"(?:(?:[0-9a-fA-F]{1,4}:){7,7}[0-9a-fA-F]{1,4}|(?:[0-9a-fA-F]{1,4}:){1,7}:|"
+    r"(?:[0-9a-fA-F]{1,4}:){1,6}:[0-9a-fA-F]{1,4}|(?:[0-9a-fA-F]{1,4}:){1,5}(?::[0-9a-fA-F]{1,4}){1,2}|"
+    r"(?:[0-9a-fA-F]{1,4}:){1,4}(?::[0-9a-fA-F]{1,4}){1,3}|(?:[0-9a-fA-F]{1,4}:){1,3}(?::[0-9a-fA-F]{1,4}){1,4}|"
+    r"(?:[0-9a-fA-F]{1,4}:){1,2}(?::[0-9a-fA-F]{1,4}){1,5}|[0-9a-fA-F]{1,4}:(?:(?::[0-9a-fA-F]{1,4}){1,6})|"
+    r":(?:(?::[0-9a-fA-F]{1,4}){1,7}|:)|fe80:(?::[0-9a-fA-F]{0,4}){0,4}%[0-9a-zA-Z]{1,}|"
+    r"::(?:ffff(?::0{1,4}){0,1}:){0,1}(?:(?:25[0-5]|(?:2[0-4]|1{0,1}[0-9]){0,1}[0-9])\.){3,3}(?:25[0-5]|"
+    r"(?:2[0-4]|1{0,1}[0-9]){0,1}[0-9])|(?:[0-9a-fA-F]{1,4}:){1,4}:(?:(?:25[0-5]|"
+    r"(?:2[0-4]|1{0,1}[0-9]){0,1}[0-9])\.){3,3}(?:25[0-5]|(?:2[0-4]|1{0,1}[0-9]){0,1}[0-9]))"
+)
+IP_REGEX = f"(?:{IPV4_REGEX}|{IPV6_REGEX})"
+IP_ONLY_REGEX = f"^{IP_REGEX}$"
+PRIVATE_IP = (
+    r"(?:(?:127|10)(?:\.(?:[2](?:[0-5][0-5]|[01234][6-9])|[1][0-9][0-9]|[1-9][0-9]|[0-9])){3})|"
+    r"(?:172\.(?:1[6-9]|2[0-9]|3[0-1])(?:\.(?:2[0-4][0-9]|25[0-5]|[1][0-9][0-9]|[1-9][0-9]|[0-9])){2}|"
+    r"(?:192\.168(?:\.(?:25[0-5]|2[0-4][0-9]|1[0-9][0-9]|[1-9][0-9]|[0-9])){2}))"
+)
+PHONE_REGEX = r"^(\+?\d{1,2})?[ .-]?(\(\d{3}\)|\d{3})[ .-](\d{3})[ .-](\d{4})$"
+SSDEEP_REGEX = r"^[0-9]{1,18}:[a-zA-Z0-9/+]{0,64}:[a-zA-Z0-9/+]{0,64}$"
+MD5_REGEX = r"^[a-f0-9]{32}$"
+SHA1_REGEX = r"^[a-f0-9]{40}$"
+SHA256_REGEX = r"^[a-f0-9]{64}$"
+HOWLER_HASH_REGEX = r"^[a-f0-9]{1,64}$"
+MAC_REGEX = r"^(?:(?:[0-9a-f]{2}-){5}[0-9a-f]{2}|(?:[0-9a-f]{2}:){5}[0-9a-f]{2})$"
+URI_PATH = r"(?:[/?#]\S*)"
+FULL_URI = f"^((?:(?:[A-Za-z]*:)?//)?(?:\\S+(?::\\S*)?@)?({IP_REGEX}|{DOMAIN_REGEX})(?::\\d{{2,5}})?){URI_PATH}?$"
+PLATFORM_REGEX = r"^(Windows|Linux|MacOS|Android|iOS)$"
+PROCESSOR_REGEX = r"^x(64|86)$"
+
+
+def flat_to_nested(data: dict[str, _Any]) -> dict[str, _Any]:
+    sub_data: dict[str, _Any] = {}
+    nested_keys = []
+    for key, value in data.items():
+        if "." in key:
+            child, sub_key = key.split(".", 1)
+            nested_keys.append(child)
+            try:
+                sub_data[child][sub_key] = value
+            except (KeyError, TypeError):
+                sub_data[child] = {sub_key: value}
+        else:
+            sub_data[key] = value
+
+    for key in nested_keys:
+        sub_data[key] = flat_to_nested(sub_data[key])
+
+    return sub_data
+
+
+class KeyMaskException(HowlerKeyError):
+    pass
+
+
+class _Field:
+    def __init__(
+        self,
+        name=None,
+        index=None,
+        store=None,
+        copyto=None,
+        default=None,
+        description=None,
+        deprecated_description=None,
+        reference=None,
+        optional=False,
+        deprecated=False,
+    ):
+        self.index = index
+        self.store = store
+        self.multivalued = False
+        self.copyto = []
+        if isinstance(copyto, str):
+            self.copyto.append(copyto)
+        elif copyto:
+            self.copyto.extend(copyto)
+
+        self.name = name
+        self.parent_name = None
+        self.getter_function = None
+        self.setter_function = None
+        self.description = description
+        self.reference = reference
+        self.optional = optional
+        self.deprecated = deprecated
+        self.deprecated_description = deprecated_description
+
+        self.default = default
+        self.default_set = default is not None
+
+    # noinspection PyProtectedMember
+    def __get__(self, obj, objtype=None):
+        """Read the value of this field from the model instance (obj)."""
+        if obj is None:
+            return obj
+        if self.name in obj._odm_removed:
+            raise KeyMaskException(self.name)
+        if self.getter_function is not None:
+            return self.getter_function(obj, obj._odm_py_obj[self.name.rstrip("_")])
+
+        return obj._odm_py_obj[self.name.rstrip("_")]
+
+    # noinspection PyProtectedMember
+    def __set__(self, obj, value):
+        """Set the value of this field, calling a setter method if available."""
