Reflow dataclasses comments (GH-6893)

To be more consistent with other code (and so people stop hassling me!), reflow the dataclasses comments to not use a single space indentation when continuing a paragraph of text.
This commit is contained in:
Eric V. Smith 2018-05-16 05:14:53 -04:00 committed by GitHub
parent 98d50cb8f5
commit f8e7549490
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 176 additions and 170 deletions

View File

@ -21,12 +21,12 @@ __all__ = ['dataclass',
] ]
# Conditions for adding methods. The boxes indicate what action the # Conditions for adding methods. The boxes indicate what action the
# dataclass decorator takes. For all of these tables, when I talk # dataclass decorator takes. For all of these tables, when I talk
# about init=, repr=, eq=, order=, unsafe_hash=, or frozen=, I'm # about init=, repr=, eq=, order=, unsafe_hash=, or frozen=, I'm
# referring to the arguments to the @dataclass decorator. When # referring to the arguments to the @dataclass decorator. When
# checking if a dunder method already exists, I mean check for an # checking if a dunder method already exists, I mean check for an
# entry in the class's __dict__. I never check to see if an # entry in the class's __dict__. I never check to see if an attribute
# attribute is defined in a base class. # is defined in a base class.
# Key: # Key:
# +=========+=========================================+ # +=========+=========================================+
@ -79,7 +79,7 @@ __all__ = ['dataclass',
# | True | add | raise | # | True | add | raise |
# +=======+=======+=======+ # +=======+=======+=======+
# Raise because not adding these methods would break the "frozen-ness" # Raise because not adding these methods would break the "frozen-ness"
# of the class. # of the class.
# __eq__ # __eq__
# #
@ -108,7 +108,7 @@ __all__ = ['dataclass',
# | True | add | raise | # | True | add | raise |
# +=======+=======+=======+ # +=======+=======+=======+
# Raise because to allow this case would interfere with using # Raise because to allow this case would interfere with using
# functools.total_ordering. # functools.total_ordering.
# __hash__ # __hash__
@ -136,34 +136,35 @@ __all__ = ['dataclass',
# | True | True | True | add | raise | Frozen, so hashable # | True | True | True | add | raise | Frozen, so hashable
# +=======+=======+=======+========+========+ # +=======+=======+=======+========+========+
# For boxes that are blank, __hash__ is untouched and therefore # For boxes that are blank, __hash__ is untouched and therefore
# inherited from the base class. If the base is object, then # inherited from the base class. If the base is object, then
# id-based hashing is used. # id-based hashing is used.
#
# Note that a class may already have __hash__=None if it specified an # Note that a class may already have __hash__=None if it specified an
# __eq__ method in the class body (not one that was created by # __eq__ method in the class body (not one that was created by
# @dataclass). # @dataclass).
#
# See _hash_action (below) for a coded version of this table. # See _hash_action (below) for a coded version of this table.
# Raised when an attempt is made to modify a frozen class. # Raised when an attempt is made to modify a frozen class.
class FrozenInstanceError(AttributeError): pass class FrozenInstanceError(AttributeError): pass
# A sentinel object for default values to signal that a # A sentinel object for default values to signal that a default
# default factory will be used. # factory will be used. This is given a nice repr() which will appear
# This is given a nice repr() which will appear in the function # in the function signature of dataclasses' constructors.
# signature of dataclasses' constructors.
class _HAS_DEFAULT_FACTORY_CLASS: class _HAS_DEFAULT_FACTORY_CLASS:
def __repr__(self): def __repr__(self):
return '<factory>' return '<factory>'
_HAS_DEFAULT_FACTORY = _HAS_DEFAULT_FACTORY_CLASS() _HAS_DEFAULT_FACTORY = _HAS_DEFAULT_FACTORY_CLASS()
# A sentinel object to detect if a parameter is supplied or not. Use # A sentinel object to detect if a parameter is supplied or not. Use
# a class to give it a better repr. # a class to give it a better repr.
class _MISSING_TYPE: class _MISSING_TYPE:
pass pass
MISSING = _MISSING_TYPE() MISSING = _MISSING_TYPE()
# Since most per-field metadata will be unused, create an empty # Since most per-field metadata will be unused, create an empty
# read-only proxy that can be shared among all fields. # read-only proxy that can be shared among all fields.
_EMPTY_METADATA = types.MappingProxyType({}) _EMPTY_METADATA = types.MappingProxyType({})
# Markers for the various kinds of fields and pseudo-fields. # Markers for the various kinds of fields and pseudo-fields.
