Hacked By AnonymousFox
# Licensed under the GPL: https://www.gnu.org/licenses/old-licenses/gpl-2.0.html
# For details: https://github.com/PyCQA/pylint/blob/main/LICENSE
# Copyright (c) https://github.com/PyCQA/pylint/blob/main/CONTRIBUTORS.txt
from __future__ import annotations
import astroid
from astroid import nodes
from pylint import checkers
from pylint.checkers import utils
from pylint.interfaces import HIGH, INFERENCE
class RecommendationChecker(checkers.BaseChecker):
name = "refactoring"
msgs = {
"C0200": (
"Consider using enumerate instead of iterating with range and len",
"consider-using-enumerate",
"Emitted when code that iterates with range and len is "
"encountered. Such code can be simplified by using the "
"enumerate builtin.",
),
"C0201": (
"Consider iterating the dictionary directly instead of calling .keys()",
"consider-iterating-dictionary",
"Emitted when the keys of a dictionary are iterated through the ``.keys()`` "
"method or when ``.keys()`` is used for a membership check. "
"It is enough to iterate through the dictionary itself, "
"``for key in dictionary``. For membership checks, "
"``if key in dictionary`` is faster.",
),
"C0206": (
"Consider iterating with .items()",
"consider-using-dict-items",
"Emitted when iterating over the keys of a dictionary and accessing the "
"value by index lookup. "
"Both the key and value can be accessed by iterating using the .items() "
"method of the dictionary instead.",
),
"C0207": (
"Use %s instead",
"use-maxsplit-arg",
"Emitted when accessing only the first or last element of str.split(). "
"The first and last element can be accessed by using "
"str.split(sep, maxsplit=1)[0] or str.rsplit(sep, maxsplit=1)[-1] "
"instead.",
),
"C0208": (
"Use a sequence type when iterating over values",
"use-sequence-for-iteration",
"When iterating over values, sequence types (e.g., ``lists``, ``tuples``, ``ranges``) "
"are more efficient than ``sets``.",
),
"C0209": (
"Formatting a regular string which could be a f-string",
"consider-using-f-string",
"Used when we detect a string that is being formatted with format() or % "
"which could potentially be a f-string. The use of f-strings is preferred. "
"Requires Python 3.6 and ``py-version >= 3.6``.",
),
}
def open(self) -> None:
py_version = self.linter.config.py_version
self._py36_plus = py_version >= (3, 6)
@staticmethod
def _is_builtin(node: nodes.NodeNG, function: str) -> bool:
inferred = utils.safe_infer(node)
if not inferred:
return False
return utils.is_builtin_object(inferred) and inferred.name == function
@utils.only_required_for_messages(
"consider-iterating-dictionary", "use-maxsplit-arg"
)
def visit_call(self, node: nodes.Call) -> None:
self._check_consider_iterating_dictionary(node)
self._check_use_maxsplit_arg(node)
def _check_consider_iterating_dictionary(self, node: nodes.Call) -> None:
if not isinstance(node.func, nodes.Attribute):
return
if node.func.attrname != "keys":
return
if isinstance(node.parent, nodes.BinOp) and node.parent.op in {"&", "|", "^"}:
return
comp_ancestor = utils.get_node_first_ancestor_of_type(node, nodes.Compare)
if (
isinstance(node.parent, (nodes.For, nodes.Comprehension))
or comp_ancestor
and any(
op
for op, comparator in comp_ancestor.ops
if op in {"in", "not in"}
and (comparator in node.node_ancestors() or comparator is node)
)
):
inferred = utils.safe_infer(node.func)
if not isinstance(inferred, astroid.BoundMethod) or not isinstance(
inferred.bound, nodes.Dict
):
return
self.add_message(
"consider-iterating-dictionary", node=node, confidence=INFERENCE
)
def _check_use_maxsplit_arg(self, node: nodes.Call) -> None:
"""Add message when accessing first or last elements of a str.split() or
str.rsplit().
