efro.dataclassio

29class Codec(Enum):
30    """Specifies expected data format exported to or imported from."""
31
32    # Use only types that will translate cleanly to/from json: lists,
33    # dicts with str keys, bools, ints, floats, and None.
34    JSON = 'json'
35
36    # Mostly like JSON but passes bytes and datetime objects through
37    # as-is instead of converting them to json-friendly types.
38    FIRESTORE = 'firestore'

 61class DataclassFieldLookup(Generic[T]):
 62    """Get info about nested dataclass fields in type-safe way."""
 63
 64    def __init__(self, cls: type[T]) -> None:
 65        self.cls = cls
 66
 67    def path(self, callback: Callable[[T], Any]) -> str:
 68        """Look up a path on child dataclass fields.
 69
 70        example:
 71          DataclassFieldLookup(MyType).path(lambda obj: obj.foo.bar)
 72
 73        The above example will return the string 'foo.bar' or something
 74        like 'f.b' if the dataclasses have custom storage names set.
 75        It will also be static-type-checked, triggering an error if
 76        MyType.foo.bar is not a valid path. Note, however, that the
 77        callback technically allows any return value but only nested
 78        dataclasses and their fields will succeed.
 79        """
 80
 81        # We tell the type system that we are returning an instance
 82        # of our class, which allows it to perform type checking on
 83        # member lookups. In reality, however, we are providing a
 84        # special object which captures path lookups, so we can build
 85        # a string from them.
 86        if not TYPE_CHECKING:
 87            out = callback(_PathCapture(self.cls))
 88            if not isinstance(out, _PathCapture):
 89                raise TypeError(
 90                    f'Expected a valid path under'
 91                    f' the provided object; got a {type(out)}.'
 92                )
 93            return out.path
 94        return ''
 95
 96    def paths(self, callback: Callable[[T], list[Any]]) -> list[str]:
 97        """Look up multiple paths on child dataclass fields.
 98
 99        Functionality is identical to path() but for multiple paths at once.
100
101        example:
102          DataclassFieldLookup(MyType).paths(lambda obj: [obj.foo, obj.bar])
103        """
104        outvals: list[str] = []
105        if not TYPE_CHECKING:
106            outs = callback(_PathCapture(self.cls))
107            assert isinstance(outs, list)
108            for out in outs:
109                if not isinstance(out, _PathCapture):
110                    raise TypeError(
111                        f'Expected a valid path under'
112                        f' the provided object; got a {type(out)}.'
113                    )
114                outvals.append(out.path)
115        return outvals