@ -177,7 +178,7 @@ _FIELD_CLASSVAR = _FIELD_BASE('_FIELD_CLASSVAR')
_FIELD_INITVAR = _FIELD_BASE('_FIELD_INITVAR') _FIELD_INITVAR = _FIELD_BASE('_FIELD_INITVAR')
# The name of an attribute on the class where we store the Field # The name of an attribute on the class where we store the Field
# objects. Also used to check if a class is a Data Class. # objects. Also used to check if a class is a Data Class.
_FIELDS = '__dataclass_fields__' _FIELDS = '__dataclass_fields__'
# The name of an attribute on the class that stores the parameters to # The name of an attribute on the class that stores the parameters to
@ -202,13 +203,15 @@ class InitVar(metaclass=_InitVarMeta):
# Instances of Field are only ever created from within this module, # Instances of Field are only ever created from within this module,
# and only from the field() function, although Field instances are # and only from the field() function, although Field instances are
# exposed externally as (conceptually) read-only objects. # exposed externally as (conceptually) read-only objects.
# name and type are filled in after the fact, not in __init__. They're #
# not known at the time this class is instantiated, but it's # name and type are filled in after the fact, not in __init__.
# convenient if they're available later. # They're not known at the time this class is instantiated, but it's
# convenient if they're available later.
#
# When cls._FIELDS is filled in with a list of Field objects, the name # When cls._FIELDS is filled in with a list of Field objects, the name
# and type fields will have been populated. # and type fields will have been populated.
class Field: class Field:
__slots__ = ('name', __slots__ = ('name',
'type', 'type',
@ -252,17 +255,18 @@ class Field:
')') ')')
# This is used to support the PEP 487 __set_name__ protocol in the # This is used to support the PEP 487 __set_name__ protocol in the
# case where we're using a field that contains a descriptor as a # case where we're using a field that contains a descriptor as a
# defaul value. For details on __set_name__, see # defaul value. For details on __set_name__, see
# https://www.python.org/dev/peps/pep-0487/#implementation-details. # https://www.python.org/dev/peps/pep-0487/#implementation-details.
# Note that in _process_class, this Field object is overwritten with #
# the default value, so the end result is a descriptor that had # Note that in _process_class, this Field object is overwritten
# __set_name__ called on it at the right time. # with the default value, so the end result is a descriptor that
# had __set_name__ called on it at the right time.
def __set_name__(self, owner, name): def __set_name__(self, owner, name):
func = getattr(type(self.default), '__set_name__', None) func = getattr(type(self.default), '__set_name__', None)
if func: if func:
# There is a __set_name__ method on the descriptor, # There is a __set_name__ method on the descriptor, call
# call it. # it.
func(self.default, owner, name) func(self.default, owner, name)
@ -295,20 +299,20 @@ class _DataclassParams:
# This function is used instead of exposing Field creation directly, # This function is used instead of exposing Field creation directly,
# so that a type checker can be told (via overloads) that this is a # so that a type checker can be told (via overloads) that this is a
# function whose type depends on its parameters. # function whose type depends on its parameters.
def field(*, default=MISSING, default_factory=MISSING, init=True, repr=True, def field(*, default=MISSING, default_factory=MISSING, init=True, repr=True,
hash=None, compare=True, metadata=None): hash=None, compare=True, metadata=None):
"""Return an object to identify dataclass fields. """Return an object to identify dataclass fields.
default is the default value of the field. default_factory is a default is the default value of the field. default_factory is a
0-argument function called to initialize a field's value. If init 0-argument function called to initialize a field's value. If init
is True, the field will be a parameter to the class's __init__() is True, the field will be a parameter to the class's __init__()
function. If repr is True, the field will be included in the function. If repr is True, the field will be included in the
object's repr(). If hash is True, the field will be included in object's repr(). If hash is True, the field will be included in
the object's hash(). If compare is True, the field will be used in the object's hash(). If compare is True, the field will be used
comparison functions. metadata, if specified, must be a mapping in comparison functions. metadata, if specified, must be a
which is stored but not otherwise examined by dataclass. mapping which is stored but not otherwise examined by dataclass.
It is an error to specify both default and default_factory. It is an error to specify both default and default_factory.