+        if self.name in obj._odm_removed:
+            raise KeyMaskException(self.name)
+        value = self.check(value)
+        if self.setter_function is not None:
+            value = self.setter_function(obj, value)
+        obj._odm_py_obj[self.name.rstrip("_")] = value
+
+    def getter(self, method):
+        """Decorator to create getter method for a field."""
+        out = copy.deepcopy(self)
+        out.getter_function = method
+        return out
+
+    def setter(self, method):
+        """Let fields be used as a decorator to define a setter method.
+
+        >>> expiry = Date()
+        >>>
+        >>> # noinspection PyUnusedLocal,PyUnresolvedReferences
+        >>> @expiry.setter
+        >>> def expiry(self, assign, value):
+        >>>     assert value
+        >>>     assign(value)
+        """
+        out = copy.deepcopy(self)
+        out.setter_function = method
+        return out
+
+    def apply_defaults(self, index, store):
+        """Used by the model decorator to pass through default parameters."""
+        if self.index is None:
+            self.index = index
+        if self.store is None:
+            self.store = store
+
+    def fields(self):
+        """Return the subfields/modified field data.
+
+        For simple fields this is an identity function.
+        """
+        return {"": self}
+
+    def check(self, value, **kwargs):
+        raise HowlerNotImplementedError(
+            "This function is not defined in the default field. " "Each fields has to have their own definition"
+        )
+
+    def __repr__(self) -> str:
+        keys = [
+            key
+            for key in self.__dir__()
+            if not key.startswith("_") and not callable(getattr(self, key)) and getattr(self, key) is not None
+        ]
+        return f"{type(self).__name__}({', '.join([f'{key}={str(getattr(self, key))}' for key in keys])})"
+
+
+class _DeletedField:
+    pass
+
+
+class Date(_Field):
+    """A field storing a datetime value."""
+
+    def check(self, value, context=[], **kwargs):
+        if value is None:
+            return None
+
+        if value == "NOW":
+            value = now_as_iso()
+
+        try:
+            try:
+                return datetime.strptime(value, DATEFORMAT).replace(tzinfo=UTC_TZ)
+            except (TypeError, ValueError):
+                return arrow.get(value).datetime
+        except Exception as e:
+            raise HowlerValueError(f"[{'.'.join(context) or self.name}]: {str(e)}")
+
+
+class Boolean(_Field):
+    """A field storing a boolean value."""
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        try:
+            return bool(value)
+        except ValueError as e:
+            raise HowlerValueError(f"[{'.'.join(context) or self.name}]: {str(e)}")
+
+
+class Json(_Field):
+    """A field storing serializeable structure with their JSON encoded representations.
+
+    Examples: metadata
+    """
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if not isinstance(value, str):
+            try:
+                return json.dumps(value)
+            except (ValueError, OverflowError, TypeError) as e:
+                raise HowlerValueError(f"[{'.'.join(context) or self.name}]: {str(e)}")
+
+        return value
+
+
+class Keyword(_Field):
+    """A field storing a short string with a technical interpretation.
+
+    Examples: file hashes, service names, document ids
+    """
+
+    def check(self, value, context=[], **kwargs):
+        # We have a special case for bytes here due to how often strings and bytes
+        # get mixed up in python apis
+        if self.optional and value is None:
+            return None
+
+        if isinstance(value, bytes):
+            raise HowlerValueError(f"[{'.'.join(context) or self.name}] Keyword doesn't accept bytes values")
+
+        if value == "" or value is None:
+            if self.default_set:
+                value = self.default
+            else:
+                raise HowlerValueError(
+                    f"[{'.'.join(context) or self.name}] Empty strings are not allowed without defaults"
+                )
+
+        if value is None:
+            return None
+
+        return str(value)
+
+
+class EmptyableKeyword(_Field):
+    """A keyword which allow to differentiate between empty and None values."""
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        # We have a special case for bytes here due to how often strings and bytes
+        # get mixed up in python apis
+        if isinstance(value, bytes):
+            raise HowlerValueError(f"[{'.'.join(context) or self.name}] EmptyableKeyword doesn't accept bytes values")
+
+        if value is None and self.default_set:
+            value = self.default
+
+        if value is None:
+            return None
+
+        return str(value)
+
+
+class UpperKeyword(Keyword):
+    """A field storing a short uppercase string with a technical interpretation."""
+
+    def check(self, value, context=[], **kwargs):
+        kw_val = super().check(value, context=context, **kwargs)
+
+        if kw_val is None:
+            return None
+
+        return kw_val.upper()
+
+
+class LowerKeyword(Keyword):
+    """
+    A field storing a short lowercase string with a technical interpretation.
+    """
+
+    def check(self, value, context=[], **kwargs):
+        kw_val = super().check(value, context=context, **kwargs)
+
+        if kw_val is None:
+            return None
+
+        return kw_val.lower()
+
+
+class CaseInsensitiveKeyword(Keyword):
+    """
+    A field storing a string with a technical interpretation, but is case-insensitive when searching.
+    """
+
+
+class Any(Keyword):
+    """A field that can hold any value whatsoever but which is stored as a
+    Keyword in the datastore index
+    """
+
+    def __init__(self, *args, **kwargs):
+        kwargs["index"] = False
+        kwargs["store"] = False
+        super().__init__(*args, **kwargs)
+
+    def check(self, value, **_):
+        return value
+
+
+class ValidatedKeyword(Keyword):
+    """Keyword field which the values are validated by a regular expression"""
+
+    def __init__(self, validation_regex, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.validation_regex = re.compile(validation_regex)
+
+    def __deepcopy__(self, memo=None):
+        # NOTE: This deepcopy code does not work with a sub-class that add args of kwargs that should be copied.