115class IOAttrs:
116    """For specifying io behavior in annotations.
117
118    'storagename', if passed, is the name used when storing to json/etc.
119    'store_default' can be set to False to avoid writing values when equal
120        to the default value. Note that this requires the dataclass field
121        to define a default or default_factory or for its IOAttrs to
122        define a soft_default value.
123    'whole_days', if True, requires datetime values to be exactly on day
124        boundaries (see efro.util.utc_today()).
125    'whole_hours', if True, requires datetime values to lie exactly on hour
126        boundaries (see efro.util.utc_this_hour()).
127    'whole_minutes', if True, requires datetime values to lie exactly on minute
128        boundaries (see efro.util.utc_this_minute()).
129    'soft_default', if passed, injects a default value into dataclass
130        instantiation when the field is not present in the input data.
131        This allows dataclasses to add new non-optional fields while
132        gracefully 'upgrading' old data. Note that when a soft_default is
133        present it will take precedence over field defaults when determining
134        whether to store a value for a field with store_default=False
135        (since the soft_default value is what we'll get when reading that
136        same data back in when the field is omitted).
137    'soft_default_factory' is similar to 'default_factory' in dataclass
138        fields; it should be used instead of 'soft_default' for mutable types
139        such as lists to prevent a single default object from unintentionally
140        changing over time.
141    """
142
143    # A sentinel object to detect if a parameter is supplied or not.  Use
144    # a class to give it a better repr.
145    class _MissingType:
146        pass
147
148    MISSING = _MissingType()
149
150    storagename: str | None = None
151    store_default: bool = True
152    whole_days: bool = False
153    whole_hours: bool = False
154    whole_minutes: bool = False
155    soft_default: Any = MISSING
156    soft_default_factory: Callable[[], Any] | _MissingType = MISSING
157
158    def __init__(
159        self,
160        storagename: str | None = storagename,
161        store_default: bool = store_default,
162        whole_days: bool = whole_days,
163        whole_hours: bool = whole_hours,
164        whole_minutes: bool = whole_minutes,
165        soft_default: Any = MISSING,
166        soft_default_factory: Callable[[], Any] | _MissingType = MISSING,
167    ):
168        # Only store values that differ from class defaults to keep
169        # our instances nice and lean.
170        cls = type(self)
171        if storagename != cls.storagename:
172            self.storagename = storagename
173        if store_default != cls.store_default:
174            self.store_default = store_default
175        if whole_days != cls.whole_days:
176            self.whole_days = whole_days
177        if whole_hours != cls.whole_hours:
178            self.whole_hours = whole_hours
179        if whole_minutes != cls.whole_minutes:
180            self.whole_minutes = whole_minutes
181        if soft_default is not cls.soft_default:
182            # Do what dataclasses does with its default types and
183            # tell the user to use factory for mutable ones.
184            if isinstance(soft_default, (list, dict, set)):
185                raise ValueError(
186                    f'mutable {type(soft_default)} is not allowed'
187                    f' for soft_default; use soft_default_factory.'
188                )
189            self.soft_default = soft_default
190        if soft_default_factory is not cls.soft_default_factory:
191            self.soft_default_factory = soft_default_factory
192            if self.soft_default is not cls.soft_default:
193                raise ValueError(
194                    'Cannot set both soft_default and soft_default_factory'
195                )
196
197    def validate_for_field(self, cls: type, field: dataclasses.Field) -> None:
198        """Ensure the IOAttrs instance is ok to use with the provided field."""
199
200        # Turning off store_default requires the field to have either
201        # a default or a default_factory or for us to have soft equivalents.
202
203        if not self.store_default:
204            field_default_factory: Any = field.default_factory
205            if (
206                field_default_factory is dataclasses.MISSING
207                and field.default is dataclasses.MISSING
208                and self.soft_default is self.MISSING
209                and self.soft_default_factory is self.MISSING
210            ):
211                raise TypeError(
212                    f'Field {field.name} of {cls} has'
213                    f' neither a default nor a default_factory'
214                    f' and IOAttrs contains neither a soft_default'
215                    f' nor a soft_default_factory;'
216                    f' store_default=False cannot be set for it.'
217                )
218
219    def validate_datetime(
220        self, value: datetime.datetime, fieldpath: str
221    ) -> None:
222        """Ensure a datetime value meets our value requirements."""
223        if self.whole_days:
224            if any(
225                x != 0
226                for x in (
227                    value.hour,
228                    value.minute,
229                    value.second,
230                    value.microsecond,
231                )
232            ):
233                raise ValueError(
234                    f'Value {value} at {fieldpath} is not a whole day.'
235                )
236        elif self.whole_hours:
237            if any(
238                x != 0 for x in (value.minute, value.second, value.microsecond)
239            ):
240                raise ValueError(
241                    f'Value {value} at {fieldpath}' f' is not a whole hour.'
242                )
243        elif self.whole_minutes:
244            if any(x != 0 for x in (value.second, value.microsecond)):
245                raise ValueError(
246                    f'Value {value} at {fieldpath}' f' is not a whole minute.'
247                )

41class IOExtendedData:
42    """A class that data types can inherit from for extra functionality."""
43
44    def will_output(self) -> None:
45        """Called before data is sent to an outputter.
46
47        Can be overridden to validate or filter data before
48        sending it on its way.
49        """
50
51    @classmethod
52    def will_input(cls, data: dict) -> None:
53        """Called on raw data before a class instance is created from it.
54
55        Can be overridden to migrate old data formats to new, etc.
56        """
57
58    def did_input(self) -> None:
59        """Called on a class instance after created from data.
60
61        Can be useful to correct values from the db, etc. in the
62        type-safe form.
63        """