""" """
@ -321,8 +325,8 @@ def field(*, default=MISSING, default_factory=MISSING, init=True, repr=True,
def _tuple_str(obj_name, fields): def _tuple_str(obj_name, fields):
# Return a string representing each field of obj_name as a tuple # Return a string representing each field of obj_name as a tuple
# member. So, if fields is ['x', 'y'] and obj_name is "self", # member. So, if fields is ['x', 'y'] and obj_name is "self",
# return "(self.x,self.y)". # return "(self.x,self.y)".
# Special case for the 0-tuple. # Special case for the 0-tuple.
if not fields: if not fields:
@ -333,9 +337,9 @@ def _tuple_str(obj_name, fields):
def _create_fn(name, args, body, *, globals=None, locals=None, def _create_fn(name, args, body, *, globals=None, locals=None,
return_type=MISSING): return_type=MISSING):
# Note that we mutate locals when exec() is called. Caller beware! # Note that we mutate locals when exec() is called. Caller
# The only callers are internal to this module, so no worries # beware! The only callers are internal to this module, so no
# about external callers. # worries about external callers.
if locals is None: if locals is None:
locals = {} locals = {}
return_annotation = '' return_annotation = ''
@ -354,10 +358,11 @@ def _create_fn(name, args, body, *, globals=None, locals=None,
def _field_assign(frozen, name, value, self_name): def _field_assign(frozen, name, value, self_name):
# If we're a frozen class, then assign to our fields in __init__ # If we're a frozen class, then assign to our fields in __init__
# via object.__setattr__. Otherwise, just use a simple # via object.__setattr__. Otherwise, just use a simple
# assignment. # assignment.
#
# self_name is what "self" is called in this function: don't # self_name is what "self" is called in this function: don't
# hard-code "self", since that might be a field name. # hard-code "self", since that might be a field name.
if frozen: if frozen:
return f'object.__setattr__({self_name},{name!r},{value})' return f'object.__setattr__({self_name},{name!r},{value})'
return f'{self_name}.{name}={value}' return f'{self_name}.{name}={value}'
@ -365,31 +370,31 @@ def _field_assign(frozen, name, value, self_name):
def _field_init(f, frozen, globals, self_name): def _field_init(f, frozen, globals, self_name):
# Return the text of the line in the body of __init__ that will # Return the text of the line in the body of __init__ that will
# initialize this field. # initialize this field.
default_name = f'_dflt_{f.name}' default_name = f'_dflt_{f.name}'
if f.default_factory is not MISSING: if f.default_factory is not MISSING:
if f.init: if f.init:
# This field has a default factory. If a parameter is # This field has a default factory. If a parameter is
# given, use it. If not, call the factory. # given, use it. If not, call the factory.
globals[default_name] = f.default_factory globals[default_name] = f.default_factory
value = (f'{default_name}() ' value = (f'{default_name}() '
f'if {f.name} is _HAS_DEFAULT_FACTORY ' f'if {f.name} is _HAS_DEFAULT_FACTORY '
f'else {f.name}') f'else {f.name}')
else: else:
# This is a field that's not in the __init__ params, but # This is a field that's not in the __init__ params, but
# has a default factory function. It needs to be # has a default factory function. It needs to be
# initialized here by calling the factory function, # initialized here by calling the factory function,
# because there's no other way to initialize it. # because there's no other way to initialize it.
# For a field initialized with a default=defaultvalue, the # For a field initialized with a default=defaultvalue, the
# class dict just has the default value # class dict just has the default value
# (cls.fieldname=defaultvalue). But that won't work for a # (cls.fieldname=defaultvalue). But that won't work for a
# default factory, the factory must be called in __init__ # default factory, the factory must be called in __init__
# and we must assign that to self.fieldname. We can't # and we must assign that to self.fieldname. We can't
# fall back to the class dict's value, both because it's # fall back to the class dict's value, both because it's
# not set, and because it might be different per-class # not set, and because it might be different per-class
# (which, after all, is why we have a factory function!). # (which, after all, is why we have a factory function!).
globals[default_name] = f.default_factory globals[default_name] = f.default_factory
value = f'{default_name}()' value = f'{default_name}()'
@ -403,13 +408,13 @@ def _field_init(f, frozen, globals, self_name):
globals[default_name] = f.default globals[default_name] = f.default
value = f.name value = f.name
else: else:
# This field does not need initialization. Signify that to # This field does not need initialization. Signify that
# the caller by returning None. # to the caller by returning None.
return None return None
# Only test this now, so that we can create variables for the # Only test this now, so that we can create variables for the
# default. However, return None to signify that we're not going # default. However, return None to signify that we're not going