+        #       If that is the case, the sub-class should implement its own deepcopy function.
+        valid_fields = ["name", "index", "store", "copyto", "default", "description"]
+        if "validation_regex" in self.__class__.__init__.__code__.co_varnames:
+            return self.__class__(
+                self.validation_regex.pattern,
+                **{k: v for k, v in self.__dict__.items() if k in valid_fields},
+            )
+        else:
+            return self.__class__(**{k: v for k, v in self.__dict__.items() if k in valid_fields})
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if not value:
+            if self.default_set:
+                value = self.default
+            else:
+                raise HowlerValueError(
+                    f"[{'.'.join(context) or self.name}]: Empty strings are not allowed without defaults"
+                )
+
+        if value is None:
+            return value
+
+        if not self.validation_regex.match(value):
+            raise HowlerValueError(
+                f"[{'.'.join(context) or self.name}]: '{value}' not match the "
+                f"validator: {self.validation_regex.pattern}"
+            )
+
+        return str(value)
+
+
+class IP(Keyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.validation_regex = re.compile(IP_ONLY_REGEX)
+
+    def check(self, value, context=[], **kwargs):
+        if not value:
+            return None
+
+        if not self.validation_regex.match(value):
+            raise HowlerValueError(
+                f"[{'.'.join(context) or self.name}]: '{value}' not match the "
+                f"validator: {self.validation_regex.pattern}"
+            )
+
+        return value
+
+
+class Domain(Keyword):
+    def __init__(self, *args, strict=True, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.strict = strict
+
+    def check(self, value, context=[], **kwargs):
+        if not value:
+            return None
+
+        domain_result = validators.domain(value)
+        # We'll only raise the exception if strict mode is enabled - otherwise, we'll check hostname validation as well
+        if isinstance(domain_result, Exception) and self.strict:
+            raise HowlerValueError(
+                f"[{'.'.join(context) or self.name}] '{value}' did not pass validation."
+            ) from domain_result
+
+        hostname_result = validators.hostname(value)
+        if isinstance(hostname_result, Exception):
+            raise HowlerValueError(
+                f"[{'.'.join(context) or self.name}] '{value}' did not pass validation."
+            ) from hostname_result
+
+        return value.lower()
+
+
+class Email(Keyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.validation_regex = re.compile(EMAIL_REGEX)
+
+    def check(self, value, context=[], **kwargs):
+        if not value:
+            return None
+
+        validation_result = validators.email(value)
+        if isinstance(validation_result, Exception):
+            raise HowlerValueError(
+                f"[{'.'.join(context) or self.name}] '{value}' did not pass validation."
+            ) from validation_result
+
+        match = self.validation_regex.match(value)
+        if not is_valid_domain(match.group(1)):
+            raise HowlerValueError(
+                f"[{'.'.join(context) or self.name}] '{match.group(1)}' in email '{value}'" " is not a valid Domain."
+            )
+
+        return value.lower()
+
+
+class URI(Keyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.validation_regex = re.compile(FULL_URI)
+
+    def check(self, value, context=[], **kwargs):
+        if not value:
+            return None
+
+        match = self.validation_regex.match(value)
+        if not match:
+            raise HowlerValueError(
+                f"[{'.'.join(context) or self.name}] '{value}' not match the "
+                f"validator: {self.validation_regex.pattern}"
+            )
+
+        if not is_valid_domain(match.group(2)) and not is_valid_ip(match.group(2)):
+            raise HowlerValueError(
+                f"[{'.'.join(context) or self.name}] '{match.group(2)}' in URI '{value}'"
+                " is not a valid Domain or IP."
+            )
+
+        return match.group(0).replace(match.group(1), match.group(1).lower())
+
+
+class URIPath(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(URI_PATH, *args, **kwargs)
+
+
+class MAC(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(MAC_REGEX, *args, **kwargs)
+
+
+class PhoneNumber(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(PHONE_REGEX, *args, **kwargs)
+
+
+class SSDeepHash(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(SSDEEP_REGEX, *args, **kwargs)
+
+
+class SHA1(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(SHA1_REGEX, *args, **kwargs)
+
+
+class SHA256(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(SHA256_REGEX, *args, **kwargs)
+
+
+class HowlerHash(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(HOWLER_HASH_REGEX, *args, **kwargs)
+
+
+class MD5(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(MD5_REGEX, *args, **kwargs)
+
+
+class Platform(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(PLATFORM_REGEX, *args, **kwargs)
+
+
+class Processor(ValidatedKeyword):
+    def __init__(self, *args, **kwargs):
+        super().__init__(PROCESSOR_REGEX, *args, **kwargs)
+
+
+class Enum(Keyword):
+    """A field storing a short string that has predefined list of possible values"""
+
+    def __init__(self, values: PyEnum | list[typing.Any] | set[typing.Any], *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        if isinstance(values, set):
+            self.values = values
+        elif isinstance(values, (list, tuple)):
+            self.values = set(values)
+        elif isinstance(values, (PyEnum, EnumMeta)):
+            self.values = set([e.value for e in values])
+        else:
+            raise HowlerTypeError(f"Type unsupported for Enum odm: {type(values)}")
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if not value:
+            if self.default_set:
+                value = self.default
+            else:
+                raise HowlerValueError(f"[{'.'.join(context)}] Empty enums are not allow without defaults")
+
+        if value not in self.values:
+            raise HowlerValueError(f"[{'.'.join(context)}] {value} not in the possible values: {self.values}")
+
+        if value is None:
+            return value
+
+        return str(value)
+
+
+class UUID(Keyword):
+    """A field storing an auto-generated unique ID if None is provided"""
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.default_set = True
+
+    def check(self, value, **kwargs):
+        if value is None:
+            value = get_random_id()
+
+        return str(value)
+
+
+class Text(_Field):
+    """A field storing human readable text data."""
