648 lines
23 KiB
Python
648 lines
23 KiB
Python
"""
|
|
Contains all classes and functions to deal with lists, dicts, generators and
|
|
iterators in general.
|
|
"""
|
|
from jedi.inference import compiled
|
|
from jedi.inference import analysis
|
|
from jedi.inference.lazy_value import LazyKnownValue, LazyKnownValues, \
|
|
LazyTreeValue
|
|
from jedi.inference.helpers import get_int_or_none, is_string, \
|
|
reraise_getitem_errors, SimpleGetItemNotFound
|
|
from jedi.inference.utils import safe_property, to_list
|
|
from jedi.inference.cache import inference_state_method_cache
|
|
from jedi.inference.filters import LazyAttributeOverwrite, publish_method
|
|
from jedi.inference.base_value import ValueSet, Value, NO_VALUES, \
|
|
ContextualizedNode, iterate_values, sentinel, \
|
|
LazyValueWrapper
|
|
from jedi.parser_utils import get_sync_comp_fors
|
|
from jedi.inference.context import CompForContext
|
|
from jedi.inference.value.dynamic_arrays import check_array_additions
|
|
|
|
|
|
class IterableMixin:
|
|
def py__next__(self, contextualized_node=None):
|
|
return self.py__iter__(contextualized_node)
|
|
|
|
def py__stop_iteration_returns(self):
|
|
return ValueSet([compiled.builtin_from_name(self.inference_state, 'None')])
|
|
|
|
# At the moment, safe values are simple values like "foo", 1 and not
|
|
# lists/dicts. Therefore as a small speed optimization we can just do the
|
|
# default instead of resolving the lazy wrapped values, that are just
|
|
# doing this in the end as well.
|
|
# This mostly speeds up patterns like `sys.version_info >= (3, 0)` in
|
|
# typeshed.
|
|
get_safe_value = Value.get_safe_value
|
|
|
|
|
|
class GeneratorBase(LazyAttributeOverwrite, IterableMixin):
|
|
array_type = None
|
|
|
|
def _get_wrapped_value(self):
|
|
instance, = self._get_cls().execute_annotation()
|
|
return instance
|
|
|
|
def _get_cls(self):
|
|
generator, = self.inference_state.typing_module.py__getattribute__('Generator')
|
|
return generator
|
|
|
|
def py__bool__(self):
|
|
return True
|
|
|
|
@publish_method('__iter__')
|
|
def _iter(self, arguments):
|
|
return ValueSet([self])
|
|
|
|
@publish_method('send')
|
|
@publish_method('__next__')
|
|
def _next(self, arguments):
|
|
return ValueSet.from_sets(lazy_value.infer() for lazy_value in self.py__iter__())
|
|
|
|
def py__stop_iteration_returns(self):
|
|
return ValueSet([compiled.builtin_from_name(self.inference_state, 'None')])
|
|
|
|
@property
|
|
def name(self):
|
|
return compiled.CompiledValueName(self, 'Generator')
|
|
|
|
def get_annotated_class_object(self):
|
|
from jedi.inference.gradual.generics import TupleGenericManager
|
|
gen_values = self.merge_types_of_iterate().py__class__()
|
|
gm = TupleGenericManager((gen_values, NO_VALUES, NO_VALUES))
|
|
return self._get_cls().with_generics(gm)
|
|
|
|
|
|
class Generator(GeneratorBase):
|
|
"""Handling of `yield` functions."""
