"""Utilities for union (sum type) disambiguation."""
from __future__ import annotations
from collections import defaultdict
from dataclasses import MISSING
from functools import reduce
from operator import or_
from typing import TYPE_CHECKING, Any, Callable, Literal, Mapping, Union
from attrs import NOTHING, Attribute, AttrsInstance
from ._compat import (
NoneType,
adapted_fields,
fields_dict,
get_args,
get_origin,
has,
is_literal,
is_union_type,
)
from .gen import AttributeOverride
if TYPE_CHECKING:
from .converters import BaseConverter
__all__ = ["is_supported_union", "create_default_dis_func"]
[docs]def is_supported_union(typ: Any) -> bool:
"""Whether the type is a union of attrs classes."""
return is_union_type(typ) and all(
e is NoneType or has(get_origin(e) or e) for e in typ.__args__
)
[docs]def create_default_dis_func(
converter: BaseConverter,
*classes: type[AttrsInstance],
use_literals: bool = True,
overrides: (
dict[str, AttributeOverride] | Literal["from_converter"]
) = "from_converter",
) -> Callable[[Mapping[Any, Any]], type[Any] | None]:
"""Given attrs classes or dataclasses, generate a disambiguation function.
The function is based on unique fields without defaults or unique values.
:param use_literals: Whether to try using fields annotated as literals for
disambiguation.
:param overrides: Attribute overrides to apply.
.. versionchanged:: 24.1.0
Dataclasses are now supported.
"""
if len(classes) < 2:
raise ValueError("At least two classes required.")
if overrides == "from_converter":
overrides = [
getattr(converter.get_structure_hook(c), "overrides", {}) for c in classes
]
else:
overrides = [overrides for _ in classes]
# first, attempt for unique values
if use_literals:
# requirements for a discriminator field:
# (... TODO: a single fallback is OK)
# - it must always be enumerated
cls_candidates = [
{
at.name
for at in adapted_fields(get_origin(cl) or cl)
if is_literal(at.type)
}
for cl in classes
]
# literal field names common to all members
discriminators: set[str] = cls_candidates[0]
for possible_discriminators in cls_candidates:
discriminators &= possible_discriminators
best_result = None
best_discriminator = None
for discriminator in discriminators:
# maps Literal values (strings, ints...) to classes
mapping = defaultdict(list)
for cl in classes:
for key in get_args(
fields_dict(get_origin(cl) or cl)[discriminator].type
):
mapping[key].append(cl)
if best_result is None or max(len(v) for v in mapping.values()) <= max(
len(v) for v in best_result.values()
):
best_result = mapping
best_discriminator = discriminator
if (
best_result
and best_discriminator
and max(len(v) for v in best_result.values()) != len(classes)
):
final_mapping = {
k: v[0] if len(v) == 1 else Union[tuple(v)]
for k, v in best_result.items()
}
def dis_func(data: Mapping[Any, Any]) -> type | None:
if not isinstance(data, Mapping):
raise ValueError("Only input mappings are supported.")
return final_mapping[data[best_discriminator]]
return dis_func
# next, attempt for unique keys
# NOTE: This could just as well work with just field availability and not
# uniqueness, returning Unions ... it doesn't do that right now.
cls_and_attrs = [
(cl, *_usable_attribute_names(cl, override))
for cl, override in zip(classes, overrides)
]
# For each class, attempt to generate a single unique required field.
uniq_attrs_dict: dict[str, type] = {}
# We start from classes with the largest number of unique fields
# so we can do easy picks first, making later picks easier.
cls_and_attrs.sort(key=lambda c_a: len(c_a[1]), reverse=True)
fallback = None # If none match, try this.
for cl, cl_reqs, back_map in cls_and_attrs:
# We do not have to consider classes we've already processed, since
# they will have been eliminated by the match dictionary already.
other_classes = [
c_and_a
for c_and_a in cls_and_attrs
if c_and_a[0] is not cl and c_and_a[0] not in uniq_attrs_dict.values()
]
other_reqs = reduce(or_, (c_a[1] for c_a in other_classes), set())
uniq = cl_reqs - other_reqs
# We want a unique attribute with no default.
cl_fields = fields_dict(get_origin(cl) or cl)
for maybe_renamed_attr_name in uniq:
orig_name = back_map[maybe_renamed_attr_name]
if cl_fields[orig_name].default in (NOTHING, MISSING):
break
else:
if fallback is None:
fallback = cl
continue
raise TypeError(f"{cl} has no usable non-default attributes")
uniq_attrs_dict[maybe_renamed_attr_name] = cl
if fallback is None:
def dis_func(data: Mapping[Any, Any]) -> type[AttrsInstance] | None:
if not isinstance(data, Mapping):
raise ValueError("Only input mappings are supported")
for k, v in uniq_attrs_dict.items():
if k in data:
return v
raise ValueError("Couldn't disambiguate")
else:
def dis_func(data: Mapping[Any, Any]) -> type[AttrsInstance] | None:
if not isinstance(data, Mapping):
raise ValueError("Only input mappings are supported")
for k, v in uniq_attrs_dict.items():
if k in data:
return v
return fallback
return dis_func
create_uniq_field_dis_func = create_default_dis_func
def _overriden_name(at: Attribute, override: AttributeOverride | None) -> str:
if override is None or override.rename is None:
return at.name
return override.rename
def _usable_attribute_names(
cl: type[Any], overrides: dict[str, AttributeOverride]
) -> tuple[set[str], dict[str, str]]:
"""Return renamed fields and a mapping to original field names."""
res = set()
mapping = {}
for at in adapted_fields(get_origin(cl) or cl):
res.add(n := _overriden_name(at, overrides.get(at.name)))
mapping[n] = at.name
return res, mapping