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if not value:
+            if self.default_set:
+                value = self.default
+            else:
+                raise HowlerValueError(f"[{'.'.join(context)}] Empty strings are not allowed without defaults")
+
+        if value is None:
+            return None
+
+        return str(value)
+
+
+class IndexText(_Field):
+    """A special field with special processing rules to simplify searching."""
+
+    def check(self, value, **kwargs):
+        if self.optional and value is None:
+            return None
+
+        return str(value)
+
+
+class Integer(_Field):
+    """A field storing an integer value."""
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if value is None or value == "":
+            if self.default_set:
+                return self.default
+
+        try:
+            return int(value)
+        except ValueError as e:
+            raise HowlerValueError(f"[{'.'.join(context)}]: {str(e)}")
+
+
+class Long(_Field):
+    """
+    A field storing a long value. Equivalent to Integer in python, but sets the ES datatype to long.
+
+    In Elasticsearch, Integer supports values from -2^31 to 2^31-1, while Long supports values from -2^63 to 2^63-1.
+    """
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if value is None or value == "":
+            if self.default_set:
+                return self.default
+
+        try:
+            return int(value)
+        except ValueError as e:
+            raise HowlerValueError(f"[{'.'.join(context)}]: {str(e)}")
+
+
+class Float(_Field):
+    """A field storing a floating point value."""
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if not value:
+            if self.default_set:
+                return self.default
+        try:
+            return float(value)
+        except ValueError as e:
+            raise HowlerValueError(f"[{'.'.join(context)}]: {str(e)}")
+
+
+class ClassificationObject(object):
+    def __init__(self, engine, value, is_uc=False):
+        self.engine = engine
+        self.is_uc = is_uc
+        self.value = engine.normalize_classification(value, skip_auto_select=is_uc)
+
+    def get_access_control_parts(self):
+        return self.engine.get_access_control_parts(self.value, user_classification=self.is_uc)
+
+    def min(self, other):
+        return ClassificationObject(
+            self.engine,
+            self.engine.min_classification(self.value, other.value),
+            is_uc=self.is_uc,
+        )
+
+    def max(self, other):
+        return ClassificationObject(
+            self.engine,
+            self.engine.max_classification(self.value, other.value),
+            is_uc=self.is_uc,
+        )
+
+    def intersect(self, other):
+        return ClassificationObject(
+            self.engine,
+            self.engine.intersect_user_classification(self.value, other.value),
+            is_uc=self.is_uc,
+        )
+
+    def long(self):
+        return self.engine.normalize_classification(self.value, skip_auto_select=self.is_uc)
+
+    def small(self):
+        return self.engine.normalize_classification(self.value, long_format=False, skip_auto_select=self.is_uc)
+
+    def __str__(self):
+        return self.value
+
+    def __eq__(self, other):
+        return self.value == other.value
+
+    def __ne__(self, other):
+        return self.value != other.value
+
+    def __le__(self, other):
+        return self.engine.is_accessible(other.value, self.value)
+
+    def __lt__(self, other):
+        return self.engine.is_accessible(other.value, self.value)
+
+    def __ge__(self, other):
+        return self.engine.is_accessible(self.value, other.value)
+
+    def __gt__(self, other):
+        return not self.engine.is_accessible(other.value, self.value)
+
+
+class Classification(Keyword):
+    """A field storing access control classification."""
+
+    def __init__(self, *args, is_user_classification=False, yml_config=None, **kwargs):
+        """An expanded classification is one that controls the access to the document
+        which holds it.
+        """
+        super().__init__(*args, **kwargs)
+        self.engine = loader.get_classification(yml_config=yml_config)
+        self.is_uc = is_user_classification
+
+    def check(self, value, **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if isinstance(value, ClassificationObject):
+            return ClassificationObject(self.engine, value.value, is_uc=self.is_uc)
+
+        return ClassificationObject(self.engine, value, is_uc=self.is_uc)
+
+
+class ClassificationString(Keyword):
+    """A field storing the classification as a string only."""
+
+    def __init__(self, *args, yml_config=None, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.engine = loader.get_classification(yml_config=yml_config)
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if not value:
+            if self.default_set:
+                value = self.default
+            else:
+                raise HowlerValueError(
+                    f"[{'.'.join(context) or self.name}]: Empty classification is not allowed without defaults"
+                )
+
+        if not self.engine.is_valid(value):
+            raise HowlerValueError(f"[{'.'.join(context) or self.name}]: Invalid classification: {value}")
+
+        return str(value)
+
+
+class TypedList(list):
+    def __init__(self, type_p, *items, context=[], **kwargs):
+        self.context = context
+        self.type = type_p
+
+        super().__init__([type_p.check(el, context=self.context, **kwargs) for el in items])
+
+    def append(self, item):
+        super().append(self.type.check(item, context=self.context))
+
+    def extend(self, sequence):
+        super().extend(self.type.check(item, context=self.context) for item in sequence)
+
+    def insert(self, index, item):
+        super().insert(index, self.type.check(item, context=self.context))
+
+    def __setitem__(self, index, item):
+        if isinstance(index, slice):
+            item = [self.type.check(val, context=self.context) for val in item]
+            super().__setitem__(index, item)
+        else:
+            super().__setitem__(index, self.type.check(item, context=self.context))
+
+
+class List(_Field):
+    """A field storing a sequence of typed elements."""
+
+    def __init__(self, child_type, **kwargs):
+        super().__init__(**kwargs)
+        self.child_type = child_type
+
+    def check(self, value, **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if isinstance(self.child_type, Compound) and isinstance(value, dict):
+            # Search queries of list of compound fields will return dotted paths of list of
+            # values. When processed through the flat_fields function, since this function
+            # has no idea about the data layout, it will transform the dotted paths into
+            # a dictionary of items then contains a list of object instead of a list
+            # of dictionaries with single items.
+
+            # The following piece of code transforms the dictionary of list into a list of
+            # dictionaries so the rest of the model validation can go through.