|
|
def __init__(self, inference_state, func_execution_context):
|
|
super().__init__(inference_state)
|
|
self._func_execution_context = func_execution_context
|
|
|
|
def py__iter__(self, contextualized_node=None):
|
|
iterators = self._func_execution_context.infer_annotations()
|
|
if iterators:
|
|
return iterators.iterate(contextualized_node)
|
|
return self._func_execution_context.get_yield_lazy_values()
|
|
|
|
def py__stop_iteration_returns(self):
|
|
return self._func_execution_context.get_return_values()
|
|
|
|
def __repr__(self):
|
|
return "<%s of %s>" % (type(self).__name__, self._func_execution_context)
|
|
|
|
|
|
def comprehension_from_atom(inference_state, value, atom):
|
|
bracket = atom.children[0]
|
|
test_list_comp = atom.children[1]
|
|
|
|
if bracket == '{':
|
|
if atom.children[1].children[1] == ':':
|
|
sync_comp_for = test_list_comp.children[3]
|
|
if sync_comp_for.type == 'comp_for':
|
|
sync_comp_for = sync_comp_for.children[1]
|
|
|
|
return DictComprehension(
|
|
inference_state,
|
|
value,
|
|
sync_comp_for_node=sync_comp_for,
|
|
key_node=test_list_comp.children[0],
|
|
value_node=test_list_comp.children[2],
|
|
)
|
|
else:
|
|
cls = SetComprehension
|
|
elif bracket == '(':
|
|
cls = GeneratorComprehension
|
|
elif bracket == '[':
|
|
cls = ListComprehension
|
|
|
|
sync_comp_for = test_list_comp.children[1]
|
|
if sync_comp_for.type == 'comp_for':
|
|
sync_comp_for = sync_comp_for.children[1]
|
|
|
|
return cls(
|
|
inference_state,
|
|
defining_context=value,
|
|
sync_comp_for_node=sync_comp_for,
|
|
entry_node=test_list_comp.children[0],
|
|
)
|
|
|
|
|
|
class ComprehensionMixin:
|
|
@inference_state_method_cache()
|
|
def _get_comp_for_context(self, parent_context, comp_for):
|
|
return CompForContext(parent_context, comp_for)
|
|
|
|
def _nested(self, comp_fors, parent_context=None):
|
|
comp_for = comp_fors[0]
|
|
|
|
is_async = comp_for.parent.type == 'comp_for'
|
|
|
|
input_node = comp_for.children[3]
|
|
parent_context = parent_context or self._defining_context
|
|
input_types = parent_context.infer_node(input_node)
|
|
|
|
cn = ContextualizedNode(parent_context, input_node)
|
|
iterated = input_types.iterate(cn, is_async=is_async)
|
|
exprlist = comp_for.children[1]
|
|
for i, lazy_value in enumerate(iterated):
|
|
types = lazy_value.infer()
|
|
dct = unpack_tuple_to_dict(parent_context, types, exprlist)
|
|
context = self._get_comp_for_context(
|
|
parent_context,
|
|
comp_for,
|
|
)
|
|
with context.predefine_names(comp_for, dct):
|
|
try:
|
|
yield from self._nested(comp_fors[1:], context)
|
|
except IndexError:
|
|
iterated = context.infer_node(self._entry_node)
|
|
if self.array_type == 'dict':
|
|
yield iterated, context.infer_node(self._value_node)
|
|
else:
|
|
yield iterated
|
|
|
|
@inference_state_method_cache(default=[])
|
|
@to_list
|
|
def _iterate(self):
|
|
comp_fors = tuple(get_sync_comp_fors(self._sync_comp_for_node))
|
|
yield from self._nested(comp_fors)
|
|
|
|
def py__iter__(self, contextualized_node=None):
|
|
for set_ in self._iterate():
|
|
yield LazyKnownValues(set_)
|
|
|
|
def __repr__(self):
|
|
return "<%s of %s>" % (type(self).__name__, self._sync_comp_for_node)
|
|
|
|
|
|
class _DictMixin:
|
|
def _get_generics(self):
|
|
return tuple(c_set.py__class__() for c_set in self.get_mapping_item_values())
|
|
|
|
|
|
class Sequence(LazyAttributeOverwrite, IterableMixin):
|
|
api_type = 'instance'
|
|
|
|
@property
|
|
def name(self):
|
|
return compiled.CompiledValueName(self, self.array_type)
|
|
|
|
def _get_generics(self):
|
|
return (self.merge_types_of_iterate().py__class__(),)
|
|
|
|
@inference_state_method_cache(default=())
|
|
def _cached_generics(self):
|
|
return self._get_generics()
|
|
|
|
def _get_wrapped_value(self):
|
|
from jedi.inference.gradual.base import GenericClass
|
|
from jedi.inference.gradual.generics import TupleGenericManager
|
|
klass = compiled.builtin_from_name(self.inference_state, self.array_type)
|
|
c, = GenericClass(
|
|
klass,
|
|
TupleGenericManager(self._cached_generics())
|
|
).execute_annotation()
|
|
return c
|
|
|
|
def py__bool__(self):
|
|
return None # We don't know the length, because of appends.