# to actually do the assignment statement for InitVars. # to actually do the assignment statement for InitVars.
if f._field_type == _FIELD_INITVAR: if f._field_type == _FIELD_INITVAR:
return None return None
@ -418,19 +423,20 @@ def _field_init(f, frozen, globals, self_name):
def _init_param(f): def _init_param(f):
# Return the __init__ parameter string for this field. # Return the __init__ parameter string for this field. For
# For example, the equivalent of 'x:int=3' (except instead of 'int', # example, the equivalent of 'x:int=3' (except instead of 'int',
# reference a variable set to int, and instead of '3', reference a # reference a variable set to int, and instead of '3', reference a
# variable set to 3). # variable set to 3).
if f.default is MISSING and f.default_factory is MISSING: if f.default is MISSING and f.default_factory is MISSING:
# There's no default, and no default_factory, just # There's no default, and no default_factory, just output the
# output the variable name and type. # variable name and type.
default = '' default = ''
elif f.default is not MISSING: elif f.default is not MISSING:
# There's a default, this will be the name that's used to look it up. # There's a default, this will be the name that's used to look
# it up.
default = f'=_dflt_{f.name}' default = f'=_dflt_{f.name}'
elif f.default_factory is not MISSING: elif f.default_factory is not MISSING:
# There's a factory function. Set a marker. # There's a factory function. Set a marker.
default = '=_HAS_DEFAULT_FACTORY' default = '=_HAS_DEFAULT_FACTORY'
return f'{f.name}:_type_{f.name}{default}' return f'{f.name}:_type_{f.name}{default}'
@ -439,10 +445,10 @@ def _init_fn(fields, frozen, has_post_init, self_name):
# fields contains both real fields and InitVar pseudo-fields. # fields contains both real fields and InitVar pseudo-fields.
# Make sure we don't have fields without defaults following fields # Make sure we don't have fields without defaults following fields
# with defaults. This actually would be caught when exec-ing the # with defaults. This actually would be caught when exec-ing the
# function source code, but catching it here gives a better error # function source code, but catching it here gives a better error
# message, and future-proofs us in case we build up the function # message, and future-proofs us in case we build up the function
# using ast. # using ast.
seen_default = False seen_default = False
for f in fields: for f in fields:
# Only consider fields in the __init__ call. # Only consider fields in the __init__ call.
@ -460,7 +466,7 @@ def _init_fn(fields, frozen, has_post_init, self_name):
for f in fields: for f in fields:
line = _field_init(f, frozen, globals, self_name) line = _field_init(f, frozen, globals, self_name)
# line is None means that this field doesn't require # line is None means that this field doesn't require
# initialization (it's a pseudo-field). Just skip it. # initialization (it's a pseudo-field). Just skip it.
if line: if line:
body_lines.append(line) body_lines.append(line)
@ -493,8 +499,8 @@ def _repr_fn(fields):
def _frozen_get_del_attr(cls, fields): def _frozen_get_del_attr(cls, fields):
# XXX: globals is modified on the first call to _create_fn, then the # XXX: globals is modified on the first call to _create_fn, then
# modified version is used in the second call. Is this okay? # the modified version is used in the second call. Is this okay?
globals = {'cls': cls, globals = {'cls': cls,
'FrozenInstanceError': FrozenInstanceError} 'FrozenInstanceError': FrozenInstanceError}
if fields: if fields:
@ -519,9 +525,9 @@ def _frozen_get_del_attr(cls, fields):
def _cmp_fn(name, op, self_tuple, other_tuple): def _cmp_fn(name, op, self_tuple, other_tuple):
# Create a comparison function. If the fields in the object are # Create a comparison function. If the fields in the object are
# named 'x' and 'y', then self_tuple is the string # named 'x' and 'y', then self_tuple is the string
# '(self.x,self.y)' and other_tuple is the string # '(self.x,self.y)' and other_tuple is the string
# '(other.x,other.y)'. # '(other.x,other.y)'.
return _create_fn(name, return _create_fn(name,
('self', 'other'), ('self', 'other'),
@ -540,7 +546,7 @@ def _hash_fn(fields):
def _is_classvar(a_type, typing): def _is_classvar(a_type, typing):
if typing: if typing:
# This test uses a typing internal class, but it's the best # This test uses a typing internal class, but it's the best
# way to test if this is a ClassVar. # way to test if this is a ClassVar.
return (a_type is typing.ClassVar return (a_type is typing.ClassVar
or (type(a_type) is typing._GenericAlias or (type(a_type) is typing._GenericAlias
and a_type.__origin__ is typing.ClassVar)) and a_type.__origin__ is typing.ClassVar))
@ -573,25 +579,25 @@ def _is_type(annotation, cls, a_module, a_type, is_type_predicate):
# instead only a module (global) lookup), there are some things it # instead only a module (global) lookup), there are some things it