 69class IOMultiType(Generic[EnumT]):
 70    """A base class for types that can map to multiple dataclass types.
 71
 72    This enables usage of high level base classes (for example
 73    a 'Message' type) in annotations, with dataclassio automatically
 74    serializing & deserializing dataclass subclasses based on their
 75    type ('MessagePing', 'MessageChat', etc.)
 76
 77    Standard usage involves creating a class which inherits from this
 78    one which acts as a 'registry', and then creating dataclass classes
 79    inheriting from that registry class. Dataclassio will then do the
 80    right thing when that registry class is used in type annotations.
 81
 82    See tests/test_efro/test_dataclassio.py for examples.
 83    """
 84
 85    @classmethod
 86    def get_type(cls, type_id: EnumT) -> type[Self]:
 87        """Return a specific subclass given a type-id."""
 88        raise NotImplementedError()
 89
 90    @classmethod
 91    def get_type_id(cls) -> EnumT:
 92        """Return the type-id for this subclass."""
 93        raise NotImplementedError()
 94
 95    @classmethod
 96    def get_type_id_type(cls) -> type[EnumT]:
 97        """Return the Enum type this class uses as its type-id."""
 98        out: type[EnumT] = cls.__orig_bases__[0].__args__[0]  # type: ignore
 99        assert issubclass(out, Enum)
100        return out
101
102    @classmethod
103    def get_type_id_storage_name(cls) -> str:
104        """Return the key used to store type id in serialized data.
105
106        The default is an obscure value so that it does not conflict
107        with members of individual type attrs, but in some cases one
108        might prefer to serialize it to something simpler like 'type'
109        by overriding this call. One just needs to make sure that no
110        encompassed types serialize anything to 'type' themself.
111        """
112        return '_dciotype'

27class JsonStyle(Enum):
28    """Different style types for json."""
29
30    # Single line, no spaces, no sorting. Not deterministic.
31    # Use this where speed is more important than determinism.
32    FAST = 'fast'
33
34    # Single line, no spaces, sorted keys. Deterministic.
35    # Use this when output may be hashed or compared for equality.
36    SORTED = 'sorted'
37
38    # Multiple lines, spaces, sorted keys. Deterministic.
39    # Use this for pretty human readable output.
40    PRETTY = 'pretty'

 95def dataclass_from_dict(
 96    cls: type[T],
 97    values: dict,
 98    codec: Codec = Codec.JSON,
 99    coerce_to_float: bool = True,
100    allow_unknown_attrs: bool = True,
101    discard_unknown_attrs: bool = False,
102) -> T:
103    """Given a dict, return a dataclass of a given type.
104
105    The dict must be formatted to match the specified codec (generally
106    json-friendly object types). This means that sequence values such as
107    tuples or sets should be passed as lists, enums should be passed as
108    their associated values, nested dataclasses should be passed as dicts,
109    etc.
110
111    All values are checked to ensure their types/values are valid.
112
113    Data for attributes of type Any will be checked to ensure they match
114    types supported directly by json. This does not include types such
115    as tuples which are implicitly translated by Python's json module
116    (as this would break the ability to do a lossless round-trip with
117    data).
118
119    If coerce_to_float is True, int values passed for float typed fields
120    will be converted to float values. Otherwise, a TypeError is raised.
121
122    If `allow_unknown_attrs` is False, AttributeErrors will be raised for
123    attributes present in the dict but not on the data class. Otherwise,
124    they will be preserved as part of the instance and included if it is
125    exported back to a dict, unless `discard_unknown_attrs` is True, in
126    which case they will simply be discarded.
127    """
128    val = _Inputter(
129        cls,
130        codec=codec,
131        coerce_to_float=coerce_to_float,
132        allow_unknown_attrs=allow_unknown_attrs,
133        discard_unknown_attrs=discard_unknown_attrs,
134    ).run(values)
135    assert isinstance(val, cls)
136    return val

139def dataclass_from_json(
140    cls: type[T],
141    json_str: str,
142    coerce_to_float: bool = True,
143    allow_unknown_attrs: bool = True,
144    discard_unknown_attrs: bool = False,
145) -> T:
146    """Utility function; return a dataclass instance given a json string.
147
148    Basically dataclass_from_dict(json.loads(...))
149    """
150    import json
151
152    return dataclass_from_dict(
153        cls=cls,
154        values=json.loads(json_str),
155        coerce_to_float=coerce_to_float,
156        allow_unknown_attrs=allow_unknown_attrs,
157        discard_unknown_attrs=discard_unknown_attrs,
158    )