|
|
|
|
@safe_property
|
|
def parent(self):
|
|
return self.inference_state.builtins_module
|
|
|
|
def py__getitem__(self, index_value_set, contextualized_node):
|
|
if self.array_type == 'dict':
|
|
return self._dict_values()
|
|
return iterate_values(ValueSet([self]))
|
|
|
|
|
|
class _BaseComprehension(ComprehensionMixin):
|
|
def __init__(self, inference_state, defining_context, sync_comp_for_node, entry_node):
|
|
assert sync_comp_for_node.type == 'sync_comp_for'
|
|
super().__init__(inference_state)
|
|
self._defining_context = defining_context
|
|
self._sync_comp_for_node = sync_comp_for_node
|
|
self._entry_node = entry_node
|
|
|
|
|
|
class ListComprehension(_BaseComprehension, Sequence):
|
|
array_type = 'list'
|
|
|
|
def py__simple_getitem__(self, index):
|
|
if isinstance(index, slice):
|
|
return ValueSet([self])
|
|
|
|
all_types = list(self.py__iter__())
|
|
with reraise_getitem_errors(IndexError, TypeError):
|
|
lazy_value = all_types[index]
|
|
return lazy_value.infer()
|
|
|
|
|
|
class SetComprehension(_BaseComprehension, Sequence):
|
|
array_type = 'set'
|
|
|
|
|
|
class GeneratorComprehension(_BaseComprehension, GeneratorBase):
|
|
pass
|
|
|
|
|
|
class _DictKeyMixin:
|
|
# TODO merge with _DictMixin?
|
|
def get_mapping_item_values(self):
|
|
return self._dict_keys(), self._dict_values()
|
|
|
|
def get_key_values(self):
|
|
# TODO merge with _dict_keys?
|
|
return self._dict_keys()
|
|
|
|
|
|
class DictComprehension(ComprehensionMixin, Sequence, _DictKeyMixin):
|
|
array_type = 'dict'
|
|
|
|
def __init__(self, inference_state, defining_context, sync_comp_for_node, key_node, value_node):
|
|
assert sync_comp_for_node.type == 'sync_comp_for'
|
|
super().__init__(inference_state)
|
|
self._defining_context = defining_context
|
|
self._sync_comp_for_node = sync_comp_for_node
|
|
self._entry_node = key_node
|
|
self._value_node = value_node
|
|
|
|
def py__iter__(self, contextualized_node=None):
|
|
for keys, values in self._iterate():
|
|
yield LazyKnownValues(keys)
|
|
|
|
def py__simple_getitem__(self, index):
|
|
for keys, values in self._iterate():
|
|
for k in keys:
|
|
# Be careful in the future if refactoring, index could be a
|
|
# slice object.
|
|
if k.get_safe_value(default=object()) == index:
|
|
return values
|
|
raise SimpleGetItemNotFound()
|
|
|
|
def _dict_keys(self):
|
|
return ValueSet.from_sets(keys for keys, values in self._iterate())
|
|
|
|
def _dict_values(self):
|
|
return ValueSet.from_sets(values for keys, values in self._iterate())
|
|
|
|
@publish_method('values')
|
|
def _imitate_values(self, arguments):
|
|
lazy_value = LazyKnownValues(self._dict_values())
|
|
return ValueSet([FakeList(self.inference_state, [lazy_value])])
|
|
|
|
@publish_method('items')
|
|
def _imitate_items(self, arguments):
|
|
lazy_values = [
|
|
LazyKnownValue(
|
|
FakeTuple(
|
|
self.inference_state,
|
|
[LazyKnownValues(key),
|
|
LazyKnownValues(value)]
|
|
)
|
|
)
|
|
for key, value in self._iterate()
|
|
]
|
|
|
|
return ValueSet([FakeList(self.inference_state, lazy_values)])
|
|
|
|
def exact_key_items(self):
|
|
# NOTE: A smarter thing can probably done here to achieve better
|
|
# completions, but at least like this jedi doesn't crash
|
|
return []
|
|
|
|
|
|
class SequenceLiteralValue(Sequence):
|
|
_TUPLE_LIKE = 'testlist_star_expr', 'testlist', 'subscriptlist'
|
|
mapping = {'(': 'tuple',
|
|
'[': 'list',
|
|
'{': 'set'}
|
|
|
|
def __init__(self, inference_state, defining_context, atom):
|
|
super().__init__(inference_state)
|
|
self.atom = atom
|
|
self._defining_context = defining_context
|
|
|
|
if self.atom.type in self._TUPLE_LIKE:
|
|
self.array_type = 'tuple'
|
|
else:
|
|
self.array_type = SequenceLiteralValue.mapping[atom.children[0]]
|
|
"""The builtin name of the array (list, set, tuple or dict)."""
