AIM-PIbd-32-Kurbanova-A-A/aimenv/Lib/site-packages/jedi/inference/value/function.py

462 lines
17 KiB
Python
Raw Normal View History

2024-10-02 22:15:59 +04:00
from parso.python import tree
from jedi import debug
from jedi.inference.cache import inference_state_method_cache, CachedMetaClass
from jedi.inference import compiled
from jedi.inference import recursion
from jedi.inference import docstrings
from jedi.inference import flow_analysis
from jedi.inference.signature import TreeSignature
from jedi.inference.filters import ParserTreeFilter, FunctionExecutionFilter, \
AnonymousFunctionExecutionFilter
from jedi.inference.names import ValueName, AbstractNameDefinition, \
AnonymousParamName, ParamName, NameWrapper
from jedi.inference.base_value import ContextualizedNode, NO_VALUES, \
ValueSet, TreeValue, ValueWrapper
from jedi.inference.lazy_value import LazyKnownValues, LazyKnownValue, \
LazyTreeValue
from jedi.inference.context import ValueContext, TreeContextMixin
from jedi.inference.value import iterable
from jedi import parser_utils
from jedi.inference.parser_cache import get_yield_exprs
from jedi.inference.helpers import values_from_qualified_names
from jedi.inference.gradual.generics import TupleGenericManager
class LambdaName(AbstractNameDefinition):
string_name = '<lambda>'
api_type = 'function'
def __init__(self, lambda_value):
self._lambda_value = lambda_value
self.parent_context = lambda_value.parent_context
@property
def start_pos(self):
return self._lambda_value.tree_node.start_pos
def infer(self):
return ValueSet([self._lambda_value])
class FunctionAndClassBase(TreeValue):
def get_qualified_names(self):
if self.parent_context.is_class():
n = self.parent_context.get_qualified_names()
if n is None:
# This means that the parent class lives within a function.
return None
return n + (self.py__name__(),)
elif self.parent_context.is_module():
return (self.py__name__(),)
else:
return None
class FunctionMixin:
api_type = 'function'
def get_filters(self, origin_scope=None):
cls = self.py__class__()
for instance in cls.execute_with_values():
yield from instance.get_filters(origin_scope=origin_scope)
def py__get__(self, instance, class_value):
from jedi.inference.value.instance import BoundMethod
if instance is None:
# Calling the Foo.bar results in the original bar function.
return ValueSet([self])
return ValueSet([BoundMethod(instance, class_value.as_context(), self)])
def get_param_names(self):
return [AnonymousParamName(self, param.name)
for param in self.tree_node.get_params()]
@property
def name(self):
if self.tree_node.type == 'lambdef':
return LambdaName(self)
return ValueName(self, self.tree_node.name)
def is_function(self):
return True
def py__name__(self):
return self.name.string_name
def get_type_hint(self, add_class_info=True):
return_annotation = self.tree_node.annotation
if return_annotation is None:
def param_name_to_str(n):
s = n.string_name
annotation = n.infer().get_type_hint()
if annotation is not None:
s += ': ' + annotation
if n.default_node is not None:
s += '=' + n.default_node.get_code(include_prefix=False)
return s
function_execution = self.as_context()
result = function_execution.infer()
return_hint = result.get_type_hint()
body = self.py__name__() + '(%s)' % ', '.join([
param_name_to_str(n)
for n in function_execution.get_param_names()
])
if return_hint is None:
return body
else:
return_hint = return_annotation.get_code(include_prefix=False)
body = self.py__name__() + self.tree_node.children[2].get_code(include_prefix=False)
return body + ' -> ' + return_hint
def py__call__(self, arguments):
function_execution = self.as_context(arguments)
return function_execution.infer()
def _as_context(self, arguments=None):
if arguments is None:
return AnonymousFunctionExecution(self)
return FunctionExecutionContext(self, arguments)
def get_signatures(self):
return [TreeSignature(f) for f in self.get_signature_functions()]
class FunctionValue(FunctionMixin, FunctionAndClassBase, metaclass=CachedMetaClass):
@classmethod
def from_context(cls, context, tree_node):
def create(tree_node):
if context.is_class():
return MethodValue(
context.inference_state,
context,
parent_context=parent_context,
tree_node=tree_node
)
else:
return cls(
context.inference_state,
parent_context=parent_context,
tree_node=tree_node
)
overloaded_funcs = list(_find_overload_functions(context, tree_node))
parent_context = context
while parent_context.is_class() or parent_context.is_instance():
parent_context = parent_context.parent_context
function = create(tree_node)
if overloaded_funcs:
return OverloadedFunctionValue(
function,
# Get them into the correct order: lower line first.
list(reversed([create(f) for f in overloaded_funcs]))
)
return function
def py__class__(self):
c, = values_from_qualified_names(self.inference_state, 'types', 'FunctionType')
return c
def get_default_param_context(self):
return self.parent_context
def get_signature_functions(self):
return [self]
class FunctionNameInClass(NameWrapper):
def __init__(self, class_context, name):
super().__init__(name)
self._class_context = class_context
def get_defining_qualified_value(self):
return self._class_context.get_value() # Might be None.