+
+            fixed_values = []
+            check_key = None
+            length = None
+            for key, val in flatten(value).items():
+                if not isinstance(val, list):
+                    val = [val]
+
+                if length is None:
+                    check_key = key
+                    length = len(val)
+
+                    for entry in val:
+                        fixed_values.append({key: entry})
+                elif len(val) != length:
+                    raise HowlerValueError(
+                        "Flattened fields creating list of ODMs must have equal length. Key "
+                        f"{key} has length {len(val)} compared to key {check_key} with length {length}."
+                    )
+                else:
+                    for i in range(len(val)):
+                        fixed_values[i][key] = val[i]
+
+            return TypedList(
+                self.child_type,
+                *fixed_values,
+                **kwargs,
+            )
+
+        if value is None:
+            logger.warning("Value is None, but optional is not set to True. Using an empty list to avoid errors.")
+            value = []
+
+        return TypedList(self.child_type, *value, **kwargs)
+
+    def apply_defaults(self, index, store):
+        """Initialize the default settings for the child field."""
+        # First apply the default to the list itself
+        super().apply_defaults(index, store)
+        # Then pass through the initialized values on the list to the child type
+        self.child_type.apply_defaults(self.index, self.store)
+
+    def fields(self):
+        out = dict()
+        for name, field_data in self.child_type.fields().items():
+            field_data = copy.deepcopy(field_data)
+            field_data.apply_defaults(self.index, self.store)
+            out[name] = field_data
+        return out
+
+
+class TypedMapping(dict):
+    def __init__(self, type_p, index, store, sanitizer, context=[], **items):
+        self.index = index
+        self.store = store
+        self.sanitizer = sanitizer
+        self.context = context
+
+        for key in items.keys():
+            if not self.sanitizer.match(key):
+                raise HowlerKeyError(f"[{'.'.join(self.context)}]: Illegal key {key}")
+
+        super().__init__({key: type_p.check(el, context=self.context) for key, el in items.items()})
+        self.type = type_p
+
+    def __setitem__(self, key, item):
+        if not self.sanitizer.match(key):
+            raise HowlerKeyError(f"[{'.'.join(self.context)}]: Illegal key: {key}")
+
+        return super().__setitem__(key, self.type.check(item, context=self.context))
+
+    def update(self, *args, **kwargs):
+        # Update supports three input layouts:
+        # 1. A single dictionary
+        if len(args) == 1 and isinstance(args[0], dict):
+            for key in args[0].keys():
+                if not self.sanitizer.match(key):
+                    raise HowlerKeyError(f"[{'.'.join(self.context)}]: Illegal key: {key}")
+
+            return super().update({key: self.type.check(item, context=self.context) for key, item in args[0].items()})
+
+        # 2. A list of key value pairs as if you were constructing a dictionary
+        elif args:
+            for key, _ in args:
+                if not self.sanitizer.match(key):
+                    raise HowlerKeyError(f"[{'.'.join(self.context)}]: Illegal key: {key}")
+
+            return super().update({key: self.type.check(item, context=self.context) for key, item in args})
+
+        # 3. Key values as arguments, can be combined with others
+        if kwargs:
+            for key in kwargs.keys():
+                if not self.sanitizer.match(key):
+                    raise HowlerKeyError(f"[{'.'.join(self.context)}]: Illegal key: {key}")
+
+            return super().update({key: self.type.check(item, context=self.context) for key, item in kwargs.items()})
+
+
+class Mapping(_Field):
+    """A field storing a sequence of typed elements."""
+
+    def __init__(self, child_type, **kwargs):
+        self.child_type = child_type
+        super().__init__(**kwargs)
+
+    def check(self, value, **kwargs):
+        if self.optional and value is None:
+            return None
+
+        if self.index or self.store:
+            sanitizer = FIELD_SANITIZER
+        else:
+            sanitizer = NOT_INDEXED_SANITIZER
+
+        return TypedMapping(self.child_type, self.index, self.store, sanitizer, **value)
+
+    def apply_defaults(self, index, store):
+        """Initialize the default settings for the child field."""
+        # First apply the default to the list itself
+        super().apply_defaults(index, store)
+        # Then pass through the initialized values on the list to the child type
+        self.child_type.apply_defaults(self.index, self.store)
+
+
+class FlattenedListObject(Mapping):
+    """A field storing a flattened object"""
+
+    def __init__(self, **kwargs):
+        super().__init__(List(Json()), **kwargs)
+
+    def check(self, value, **kwargs):
+        if self.optional and value is None:
+            return None
+
+        return TypedMapping(self.child_type, self.index, self.store, FLATTENED_OBJECT_SANITIZER, **value)
+
+    def apply_defaults(self, index, store):
+        """Initialize the default settings for the child field."""
+        # First apply the default to the list itself
+        super().apply_defaults(index, store)
+        # Then pass through the initialized values on the list to the child type
+        self.child_type.apply_defaults(self.index, self.store)
+
+
+class FlattenedObject(Mapping):
+    """A field storing a flattened object"""
+
+    def __init__(self, **kwargs):
+        super().__init__(Json(), **kwargs)
+
+    def check(self, value, context=[], **kwargs):
+        if self.optional and value is None:
+            return None
+
+        return TypedMapping(
+            self.child_type,
+            self.index,
+            self.store,
+            FLATTENED_OBJECT_SANITIZER,
+            context=context,
+            **value,
+        )
+
+    def apply_defaults(self, index, store):
+        """Initialize the default settings for the child field."""