|
|
|
|
def _get_generics(self):
|
|
if self.array_type == 'tuple':
|
|
return tuple(x.infer().py__class__() for x in self.py__iter__())
|
|
return super()._get_generics()
|
|
|
|
def py__simple_getitem__(self, index):
|
|
"""Here the index is an int/str. Raises IndexError/KeyError."""
|
|
if isinstance(index, slice):
|
|
return ValueSet([self])
|
|
else:
|
|
with reraise_getitem_errors(TypeError, KeyError, IndexError):
|
|
node = self.get_tree_entries()[index]
|
|
if node == ':' or node.type == 'subscript':
|
|
return NO_VALUES
|
|
return self._defining_context.infer_node(node)
|
|
|
|
def py__iter__(self, contextualized_node=None):
|
|
"""
|
|
While values returns the possible values for any array field, this
|
|
function returns the value for a certain index.
|
|
"""
|
|
for node in self.get_tree_entries():
|
|
if node == ':' or node.type == 'subscript':
|
|
# TODO this should probably use at least part of the code
|
|
# of infer_subscript_list.
|
|
yield LazyKnownValue(Slice(self._defining_context, None, None, None))
|
|
else:
|
|
yield LazyTreeValue(self._defining_context, node)
|
|
yield from check_array_additions(self._defining_context, self)
|
|
|
|
def py__len__(self):
|
|
# This function is not really used often. It's more of a try.
|
|
return len(self.get_tree_entries())
|
|
|
|
def get_tree_entries(self):
|
|
c = self.atom.children
|
|
|
|
if self.atom.type in self._TUPLE_LIKE:
|
|
return c[::2]
|
|
|
|
array_node = c[1]
|
|
if array_node in (']', '}', ')'):
|
|
return [] # Direct closing bracket, doesn't contain items.
|
|
|
|
if array_node.type == 'testlist_comp':
|
|
# filter out (for now) pep 448 single-star unpacking
|
|
return [value for value in array_node.children[::2]
|
|
if value.type != "star_expr"]
|
|
elif array_node.type == 'dictorsetmaker':
|
|
kv = []
|
|
iterator = iter(array_node.children)
|
|
for key in iterator:
|
|
if key == "**":
|
|
# dict with pep 448 double-star unpacking
|
|
# for now ignoring the values imported by **
|
|
next(iterator)
|
|
next(iterator, None) # Possible comma.
|
|
else:
|
|
op = next(iterator, None)
|
|
if op is None or op == ',':
|
|
if key.type == "star_expr":
|
|
# pep 448 single-star unpacking
|
|
# for now ignoring values imported by *
|
|
pass
|
|
else:
|
|
kv.append(key) # A set.
|
|
else:
|
|
assert op == ':' # A dict.
|
|
kv.append((key, next(iterator)))
|
|
next(iterator, None) # Possible comma.
|
|
return kv
|
|
else:
|
|
if array_node.type == "star_expr":
|
|
# pep 448 single-star unpacking
|
|
# for now ignoring values imported by *
|
|
return []
|
|
else:
|
|
return [array_node]
|
|
|
|
def __repr__(self):
|
|
return "<%s of %s>" % (self.__class__.__name__, self.atom)
|
|
|
|
|
|
class DictLiteralValue(_DictMixin, SequenceLiteralValue, _DictKeyMixin):
|
|
array_type = 'dict'
|
|
|
|
def __init__(self, inference_state, defining_context, atom):
|
|
# Intentionally don't call the super class. This is definitely a sign
|
|
# that the architecture is bad and we should refactor.
|
|
Sequence.__init__(self, inference_state)
|
|
self._defining_context = defining_context
|
|
self.atom = atom
|
|
|
|
def py__simple_getitem__(self, index):
|
|
"""Here the index is an int/str. Raises IndexError/KeyError."""