# gets wrong. # gets wrong.
# With string annotations, this will work: # With string annotations, cv0 will be detected as a ClassVar:
# CV = ClassVar # CV = ClassVar
# @dataclass # @dataclass
# class C0: # class C0:
# cv0: CV # cv0: CV
# But this will not: # But in this example cv1 will not be detected as a ClassVar:
# @dataclass # @dataclass
# class C1: # class C1:
# CV = ClassVar # CV = ClassVar
# cv1: CV # cv1: CV
# In C1, the code in this function will look up "CV" in the module # In C1, the code in this function (_is_type) will look up "CV" in
# and not find it, so it will not consider cv1 as a ClassVar. # the module and not find it, so it will not consider cv1 as a
# This is a fairly obscure corner case, and the best way to fix it # ClassVar. This is a fairly obscure corner case, and the best
# would be to eval() the string "CV" with the correct global and # way to fix it would be to eval() the string "CV" with the
# local namespaces. However that would involve a eval() penalty # correct global and local namespaces. However that would involve
# for every single field of every dataclass that's defined. It # a eval() penalty for every single field of every dataclass
# was judged not worth it. # that's defined. It was judged not worth it.
match = _MODULE_IDENTIFIER_RE.match(annotation) match = _MODULE_IDENTIFIER_RE.match(annotation)
if match: if match:
@ -613,11 +619,11 @@ def _is_type(annotation, cls, a_module, a_type, is_type_predicate):
def _get_field(cls, a_name, a_type): def _get_field(cls, a_name, a_type):
# Return a Field object for this field name and type. ClassVars # Return a Field object for this field name and type. ClassVars
# and InitVars are also returned, but marked as such (see # and InitVars are also returned, but marked as such (see
# f._field_type). # f._field_type).
# If the default value isn't derived from Field, then it's # If the default value isn't derived from Field, then it's only a
# only a normal default value. Convert it to a Field(). # normal default value. Convert it to a Field().
default = getattr(cls, a_name, MISSING) default = getattr(cls, a_name, MISSING)
if isinstance(default, Field): if isinstance(default, Field):
f = default f = default
@ -627,33 +633,31 @@ def _get_field(cls, a_name, a_type):
default = MISSING default = MISSING
f = field(default=default) f = field(default=default)
# Only at this point do we know the name and the type. Set them. # Only at this point do we know the name and the type. Set them.
f.name = a_name f.name = a_name
f.type = a_type f.type = a_type
# Assume it's a normal field until proven otherwise. We're next # Assume it's a normal field until proven otherwise. We're next
# going to decide if it's a ClassVar or InitVar, everything else # going to decide if it's a ClassVar or InitVar, everything else
# is just a normal field. # is just a normal field.
f._field_type = _FIELD f._field_type = _FIELD
# In addition to checking for actual types here, also check for # In addition to checking for actual types here, also check for
# string annotations. get_type_hints() won't always work for us # string annotations. get_type_hints() won't always work for us
# (see https://github.com/python/typing/issues/508 for example), # (see https://github.com/python/typing/issues/508 for example),
# plus it's expensive and would require an eval for every stirng # plus it's expensive and would require an eval for every stirng
# annotation. So, make a best effort to see if this is a # annotation. So, make a best effort to see if this is a ClassVar
# ClassVar or InitVar using regex's and checking that the thing # or InitVar using regex's and checking that the thing referenced