+        # First apply the default to the list itself
+        super().apply_defaults(index, store)
+        # Then pass through the initialized values on the list to the child type
+        self.child_type.apply_defaults(self.index, self.store)
+
+
+class Compound(_Field):
+    def __init__(self, field_type, **kwargs):
+        super().__init__(**kwargs)
+        self.child_type = field_type
+
+    def check(
+        self,
+        value,
+        mask=None,
+        ignore_extra_values=False,
+        extra_fields={},
+        context=[],
+        **kwargs,
+    ):
+        if self.optional and value is None:
+            return None
+
+        if isinstance(value, self.child_type):
+            return value
+
+        return self.child_type(
+            value,
+            mask=mask,
+            ignore_extra_values=ignore_extra_values,
+            extra_fields=extra_fields,
+            context=context,
+        )
+
+    def fields(self):
+        out = dict()
+        for name, field_data in self.child_type.fields().items():
+            field_data = copy.deepcopy(field_data)
+            field_data.apply_defaults(self.index, self.store)
+            out[name] = field_data
+        return out
+
+
+class Optional(_Field):
+    """A wrapper field to allow simple types (int, float, bool) to take None values."""
+
+    def __init__(self, child_type: _Field, **kwargs):
+        if child_type.default_set:
+            kwargs["default"] = child_type.default
+        else:
+            child_type.default_set = True
+        super().__init__(**kwargs)
+        self.default_set = True
+        self.child_type = child_type
+        self.child_type.optional = True
+
+    def check(self, value, *args, **kwargs):
+        if value is None:
+            return None
+
+        return self.child_type.check(value, *args, **kwargs)
+
+    def fields(self):
+        return self.child_type.fields()
+
+    def apply_defaults(self, index, store):
+        super().apply_defaults(index, store)
+        self.child_type.apply_defaults(self.index, self.store)
+
+
+class Model:
+    @classmethod
+    def fields(cls, skip_mappings=False) -> dict[str, _Field]:
+        """Describe the elements of the model.
+
+        For compound fields return the field object.
+
+        Args:
+            skip_mappings (bool): Skip over mappings where the real subfield names are unknown.
+        """
+        if skip_mappings and hasattr(cls, "_odm_field_cache_skip"):
+            return cls._odm_field_cache_skip
+
+        if not skip_mappings and hasattr(cls, "_odm_field_cache"):
+            return cls._odm_field_cache
+
+        out = dict()
+        for name, field_data in cls.__dict__.items():
+            if isinstance(field_data, _Field):
+                if skip_mappings and isinstance(field_data, Mapping):
+                    continue
+                out[name.rstrip("_")] = field_data
+
+        if skip_mappings:
+            cls._odm_field_cache_skip = out
+        else:
+            cls._odm_field_cache = out
+        return out
+
+    @classmethod
+    def add_namespace(cls, namespace: str, field: _Field, index=None, store=None, description=None):
+        recursive_set_name(field, namespace)
+
+        if hasattr(cls, "_odm_field_cache_skip"):
+            cls._odm_field_cache_skip[namespace.rstrip("_")] = field
+
+        if hasattr(cls, "_odm_field_cache"):
+            cls._odm_field_cache[namespace.rstrip("_")] = field
+
+        setattr(cls, namespace, field)
+
+        field._Model__description = description
+        for name, field_data in field.fields().items():
+            if not FIELD_SANITIZER.match(name) or name in BANNED_FIELDS:
+                raise HowlerValueError(f"Illegal variable name: {name}")
+
+            recursive_set_name(field_data, name)
+            field_data.apply_defaults(index=index, store=store)
+
+    @classmethod
+    def remove_namespace(cls, namespace: str):
+        if hasattr(cls, "_odm_field_cache_skip"):
+            del cls._odm_field_cache_skip[namespace.rstrip("_")]
+
+        if hasattr(cls, "_odm_field_cache"):
+            del cls._odm_field_cache[namespace.rstrip("_")]
+
+        delattr(cls, namespace)
+
+    @staticmethod
+    def _recurse_fields(name, field, show_compound, skip_mappings, multivalued=False):
+        name = name.rstrip("_")
+        out = dict()
+        for sub_name, sub_field in field.fields().items():
+            sub_field.multivalued = multivalued or isinstance(field, List)
+
+            if skip_mappings and isinstance(sub_field, Mapping):
+                continue
+
+            elif isinstance(sub_field, (List, Optional, Compound)) and sub_name != "":
+                out.update(
+                    Model._recurse_fields(
+                        f"{name}.{sub_name}",
+                        sub_field.child_type,
+                        show_compound,
+                        skip_mappings,
+                        multivalued=multivalued or isinstance(sub_field, List),
+                    )
+                )
+
+            elif sub_name:
+                out[f"{name}.{sub_name}"] = sub_field
+
+            else:
+                out[name] = sub_field
+
+        if isinstance(field, Compound) and show_compound:
+            out[name] = field
+
+        return out
+
+    @classmethod
+    def flat_fields(cls, show_compound=False, skip_mappings=False) -> dict[str, _Field]:
+        """Describe the elements of the model.
+
+        Recurse into compound fields, concatenating the names with '.' separators.
+
+        Args:
+            show_compound (bool): Show compound as valid fields.
+            skip_mappings (bool): Skip over mappings where the real subfield names are unknown.