|
|
compiled_value_index = compiled.create_simple_object(self.inference_state, index)
|
|
for key, value in self.get_tree_entries():
|
|
for k in self._defining_context.infer_node(key):
|
|
for key_v in k.execute_operation(compiled_value_index, '=='):
|
|
if key_v.get_safe_value():
|
|
return self._defining_context.infer_node(value)
|
|
raise SimpleGetItemNotFound('No key found in dictionary %s.' % self)
|
|
|
|
def py__iter__(self, contextualized_node=None):
|
|
"""
|
|
While values returns the possible values for any array field, this
|
|
function returns the value for a certain index.
|
|
"""
|
|
# Get keys.
|
|
types = NO_VALUES
|
|
for k, _ in self.get_tree_entries():
|
|
types |= self._defining_context.infer_node(k)
|
|
# We don't know which dict index comes first, therefore always
|
|
# yield all the types.
|
|
for _ in types:
|
|
yield LazyKnownValues(types)
|
|
|
|
@publish_method('values')
|
|
def _imitate_values(self, arguments):
|
|
lazy_value = LazyKnownValues(self._dict_values())
|
|
return ValueSet([FakeList(self.inference_state, [lazy_value])])
|
|
|
|
@publish_method('items')
|
|
def _imitate_items(self, arguments):
|
|
lazy_values = [
|
|
LazyKnownValue(FakeTuple(
|
|
self.inference_state,
|
|
(LazyTreeValue(self._defining_context, key_node),
|
|
LazyTreeValue(self._defining_context, value_node))
|
|
)) for key_node, value_node in self.get_tree_entries()
|
|
]
|
|
|
|
return ValueSet([FakeList(self.inference_state, lazy_values)])
|
|
|
|
def exact_key_items(self):
|
|
"""
|
|
Returns a generator of tuples like dict.items(), where the key is
|
|
resolved (as a string) and the values are still lazy values.
|
|
"""
|
|
for key_node, value in self.get_tree_entries():
|
|
for key in self._defining_context.infer_node(key_node):
|
|
if is_string(key):
|
|
yield key.get_safe_value(), LazyTreeValue(self._defining_context, value)
|
|
|
|
def _dict_values(self):
|
|
return ValueSet.from_sets(
|
|
self._defining_context.infer_node(v)
|
|
for k, v in self.get_tree_entries()
|
|
)
|
|
|
|
def _dict_keys(self):
|
|
return ValueSet.from_sets(
|
|
self._defining_context.infer_node(k)
|
|
for k, v in self.get_tree_entries()
|
|
)
|
|
|
|
|
|
class _FakeSequence(Sequence):
|
|
def __init__(self, inference_state, lazy_value_list):
|
|
"""
|
|
type should be one of "tuple", "list"
|
|
"""
|
|
super().__init__(inference_state)
|
|
self._lazy_value_list = lazy_value_list
|
|
|
|
def py__simple_getitem__(self, index):
|
|
if isinstance(index, slice):
|
|
return ValueSet([self])
|
|
|
|
with reraise_getitem_errors(IndexError, TypeError):
|
|
lazy_value = self._lazy_value_list[index]
|
|
return lazy_value.infer()
|
|
|
|
def py__iter__(self, contextualized_node=None):
|
|
return self._lazy_value_list
|
|
|
|
def py__bool__(self):
|
|
return bool(len(self._lazy_value_list))
|
|
|
|
def __repr__(self):
|
|
return "<%s of %s>" % (type(self).__name__, self._lazy_value_list)
|
|
|
|
|
|
class FakeTuple(_FakeSequence):
|
|
array_type = 'tuple'
|
|
|
|
|
|
class FakeList(_FakeSequence):
|
|
array_type = 'tuple'
|
|
|
|
|
|
class FakeDict(_DictMixin, Sequence, _DictKeyMixin):
|
|
array_type = 'dict'
|
|
|
|
def __init__(self, inference_state, dct):
|
|
super().__init__(inference_state)
|
|
self._dct = dct
|
|
|
|
def py__iter__(self, contextualized_node=None):
|
|
for key in self._dct:
|
|
yield LazyKnownValue(compiled.create_simple_object(self.inference_state, key))
|
|
|
|
def py__simple_getitem__(self, index):
|
|
with reraise_getitem_errors(KeyError, TypeError):
|
|
lazy_value = self._dct[index]
|
|
return lazy_value.infer()
|
|
|
|
@publish_method('values')