# referenced is actually of the correct type. # is actually of the correct type.
# For the complete discussion, see https://bugs.python.org/issue33453 # For the complete discussion, see https://bugs.python.org/issue33453
# If typing has not been imported, then it's impossible for any # If typing has not been imported, then it's impossible for any
# annotation to be a ClassVar. So, only look for ClassVar if # annotation to be a ClassVar. So, only look for ClassVar if
# typing has been imported by any module (not necessarily cls's # typing has been imported by any module (not necessarily cls's
# module). # module).
typing = sys.modules.get('typing') typing = sys.modules.get('typing')
if typing: if typing:
# This test uses a typing internal class, but it's the best
# way to test if this is a ClassVar.
if (_is_classvar(a_type, typing) if (_is_classvar(a_type, typing)
or (isinstance(f.type, str) or (isinstance(f.type, str)
and _is_type(f.type, cls, typing, typing.ClassVar, and _is_type(f.type, cls, typing, typing.ClassVar,
@ -682,10 +686,10 @@ def _get_field(cls, a_name, a_type):
raise TypeError(f'field {f.name} cannot have a ' raise TypeError(f'field {f.name} cannot have a '
'default factory') 'default factory')
# Should I check for other field settings? default_factory # Should I check for other field settings? default_factory
# seems the most serious to check for. Maybe add others. # seems the most serious to check for. Maybe add others. For
# For example, how about init=False (or really, # example, how about init=False (or really,
# init=<not-the-default-init-value>)? It makes no sense for # init=<not-the-default-init-value>)? It makes no sense for
# ClassVar and InitVar to specify init=<anything>. # ClassVar and InitVar to specify init=<anything>.
# For real fields, disallow mutable defaults for known types. # For real fields, disallow mutable defaults for known types.
if f._field_type is _FIELD and isinstance(f.default, (list, dict, set)): if f._field_type is _FIELD and isinstance(f.default, (list, dict, set)):
@ -697,7 +701,7 @@ def _get_field(cls, a_name, a_type):
def _set_new_attribute(cls, name, value): def _set_new_attribute(cls, name, value):
# Never overwrites an existing attribute. Returns True if the # Never overwrites an existing attribute. Returns True if the
# attribute already exists. # attribute already exists.
if name in cls.__dict__: if name in cls.__dict__:
return True return True
setattr(cls, name, value) setattr(cls, name, value)
@ -705,9 +709,9 @@ def _set_new_attribute(cls, name, value):
# Decide if/how we're going to create a hash function. Key is # Decide if/how we're going to create a hash function. Key is
# (unsafe_hash, eq, frozen, does-hash-exist). Value is the action to # (unsafe_hash, eq, frozen, does-hash-exist). Value is the action to
# take. The common case is to do nothing, so instead of providing a # take. The common case is to do nothing, so instead of providing a
# function that is a no-op, use None to signify that. # function that is a no-op, use None to signify that.
def _hash_set_none(cls, fields): def _hash_set_none(cls, fields):
return None return None
@ -748,28 +752,28 @@ _hash_action = {(False, False, False, False): None,
(True, True, True, True ): _hash_exception, (True, True, True, True ): _hash_exception,
} }
# See https://bugs.python.org/issue32929#msg312829 for an if-statement # See https://bugs.python.org/issue32929#msg312829 for an if-statement
# version of this table. # version of this table.
def _process_class(cls, init, repr, eq, order, unsafe_hash, frozen): def _process_class(cls, init, repr, eq, order, unsafe_hash, frozen):
# Now that dicts retain insertion order, there's no reason to use # Now that dicts retain insertion order, there's no reason to use
# an ordered dict. I am leveraging that ordering here, because # an ordered dict. I am leveraging that ordering here, because
# derived class fields overwrite base class fields, but the order # derived class fields overwrite base class fields, but the order
# is defined by the base class, which is found first. # is defined by the base class, which is found first.
fields = {} fields = {}
setattr(cls, _PARAMS, _DataclassParams(init, repr, eq, order, setattr(cls, _PARAMS, _DataclassParams(init, repr, eq, order,
unsafe_hash, frozen)) unsafe_hash, frozen))
# Find our base classes in reverse MRO order, and exclude # Find our base classes in reverse MRO order, and exclude
# ourselves. In reversed order so that more derived classes # ourselves. In reversed order so that more derived classes
# override earlier field definitions in base classes. # override earlier field definitions in base classes. As long as
# As long as we're iterating over them, see if any are frozen. # we're iterating over them, see if any are frozen.
any_frozen_base = False any_frozen_base = False
has_dataclass_bases = False has_dataclass_bases = False
for b in cls.__mro__[-1:0:-1]: for b in cls.__mro__[-1:0:-1]:
# Only process classes that have been processed by our # Only process classes that have been processed by our
# decorator. That is, they have a _FIELDS attribute. # decorator. That is, they have a _FIELDS attribute.
base_fields = getattr(b, _FIELDS, None) base_fields = getattr(b, _FIELDS, None)
if base_fields: if base_fields:
has_dataclass_bases = True has_dataclass_bases = True
@ -779,38 +783,39 @@ def _process_class(cls, init, repr, eq, order, unsafe_hash, frozen):
any_frozen_base = True any_frozen_base = True
# Annotations that are defined in this class (not in base # Annotations that are defined in this class (not in base
# classes). If __annotations__ isn't present, then this class # classes). If __annotations__ isn't present, then this class
# adds no new annotations. We use this to compute fields that # adds no new annotations. We use this to compute fields that are