+        """
+        out = dict()
+        for name, field in cls.__dict__.items():
+            if isinstance(field, _Field):
+                if skip_mappings and isinstance(field, Mapping):
+                    continue
+                out.update(
+                    Model._recurse_fields(
+                        name,
+                        field,
+                        show_compound,
+                        skip_mappings,
+                        multivalued=isinstance(field, List),
+                    )
+                )
+        return out
+
+    @classmethod
+    def markdown(
+        cls,
+        toc_depth=1,
+        include_autogen_note=True,
+        defaults=None,
+        url_prefix="/howler/odm/class/",
+    ) -> Union[str, Dict]:
+        markdown_content = (
+            (
+                '??? success "Auto-Generated Documentation"\n    '
+                "This set of documentation is automatically generated from source, and will help ensure any change to "
+                "functionality will always be documented and available on release.\n\n"
+            )
+            if include_autogen_note
+            else ""
+        )
+
+        # Header
+        markdown_content += f"{'#'*toc_depth} {cls.__name__}\n\n> {cls.__description}\n\n"
+
+        # Table
+        table = "| Field | Type | Description | Required | Default |\n| :--- | :--- | :--- | :--- | :--- |\n"
+
+        # Determine the type of Field we're dealing with
+        # if possible return the Model class if wrapped in Compound
+        def get_type(field_class: _Field) -> Tuple[str, Model]:
+            if field_class.__class__ == Optional:
+                return get_type(field_class.child_type)
+            elif field_class.__class__ == Compound:
+                name = field_class.child_type.__name__
+
+                return (
+                    f"[{name}]({url_prefix}{name.lower()})",
+                    field_class.child_type,
+                )
+            elif field_class.__class__ in [Mapping, List]:
+                child_type, child_class = (
+                    field_class.child_type.__class__.__name__,
+                    field_class.child_type.__class__,
+                )
+                if field_class.child_type.__class__ == Compound:
+                    child_type, child_class = get_type(field_class.child_type)
+                return f"{field_class.__class__.__name__} [{child_type}]", child_class
+            elif field_class.__class__.__name__ == "type":
+                return field_class.__name__, None
+
+            return field_class.__class__.__name__, None
+
+        for field, info in cls.fields().items():
+            field_type, field_class = get_type(info)
+
+            # Field description
+            description = info.description
+            if description is None and info.__class__ == Optional:
+                description = info.child_type.description
+                if info.child_type.reference:
+                    description += f'<br><a href="{info.child_type.reference}">Reference Link</a><br>'
+            elif info.reference:
+                description += f'<br><a href="{info.reference}">Reference Link</a><br>'
+
+            # If field type is Enum, then show the possible values that can be used in the description
+            if field_type == "Enum":
+                values = info.child_type.values if info.__class__ != Enum else info.values
+                none_value = False
+                if None in values:
+                    none_value = True
+                    values.remove(None)
+
+                values = [f'"{v}"' if v else str(v) for v in sorted(values)]
+                values.append("None") if none_value else None
+                description = f'{description}<br>Values:<br>`{", ".join(values)}`'
+
+            # Is this a required field?
+            if info.__class__ != Optional and not info.optional:
+                required = ":material-checkbox-marked-outline: Yes"
+            else:
+                required = ":material-minus-box-outline: Optional"
+
+            if info.deprecated:
+                required += " :material-alert-box-outline: Deprecated - "
+                required += info.deprecated_description
+            elif info.__class__ == Optional and info.child_type.deprecated:
+                required += " :material-alert-box-outline: Deprecated - "
+                required += info.child_type.deprecated_description
+
+            # Determine the correct default values to display
+            default = f"`{info.default}`"
+            # If the field is a model, then provide a link to that documentation
+            if field_class and issubclass(field_class, Model) and isinstance(info.default, dict):
+                ref_link = field_type[field_type.index("(") : field_type.index(")") + 1]
+                default = f"See [{field_class.__name__}]{ref_link} for more details."
+
+            # Handle how to display values from provided defaults (different from field defaults)
+            elif isinstance(defaults, dict):
+                val = defaults.get(field, {})
+                default = f"`{val if not isinstance(val, dict) else info.default}`"
+            elif isinstance(defaults, list):
+                default = f"`{defaults}`"
+            row = f"| {field} | {field_type} | {description} | {required} | {default} |\n"
+            table += row
+
+        markdown_content += table + "\n\n"
+
+        return markdown_content
+
+    # Allow attribute assignment by default in the constructor until it is removed
+    __frozen = False
+    # Descriptions of the model should be class-accessible only for markdown()
+    __description = None
+
+    def __init__(
+        self,
+        data: dict = None,
+        mask: list = None,
+        docid=None,
+        ignore_extra_values=True,
+        extra_fields={},
+        context=[],
+    ):
+        if len(context) == 0:
+            context = [self.__class__.__name__.lower()]
+
+        if data is None:
+            data = {}
+
+        if not hasattr(data, "items"):
+            raise HowlerTypeError(f"'{self.__class__.__name__}' object must be constructed with dict like")
+        self._odm_py_obj = {}
+        self._id = docid
+        self.context = context
+
+        # Parse the field mask for sub models
+        mask_map = {}
+        if mask is not None:
+            for entry in mask:
+                if "." in entry:
+                    child, sub_key = entry.split(".", 1)
+                    try:
+                        mask_map[child].append(sub_key)
+                    except KeyError:
+                        mask_map[child] = [sub_key]
+                else:
+                    mask_map[entry] = None
+
+        # Get the list of fields we expect this object to have
+        fields = self.fields()
+        self._odm_removed = {}
+        if mask is not None:
+            self._odm_removed = {k: v for k, v in fields.items() if k not in mask_map}
+            fields = {k: v for k, v in fields.items() if k in mask_map}
+
+        # Trim out keys that actually belong to sub sections
+        data = flat_to_nested(data)
+
+        # Check to make sure we can use all the data we are given
+        self.unused_keys = set(data.keys()) - set(fields.keys()) - BANNED_FIELDS
+        extra_keys = set(extra_fields.keys()) - set(data.keys())
+        if self.unused_keys and not ignore_extra_values:
+            raise HowlerValueError(
+                f"[{'.'.join(context)}]: object was created with invalid parameters: " f"{', '.join(self.unused_keys)}"
+            )
+
+        # Pass each value through it's respective validator, and store it
+        for name, field_type in fields.items():
+            params = {"ignore_extra_values": ignore_extra_values}
+            if name in mask_map and mask_map[name]:
+                params["mask"] = mask_map[name]
+            if name in extra_fields and extra_fields[name]:
+                params["extra_fields"] = extra_fields[name]
+
+            try:
+                value = data[name]
+            except KeyError:
+                if field_type.default_set:
+                    value = copy.copy(field_type.default)
+                elif not field_type.optional:
+                    raise HowlerValueError(f"[{'.'.join([*context, name])}]: value is missing from the object!")