"""
# Check if call is split() or rsplit()
if not (
isinstance(node.func, nodes.Attribute)
and node.func.attrname in {"split", "rsplit"}
and isinstance(utils.safe_infer(node.func), astroid.BoundMethod)
):
return
inferred_expr = utils.safe_infer(node.func.expr)
if isinstance(inferred_expr, astroid.Instance) and any(
inferred_expr.nodes_of_class(nodes.ClassDef)
):
return
try:
sep = utils.get_argument_from_call(node, 0, "sep")
except utils.NoSuchArgumentError:
return
try:
# Ignore if maxsplit arg has been set
utils.get_argument_from_call(node, 1, "maxsplit")
return
except utils.NoSuchArgumentError:
pass
if isinstance(node.parent, nodes.Subscript):
try:
subscript_value = utils.get_subscript_const_value(node.parent).value
except utils.InferredTypeError:
return
# Check for cases where variable (Name) subscripts may be mutated within a loop
if isinstance(node.parent.slice, nodes.Name):
# Check if loop present within the scope of the node
scope = node.scope()
for loop_node in scope.nodes_of_class((nodes.For, nodes.While)):
if not loop_node.parent_of(node):
continue
# Check if var is mutated within loop (Assign/AugAssign)
for assignment_node in loop_node.nodes_of_class(nodes.AugAssign):
if node.parent.slice.name == assignment_node.target.name:
return
for assignment_node in loop_node.nodes_of_class(nodes.Assign):
if node.parent.slice.name in [
n.name for n in assignment_node.targets
]:
return
if subscript_value in (-1, 0):
fn_name = node.func.attrname
new_fn = "rsplit" if subscript_value == -1 else "split"
new_name = (
node.func.as_string().rsplit(fn_name, maxsplit=1)[0]
+ new_fn
+ f"({sep.as_string()}, maxsplit=1)[{subscript_value}]"
)
self.add_message("use-maxsplit-arg", node=node, args=(new_name,))
@utils.only_required_for_messages(
"consider-using-enumerate",
"consider-using-dict-items",
"use-sequence-for-iteration",
)
def visit_for(self, node: nodes.For) -> None:
self._check_consider_using_enumerate(node)
self._check_consider_using_dict_items(node)
self._check_use_sequence_for_iteration(node)
def _check_consider_using_enumerate(self, node: nodes.For) -> None:
"""Emit a convention whenever range and len are used for indexing."""
# Verify that we have a `range([start], len(...), [stop])` call and
# that the object which is iterated is used as a subscript in the
# body of the for.
# Is it a proper range call?
if not isinstance(node.iter, nodes.Call):
return
if not self._is_builtin(node.iter.func, "range"):
return
if not node.iter.args:
return
is_constant_zero = (
isinstance(node.iter.args[0], nodes.Const) and node.iter.args[0].value == 0
)
if len(node.iter.args) == 2 and not is_constant_zero:
return
if len(node.iter.args) > 2:
return
# Is it a proper len call?
if not isinstance(node.iter.args[-1], nodes.Call):
return
second_func = node.iter.args[-1].func
if not self._is_builtin(second_func, "len"):
return
len_args = node.iter.args[-1].args
if not len_args or len(len_args) != 1:
return
iterating_object = len_args[0]
if isinstance(iterating_object, nodes.Name):
expected_subscript_val_type = nodes.Name
elif isinstance(iterating_object, nodes.Attribute):
expected_subscript_val_type = nodes.Attribute
else:
return
# If we're defining __iter__ on self, enumerate won't work
scope = node.scope()
if (
isinstance(iterating_object, nodes.Name)
and iterating_object.name == "self"
and scope.name == "__iter__"
):
return
# Verify that the body of the for loop uses a subscript
# with the object that was iterated. This uses some heuristics
# in order to make sure that the same object is used in the
# for body.
for child in node.body:
for subscript in child.nodes_of_class(nodes.Subscript):
if not isinstance(subscript.value, expected_subscript_val_type):
continue
value = subscript.slice
if not isinstance(value, nodes.Name):
continue
if subscript.value.scope() != node.scope():
# Ignore this subscript if it's not in the same
# scope. This means that in the body of the for
# loop, another scope was created, where the same
# name for the iterating object was used.
continue
if value.name == node.target.name and (
isinstance(subscript.value, nodes.Name)
and iterating_object.name == subscript.value.name
or isinstance(subscript.value, nodes.Attribute)
and iterating_object.attrname == subscript.value.attrname
):
self.add_message("consider-using-enumerate", node=node)
return
def _check_consider_using_dict_items(self, node: nodes.For) -> None:
"""Add message when accessing dict values by index lookup."""