# are added by this class. # added by this class.
#
# Fields are found from cls_annotations, which is guaranteed to be # Fields are found from cls_annotations, which is guaranteed to be
# ordered. Default values are from class attributes, if a field # ordered. Default values are from class attributes, if a field
# has a default. If the default value is a Field(), then it # has a default. If the default value is a Field(), then it
# contains additional info beyond (and possibly including) the # contains additional info beyond (and possibly including) the
# actual default value. Pseudo-fields ClassVars and InitVars are # actual default value. Pseudo-fields ClassVars and InitVars are
# included, despite the fact that they're not real fields. # included, despite the fact that they're not real fields. That's
# That's dealt with later. # dealt with later.
cls_annotations = cls.__dict__.get('__annotations__', {}) cls_annotations = cls.__dict__.get('__annotations__', {})
# Now find fields in our class. While doing so, validate some # Now find fields in our class. While doing so, validate some
# things, and set the default values (as class attributes) # things, and set the default values (as class attributes) where
# where we can. # we can.
cls_fields = [_get_field(cls, name, type) cls_fields = [_get_field(cls, name, type)
for name, type in cls_annotations.items()] for name, type in cls_annotations.items()]
for f in cls_fields: for f in cls_fields:
fields[f.name] = f fields[f.name] = f
# If the class attribute (which is the default value for # If the class attribute (which is the default value for this
# this field) exists and is of type 'Field', replace it # field) exists and is of type 'Field', replace it with the
# with the real default. This is so that normal class # real default. This is so that normal class introspection
# introspection sees a real default value, not a Field. # sees a real default value, not a Field.
if isinstance(getattr(cls, f.name, None), Field): if isinstance(getattr(cls, f.name, None), Field):
if f.default is MISSING: if f.default is MISSING:
# If there's no default, delete the class attribute. # If there's no default, delete the class attribute.
# This happens if we specify field(repr=False), for # This happens if we specify field(repr=False), for
# example (that is, we specified a field object, but # example (that is, we specified a field object, but
# no default value). Also if we're using a default # no default value). Also if we're using a default
# factory. The class attribute should not be set at # factory. The class attribute should not be set at
# all in the post-processed class. # all in the post-processed class.
delattr(cls, f.name) delattr(cls, f.name)
else: else:
setattr(cls, f.name, f.default) setattr(cls, f.name, f.default)
@ -832,21 +837,21 @@ def _process_class(cls, init, repr, eq, order, unsafe_hash, frozen):
raise TypeError('cannot inherit frozen dataclass from a ' raise TypeError('cannot inherit frozen dataclass from a '
'non-frozen one') 'non-frozen one')
# Remember all of the fields on our class (including bases). This also # Remember all of the fields on our class (including bases). This
# marks this class as being a dataclass. # also marks this class as being a dataclass.
setattr(cls, _FIELDS, fields) setattr(cls, _FIELDS, fields)
# Was this class defined with an explicit __hash__? Note that if # Was this class defined with an explicit __hash__? Note that if
# __eq__ is defined in this class, then python will automatically # __eq__ is defined in this class, then python will automatically
# set __hash__ to None. This is a heuristic, as it's possible # set __hash__ to None. This is a heuristic, as it's possible
# that such a __hash__ == None was not auto-generated, but it # that such a __hash__ == None was not auto-generated, but it
# close enough. # close enough.
class_hash = cls.__dict__.get('__hash__', MISSING) class_hash = cls.__dict__.get('__hash__', MISSING)
has_explicit_hash = not (class_hash is MISSING or has_explicit_hash = not (class_hash is MISSING or
(class_hash is None and '__eq__' in cls.__dict__)) (class_hash is None and '__eq__' in cls.__dict__))
# If we're generating ordering methods, we must be generating # If we're generating ordering methods, we must be generating the
# the eq methods. # eq methods.
if order and not eq: if order and not eq:
raise ValueError('eq must be true if order is true') raise ValueError('eq must be true if order is true')
@ -861,14 +866,15 @@ def _process_class(cls, init, repr, eq, order, unsafe_hash, frozen):