|
|
def _values(self, arguments):
|
|
return ValueSet([FakeTuple(
|
|
self.inference_state,
|
|
[LazyKnownValues(self._dict_values())]
|
|
)])
|
|
|
|
def _dict_values(self):
|
|
return ValueSet.from_sets(lazy_value.infer() for lazy_value in self._dct.values())
|
|
|
|
def _dict_keys(self):
|
|
return ValueSet.from_sets(lazy_value.infer() for lazy_value in self.py__iter__())
|
|
|
|
def exact_key_items(self):
|
|
return self._dct.items()
|
|
|
|
def __repr__(self):
|
|
return '<%s: %s>' % (self.__class__.__name__, self._dct)
|
|
|
|
|
|
class MergedArray(Sequence):
|
|
def __init__(self, inference_state, arrays):
|
|
super().__init__(inference_state)
|
|
self.array_type = arrays[-1].array_type
|
|
self._arrays = arrays
|
|
|
|
def py__iter__(self, contextualized_node=None):
|
|
for array in self._arrays:
|
|
yield from array.py__iter__()
|
|
|
|
def py__simple_getitem__(self, index):
|
|
return ValueSet.from_sets(lazy_value.infer() for lazy_value in self.py__iter__())
|
|
|
|
|
|
def unpack_tuple_to_dict(context, types, exprlist):
|
|
"""
|
|
Unpacking tuple assignments in for statements and expr_stmts.
|
|
"""
|
|
if exprlist.type == 'name':
|
|
return {exprlist.value: types}
|
|
elif exprlist.type == 'atom' and exprlist.children[0] in ('(', '['):
|
|
return unpack_tuple_to_dict(context, types, exprlist.children[1])
|
|
elif exprlist.type in ('testlist', 'testlist_comp', 'exprlist',
|
|
'testlist_star_expr'):
|
|
dct = {}
|
|
parts = iter(exprlist.children[::2])
|
|
n = 0
|
|
for lazy_value in types.iterate(ContextualizedNode(context, exprlist)):
|
|
n += 1
|
|
try:
|
|
part = next(parts)
|
|
except StopIteration:
|
|
analysis.add(context, 'value-error-too-many-values', part,
|
|
message="ValueError: too many values to unpack (expected %s)" % n)
|
|
else:
|
|
dct.update(unpack_tuple_to_dict(context, lazy_value.infer(), part))
|
|
has_parts = next(parts, None)
|
|
if types and has_parts is not None:
|
|
analysis.add(context, 'value-error-too-few-values', has_parts,
|
|
message="ValueError: need more than %s values to unpack" % n)
|
|
return dct
|
|
elif exprlist.type == 'power' or exprlist.type == 'atom_expr':
|
|
# Something like ``arr[x], var = ...``.
|
|
# This is something that is not yet supported, would also be difficult
|
|
# to write into a dict.
|
|
return {}
|
|
elif exprlist.type == 'star_expr': # `a, *b, c = x` type unpackings
|
|
# Currently we're not supporting them.
|
|
return {}
|
|
raise NotImplementedError
|
|
|
|
|
|
class Slice(LazyValueWrapper):
|
|
def __init__(self, python_context, start, stop, step):
|
|
self.inference_state = python_context.inference_state
|
|
self._context = python_context
|
|
# All of them are either a Precedence or None.
|
|
self._start = start
|
|
self._stop = stop
|
|
self._step = step
|
|
|
|
def _get_wrapped_value(self):
|
|
value = compiled.builtin_from_name(self._context.inference_state, 'slice')
|
|
slice_value, = value.execute_with_values()
|
|
return slice_value
|
|
|
|
def get_safe_value(self, default=sentinel):
|
|
"""
|
|
Imitate CompiledValue.obj behavior and return a ``builtin.slice()``
|
|
object.
|
|
"""
|
|
def get(element):
|
|
if element is None:
|
|
return None
|
|
|
|
result = self._context.infer_node(element)
|
|
if len(result) != 1:
|
|
# For simplicity, we want slices to be clear defined with just
|
|
# one type. Otherwise we will return an empty slice object.
|
|
raise IndexError
|
|
|
|
value, = result
|
|
return get_int_or_none(value)
|
|
|
|
try:
|
|
return slice(get(self._start), get(self._stop), get(self._step))
|
|
except IndexError:
|
|
return slice(None, None, None)
|