class MethodValue(FunctionValue):
def __init__(self, inference_state, class_context, *args, **kwargs):
super().__init__(inference_state, *args, **kwargs)
self.class_context = class_context
def get_default_param_context(self):
return self.class_context
def get_qualified_names(self):
# Need to implement this, because the parent value of a method
# value is not the class value but the module.
names = self.class_context.get_qualified_names()
if names is None:
return None
return names + (self.py__name__(),)
@property
def name(self):
return FunctionNameInClass(self.class_context, super().name)
class BaseFunctionExecutionContext(ValueContext, TreeContextMixin):
def infer_annotations(self):
raise NotImplementedError
@inference_state_method_cache(default=NO_VALUES)
@recursion.execution_recursion_decorator()
def get_return_values(self, check_yields=False):
funcdef = self.tree_node
if funcdef.type == 'lambdef':
return self.infer_node(funcdef.children[-1])
if check_yields:
value_set = NO_VALUES
returns = get_yield_exprs(self.inference_state, funcdef)
else:
value_set = self.infer_annotations()
if value_set:
# If there are annotations, prefer them over anything else.
# This will make it faster.
return value_set
value_set |= docstrings.infer_return_types(self._value)
returns = funcdef.iter_return_stmts()
for r in returns:
if check_yields:
value_set |= ValueSet.from_sets(
lazy_value.infer()
for lazy_value in self._get_yield_lazy_value(r)
)
else:
check = flow_analysis.reachability_check(self, funcdef, r)
if check is flow_analysis.UNREACHABLE:
debug.dbg('Return unreachable: %s', r)
else:
try:
children = r.children
except AttributeError:
ctx = compiled.builtin_from_name(self.inference_state, 'None')
value_set |= ValueSet([ctx])
else:
value_set |= self.infer_node(children[1])
if check is flow_analysis.REACHABLE:
debug.dbg('Return reachable: %s', r)
break
return value_set
def _get_yield_lazy_value(self, yield_expr):
if yield_expr.type == 'keyword':
# `yield` just yields None.
ctx = compiled.builtin_from_name(self.inference_state, 'None')
yield LazyKnownValue(ctx)
return
node = yield_expr.children[1]
if node.type == 'yield_arg': # It must be a yield from.
cn = ContextualizedNode(self, node.children[1])
yield from cn.infer().iterate(cn)
else:
yield LazyTreeValue(self, node)
@recursion.execution_recursion_decorator(default=iter([]))
def get_yield_lazy_values(self, is_async=False):
# TODO: if is_async, wrap yield statements in Awaitable/async_generator_asend
for_parents = [(y, tree.search_ancestor(y, 'for_stmt', 'funcdef',
'while_stmt', 'if_stmt'))
for y in get_yield_exprs(self.inference_state, self.tree_node)]
# Calculate if the yields are placed within the same for loop.
yields_order = []
last_for_stmt = None
for yield_, for_stmt in for_parents:
# For really simple for loops we can predict the order. Otherwise
# we just ignore it.
parent = for_stmt.parent
if parent.type == 'suite':
parent = parent.parent
if for_stmt.type == 'for_stmt' and parent == self.tree_node \
and parser_utils.for_stmt_defines_one_name(for_stmt): # Simplicity for now.
if for_stmt == last_for_stmt:
yields_order[-1][1].append(yield_)
else:
yields_order.append((for_stmt, [yield_]))
elif for_stmt == self.tree_node:
yields_order.append((None, [yield_]))
else:
types = self.get_return_values(check_yields=True)
if types:
yield LazyKnownValues(types, min=0, max=float('inf'))
return
last_for_stmt = for_stmt
for for_stmt, yields in yields_order:
if for_stmt is None:
# No for_stmt, just normal yields.
for yield_ in yields:
yield from self._get_yield_lazy_value(yield_)
else:
input_node = for_stmt.get_testlist()
cn = ContextualizedNode(self, input_node)
ordered = cn.infer().iterate(cn)
ordered = list(ordered)
for lazy_value in ordered:
dct = {str(for_stmt.children[1].value): lazy_value.infer()}
with self.predefine_names(for_stmt, dct):
for yield_in_same_for_stmt in yields:
yield from self._get_yield_lazy_value(yield_in_same_for_stmt)
def merge_yield_values(self, is_async=False):
return ValueSet.from_sets(
lazy_value.infer()
for lazy_value in self.get_yield_lazy_values()
)
def is_generator(self):
return bool(get_yield_exprs(self.inference_state, self.tree_node))
def infer(self):
"""
Created to be used by inheritance.