_init_fn(flds, _init_fn(flds,
frozen, frozen,
has_post_init, has_post_init,
# The name to use for the "self" param # The name to use for the "self"
# in __init__. Use "self" if possible. # param in __init__. Use "self"
# if possible.
'__dataclass_self__' if 'self' in fields '__dataclass_self__' if 'self' in fields
else 'self', else 'self',
)) ))
# Get the fields as a list, and include only real fields. This is # Get the fields as a list, and include only real fields. This is
# used in all of the following methods. # used in all of the following methods.
field_list = [f for f in fields.values() if f._field_type is _FIELD] field_list = [f for f in fields.values() if f._field_type is _FIELD]
if repr: if repr:
@ -877,7 +883,7 @@ def _process_class(cls, init, repr, eq, order, unsafe_hash, frozen):
if eq: if eq:
# Create _eq__ method. There's no need for a __ne__ method, # Create _eq__ method. There's no need for a __ne__ method,
# since python will call __eq__ and negate it. # since python will call __eq__ and negate it.
flds = [f for f in field_list if f.compare] flds = [f for f in field_list if f.compare]
self_tuple = _tuple_str('self', flds) self_tuple = _tuple_str('self', flds)
other_tuple = _tuple_str('other', flds) other_tuple = _tuple_str('other', flds)
@ -914,7 +920,7 @@ def _process_class(cls, init, repr, eq, order, unsafe_hash, frozen):
has_explicit_hash] has_explicit_hash]
if hash_action: if hash_action:
# No need to call _set_new_attribute here, since by the time # No need to call _set_new_attribute here, since by the time
# we're here the overwriting is unconditional. # we're here the overwriting is unconditional.
cls.__hash__ = hash_action(cls, field_list) cls.__hash__ = hash_action(cls, field_list)
if not getattr(cls, '__doc__'): if not getattr(cls, '__doc__'):
@ -926,8 +932,8 @@ def _process_class(cls, init, repr, eq, order, unsafe_hash, frozen):
# _cls should never be specified by keyword, so start it with an # _cls should never be specified by keyword, so start it with an
# underscore. The presence of _cls is used to detect if this # underscore. The presence of _cls is used to detect if this
# decorator is being called with parameters or not. # decorator is being called with parameters or not.
def dataclass(_cls=None, *, init=True, repr=True, eq=True, order=False, def dataclass(_cls=None, *, init=True, repr=True, eq=True, order=False,
unsafe_hash=False, frozen=False): unsafe_hash=False, frozen=False):
"""Returns the same class as was passed in, with dunder methods """Returns the same class as was passed in, with dunder methods
@ -968,7 +974,7 @@ def fields(class_or_instance):
raise TypeError('must be called with a dataclass type or instance') raise TypeError('must be called with a dataclass type or instance')
# Exclude pseudo-fields. Note that fields is sorted by insertion # Exclude pseudo-fields. Note that fields is sorted by insertion
# order, so the order of the tuple is as the fields were defined. # order, so the order of the tuple is as the fields were defined.
return tuple(f for f in fields.values() if f._field_type is _FIELD) return tuple(f for f in fields.values() if f._field_type is _FIELD)
@ -1130,8 +1136,8 @@ def replace(obj, **changes):
assert c1.x == 3 and c1.y == 2 assert c1.x == 3 and c1.y == 2
""" """
# We're going to mutate 'changes', but that's okay because it's a new # We're going to mutate 'changes', but that's okay because it's a
# dict, even if called with 'replace(obj, **my_changes)'. # new dict, even if called with 'replace(obj, **my_changes)'.
if not _is_dataclass_instance(obj): if not _is_dataclass_instance(obj):
raise TypeError("replace() should be called on dataclass instances") raise TypeError("replace() should be called on dataclass instances")
@ -1152,8 +1158,8 @@ def replace(obj, **changes):
changes[f.name] = getattr(obj, f.name) changes[f.name] = getattr(obj, f.name)
# Create the new object, which calls __init__() and # Create the new object, which calls __init__() and
# __post_init__() (if defined), using all of the init fields # __post_init__() (if defined), using all of the init fields we've
# we've added and/or left in 'changes'. If there are values # added and/or left in 'changes'. If there are values supplied in
# supplied in changes that aren't fields, this will correctly # changes that aren't fields, this will correctly raise a
# raise a TypeError. # TypeError.
return obj.__class__(**changes) return obj.__class__(**changes)