+                else:
+                    value = None
+
+            self._odm_py_obj[name.rstrip("_")] = field_type.check(value, context=[*context, name], **params)
+
+            value = None
+
+        for key in extra_keys:
+            self._odm_py_obj[key.rstrip("_")] = Any().check(extra_fields[key], context=[*context, name])
+
+        # Since the layout of model objects should be fixed, don't allow any further
+        # attribute assignment
+        self.__frozen = True
+
+    def as_primitives(self, hidden_fields=False, strip_null=True) -> dict[str, typing.Any]:
+        """Convert the object back into primitives that can be json serialized."""
+        out = {}
+
+        fields = self.fields()
+        for key, value in self._odm_py_obj.items():
+            field_type = fields.get(key, Any)
+            if value is not None or (value is None and field_type.default_set):
+                if strip_null and value is None:
+                    continue
+
+                if isinstance(value, Model):
+                    out[key] = value.as_primitives(strip_null=strip_null)
+                elif isinstance(value, datetime):
+                    out[key] = value.strftime(DATEFORMAT)
+                elif isinstance(value, TypedMapping):
+                    out[key] = {
+                        k: (v.as_primitives(strip_null=strip_null) if isinstance(v, Model) else v)
+                        for k, v in value.items()
+                    }
+                elif isinstance(value, TypedList):
+                    out[key] = [(v.as_primitives(strip_null=strip_null) if isinstance(v, Model) else v) for v in value]
+                elif isinstance(value, ClassificationObject):
+                    out[key] = str(value)
+                    if hidden_fields:
+                        out.update(value.get_access_control_parts())
+                else:
+                    out[key] = value
+        return out
+
+    def json(self):
+        return json.dumps(self.as_primitives())
+
+    def __eq__(self, other):
+        if isinstance(other, dict):
+            try:
+                other = self.__class__(other)
+            except (ValueError, KeyError):
+                return False
+
+        elif not isinstance(other, self.__class__):
+            return False
+
+        if len(self._odm_py_obj) != len(other._odm_py_obj):
+            return False
+
+        for name, field in self.fields().items():
+            if name in self._odm_removed:
+                continue
+            if field.__get__(self) != field.__get__(other):
+                return False
+
+        return True
+
+    def __repr__(self):
+        if self._id:
+            return f"<{self.__class__.__name__} [{self._id}] {self.json()}>"
+        return f"<{self.__class__.__name__} {self.json()}>"
+
+    def __getitem__(self, name):
+        data = self._odm_py_obj
+        for component in name.split("."):
+            data = data[component.rstrip("_")]
+
+        return data
+
+    def get(self, name, default=None):
+        try:
+            return self[name]
+        except KeyError:
+            return default
+
+    def __setitem__(self, name, value):
+        if name not in self._odm_field_cache:
+            raise HowlerKeyError(f"[{'.'.join(self.context)}]: {name}")
+        return self.__setattr__(name, value)
+
+    def __getattr__(self, name):
+        # Any attribute that hasn't been explicitly declared is forbidden
+        if name.rstrip("_") not in self.fields():
+            raise HowlerKeyError(f"[{'.'.join(self.context)}]: {name}")
+
+        return super().__getattr__(name)
+
+    def __setattr__(self, name, value):
+        # Any attribute that hasn't been explicitly declared is forbidden
+        if self.__frozen and name.rstrip("_") not in self.fields():
+            raise HowlerKeyError(f"[{'.'.join(self.context)}]: {name}")
+        return object.__setattr__(self, name, value)
+
+    def __contains__(self, name):
+        return name.rstrip("_") in self.fields()
+
+
+def recursive_set_name(field, name, to_parent=False):
+    if not to_parent:
+        field.name = name
+    else:
+        field.parent_name = name
+
+    if isinstance(field, Optional):
+        recursive_set_name(field.child_type, name)
+    if isinstance(field, List):
+        recursive_set_name(field.child_type, name, to_parent=True)
+
+
+def model(index=None, store=None, description=None):
+    """Decorator to create model objects."""
+
+    def _finish_model(cls):
+        cls._Model__description = description
+        for name, field_data in cls.fields().items():
+            if not FIELD_SANITIZER.match(name) or name in BANNED_FIELDS:
+                raise HowlerValueError(f"Illegal variable name: {name}")
+
+            recursive_set_name(field_data, name)
+            field_data.apply_defaults(index=index, store=store)
+        return cls
+
+    return _finish_model
+
+
+def _construct_field(field, value):
+    if isinstance(field, List):
+        clean, dropped = [], []
+        for item in value:
+            _c, _d = _construct_field(field.child_type, item)
+            if _c is not None:
+                clean.append(_c)
+            if _d is not None and _d != "":
+                dropped.append(_d)
+        return clean or None, dropped or None
+
+    elif isinstance(field, Compound):
+        _c, _d = construct_safe(field.child_type, value)
+        if len(_d) == 0:
+            _d = None
+        return _c, _d
+    elif isinstance(field, Optional):
+        return _construct_field(field.child_type, value)
+    else:
+        try:
+            return field.check(value), None
+        except (ValueError, TypeError):
+            return None, value
+
+
+def construct_safe(mod, data) -> tuple[_Any, dict]:
+    if not isinstance(data, dict):
+        return None, data
+    fields = mod.fields()
+    clean = {}
+    dropped = {}
+    for key, value in data.items():
+        if key not in fields:
+            dropped[key] = value
+            continue
+
+        _c, _d = _construct_field(fields[key], value)
+
+        if _c is not None:
+            clean[key] = _c
+        if _d is not None:
+            dropped[key] = _d
+
+    try:
+        return mod(clean), dropped
+    except ValueError:
+        return None, recursive_update(dropped, clean)