"""
inference_state = self.inference_state
is_coroutine = self.tree_node.parent.type in ('async_stmt', 'async_funcdef')
from jedi.inference.gradual.base import GenericClass
if is_coroutine:
if self.is_generator():
async_generator_classes = inference_state.typing_module \
.py__getattribute__('AsyncGenerator')
yield_values = self.merge_yield_values(is_async=True)
# The contravariant doesn't seem to be defined.
generics = (yield_values.py__class__(), NO_VALUES)
return ValueSet(
GenericClass(c, TupleGenericManager(generics))
for c in async_generator_classes
).execute_annotation()
else:
async_classes = inference_state.typing_module.py__getattribute__('Coroutine')
return_values = self.get_return_values()
# Only the first generic is relevant.
generics = (return_values.py__class__(), NO_VALUES, NO_VALUES)
return ValueSet(
GenericClass(c, TupleGenericManager(generics)) for c in async_classes
).execute_annotation()
else:
# If there are annotations, prefer them over anything else.
if self.is_generator() and not self.infer_annotations():
return ValueSet([iterable.Generator(inference_state, self)])
else:
return self.get_return_values()
class FunctionExecutionContext(BaseFunctionExecutionContext):
def __init__(self, function_value, arguments):
super().__init__(function_value)
self._arguments = arguments
def get_filters(self, until_position=None, origin_scope=None):
yield FunctionExecutionFilter(
self, self._value,
until_position=until_position,
origin_scope=origin_scope,
arguments=self._arguments
)
def infer_annotations(self):
from jedi.inference.gradual.annotation import infer_return_types
return infer_return_types(self._value, self._arguments)
def get_param_names(self):
return [
ParamName(self._value, param.name, self._arguments)
for param in self._value.tree_node.get_params()
]
class AnonymousFunctionExecution(BaseFunctionExecutionContext):
def infer_annotations(self):
# I don't think inferring anonymous executions is a big thing.
# Anonymous contexts are mostly there for the user to work in. ~ dave
return NO_VALUES
def get_filters(self, until_position=None, origin_scope=None):
yield AnonymousFunctionExecutionFilter(
self, self._value,
until_position=until_position,
origin_scope=origin_scope,
)
def get_param_names(self):
return self._value.get_param_names()
class OverloadedFunctionValue(FunctionMixin, ValueWrapper):
def __init__(self, function, overloaded_functions):
super().__init__(function)
self._overloaded_functions = overloaded_functions
def py__call__(self, arguments):
debug.dbg("Execute overloaded function %s", self._wrapped_value, color='BLUE')
function_executions = []
for signature in self.get_signatures():
function_execution = signature.value.as_context(arguments)
function_executions.append(function_execution)
if signature.matches_signature(arguments):
return function_execution.infer()
if self.inference_state.is_analysis:
# In this case we want precision.
return NO_VALUES
return ValueSet.from_sets(fe.infer() for fe in function_executions)
def get_signature_functions(self):
return self._overloaded_functions
def get_type_hint(self, add_class_info=True):
return 'Union[%s]' % ', '.join(f.get_type_hint() for f in self._overloaded_functions)
def _find_overload_functions(context, tree_node):
def _is_overload_decorated(funcdef):
if funcdef.parent.type == 'decorated':
decorators = funcdef.parent.children[0]
if decorators.type == 'decorator':
decorators = [decorators]
else:
decorators = decorators.children
for decorator in decorators:
dotted_name = decorator.children[1]
if dotted_name.type == 'name' and dotted_name.value == 'overload':
# TODO check with values if it's the right overload
return True
return False
if tree_node.type == 'lambdef':
return
if _is_overload_decorated(tree_node):
yield tree_node
while True:
filter = ParserTreeFilter(
context,
until_position=tree_node.start_pos
)
names = filter.get(tree_node.name.value)
assert isinstance(names, list)
if not names:
break
found = False
for name in names:
funcdef = name.tree_name.parent
if funcdef.type == 'funcdef' and _is_overload_decorated(funcdef):
tree_node = funcdef
found = True
yield funcdef
if not found:
break