43def dataclass_to_dict(
44    obj: Any,
45    codec: Codec = Codec.JSON,
46    coerce_to_float: bool = True,
47) -> dict:
48    """Given a dataclass object, return a json-friendly dict.
49
50    All values will be checked to ensure they match the types specified
51    on fields. Note that a limited set of types and data configurations is
52    supported.
53
54    Values with type Any will be checked to ensure they match types supported
55    directly by json. This does not include types such as tuples which are
56    implicitly translated by Python's json module (as this would break
57    the ability to do a lossless round-trip with data).
58
59    If coerce_to_float is True, integer values present on float typed fields
60    will be converted to float in the dict output. If False, a TypeError
61    will be triggered.
62    """
63
64    out = _Outputter(
65        obj, create=True, codec=codec, coerce_to_float=coerce_to_float
66    ).run()
67    assert isinstance(out, dict)
68    return out

71def dataclass_to_json(
72    obj: Any,
73    coerce_to_float: bool = True,
74    pretty: bool = False,
75    sort_keys: bool | None = None,
76) -> str:
77    """Utility function; return a json string from a dataclass instance.
78
79    Basically json.dumps(dataclass_to_dict(...)).
80    By default, keys are sorted for pretty output and not otherwise, but
81    this can be overridden by supplying a value for the 'sort_keys' arg.
82    """
83    import json
84
85    jdict = dataclass_to_dict(
86        obj=obj, coerce_to_float=coerce_to_float, codec=Codec.JSON
87    )
88    if sort_keys is None:
89        sort_keys = pretty
90    if pretty:
91        return json.dumps(jdict, indent=2, sort_keys=sort_keys)
92    return json.dumps(jdict, separators=(',', ':'), sort_keys=sort_keys)

161def dataclass_validate(
162    obj: Any, coerce_to_float: bool = True, codec: Codec = Codec.JSON
163) -> None:
164    """Ensure that values in a dataclass instance are the correct types."""
165
166    # Simply run an output pass but tell it not to generate data;
167    # only run validation.
168    _Outputter(
169        obj, create=False, codec=codec, coerce_to_float=coerce_to_float
170    ).run()

47def ioprep(cls: type, globalns: dict | None = None) -> None:
48    """Prep a dataclass type for use with this module's functionality.
49
50    Prepping ensures that all types contained in a data class as well as
51    the usage of said types are supported by this module and pre-builds
52    necessary constructs needed for encoding/decoding/etc.
53
54    Prepping will happen on-the-fly as needed, but a warning will be
55    emitted in such cases, as it is better to explicitly prep all used types
56    early in a process to ensure any invalid types or configuration are caught
57    immediately.
58
59    Prepping a dataclass involves evaluating its type annotations, which,
60    as of PEP 563, are stored simply as strings. This evaluation is done
61    with localns set to the class dict (so that types defined in the class
62    can be used) and globalns set to the containing module's class.
63    It is possible to override globalns for special cases such as when
64    prepping happens as part of an execed string instead of within a
65    module.
66    """
67    PrepSession(explicit=True, globalns=globalns).prep_dataclass(
68        cls, recursion_level=0
69    )

72def ioprepped(cls: type[T]) -> type[T]:
73    """Class decorator for easily prepping a dataclass at definition time.
74
75    Note that in some cases it may not be possible to prep a dataclass
76    immediately (such as when its type annotations refer to forward-declared
77    types). In these cases, dataclass_prep() should be explicitly called for
78    the class as soon as possible; ideally at module import time to expose any
79    errors as early as possible in execution.
80    """
81    ioprep(cls)
82    return cls

102def is_ioprepped_dataclass(obj: Any) -> bool:
103    """Return whether the obj is an ioprepped dataclass type or instance."""
104    cls = obj if isinstance(obj, type) else type(obj)
105    return dataclasses.is_dataclass(cls) and hasattr(cls, PREP_ATTR)

85def will_ioprep(cls: type[T]) -> type[T]:
86    """Class decorator hinting that we will prep a class later.
87
88    In some cases (such as recursive types) we cannot use the @ioprepped
89    decorator and must instead call ioprep() explicitly later. However,
90    some of our custom pylint checking behaves differently when the
91    @ioprepped decorator is present, in that case requiring type annotations
92    to be present and not simply forward declared under an "if TYPE_CHECKING"
93    block. (since they are used at runtime).
94
95    The @will_ioprep decorator triggers the same pylint behavior
96    differences as @ioprepped (which are necessary for the later ioprep() call
97    to work correctly) but without actually running any prep itself.
98    """
99    return cls