# Verify that we have a .keys() call and
# that the object which is iterated is used as a subscript in the
# body of the for.
iterating_object_name = utils.get_iterating_dictionary_name(node)
if iterating_object_name is None:
return
# Verify that the body of the for loop uses a subscript
# with the object that was iterated. This uses some heuristics
# in order to make sure that the same object is used in the
# for body.
for child in node.body:
for subscript in child.nodes_of_class(nodes.Subscript):
if not isinstance(subscript.value, (nodes.Name, nodes.Attribute)):
continue
value = subscript.slice
if (
not isinstance(value, nodes.Name)
or value.name != node.target.name
or iterating_object_name != subscript.value.as_string()
):
continue
last_definition_lineno = value.lookup(value.name)[1][-1].lineno
if last_definition_lineno > node.lineno:
# Ignore this subscript if it has been redefined after
# the for loop. This checks for the line number using .lookup()
# to get the line number where the iterating object was last
# defined and compare that to the for loop's line number
continue
if (
isinstance(subscript.parent, nodes.Assign)
and subscript in subscript.parent.targets
or isinstance(subscript.parent, nodes.AugAssign)
and subscript == subscript.parent.target
):
# Ignore this subscript if it is the target of an assignment
# Early termination as dict index lookup is necessary
return
self.add_message("consider-using-dict-items", node=node)
return
@utils.only_required_for_messages(
"consider-using-dict-items",
"use-sequence-for-iteration",
)
def visit_comprehension(self, node: nodes.Comprehension) -> None:
self._check_consider_using_dict_items_comprehension(node)
self._check_use_sequence_for_iteration(node)
def _check_consider_using_dict_items_comprehension(
self, node: nodes.Comprehension
) -> None:
"""Add message when accessing dict values by index lookup."""
iterating_object_name = utils.get_iterating_dictionary_name(node)
if iterating_object_name is None:
return
for child in node.parent.get_children():
for subscript in child.nodes_of_class(nodes.Subscript):
if not isinstance(subscript.value, (nodes.Name, nodes.Attribute)):
continue
value = subscript.slice
if (
not isinstance(value, nodes.Name)
or value.name != node.target.name
or iterating_object_name != subscript.value.as_string()
):
continue
self.add_message("consider-using-dict-items", node=node)
return
def _check_use_sequence_for_iteration(
self, node: nodes.For | nodes.Comprehension
) -> None:
"""Check if code iterates over an in-place defined set.
Sets using `*` are not considered in-place.
"""
if isinstance(node.iter, nodes.Set) and not any(
utils.has_starred_node_recursive(node)
):
self.add_message(
"use-sequence-for-iteration", node=node.iter, confidence=HIGH
)
@utils.only_required_for_messages("consider-using-f-string")
def visit_const(self, node: nodes.Const) -> None:
if self._py36_plus:
# f-strings require Python 3.6
if node.pytype() == "builtins.str" and not isinstance(
node.parent, nodes.JoinedStr
):
self._detect_replacable_format_call(node)
def _detect_replacable_format_call(self, node: nodes.Const) -> None:
"""Check whether a string is used in a call to format() or '%' and whether it
can be replaced by an f-string.
"""
if (
isinstance(node.parent, nodes.Attribute)
and node.parent.attrname == "format"
):
# Don't warn on referencing / assigning .format without calling it
if not isinstance(node.parent.parent, nodes.Call):
return
if node.parent.parent.args:
for arg in node.parent.parent.args:
# If star expressions with more than 1 element are being used
if isinstance(arg, nodes.Starred):
inferred = utils.safe_infer(arg.value)
if (
isinstance(inferred, astroid.List)
and len(inferred.elts) > 1
):
return
# Backslashes can't be in f-string expressions
if "\\" in arg.as_string():
return
elif node.parent.parent.keywords:
keyword_args = [
i[0] for i in utils.parse_format_method_string(node.value)[0]
]
for keyword in node.parent.parent.keywords:
# If keyword is used multiple times
if keyword_args.count(keyword.arg) > 1:
return
keyword = utils.safe_infer(keyword.value)
# If lists of more than one element are being unpacked
if isinstance(keyword, nodes.Dict):
if len(keyword.items) > 1 and len(keyword_args) > 1:
return
# If all tests pass, then raise message
self.add_message(
"consider-using-f-string",
node=node,
line=node.lineno,
col_offset=node.col_offset,
)
elif isinstance(node.parent, nodes.BinOp) and node.parent.op == "%":
# Backslashes can't be in f-string expressions
if "\\" in node.parent.right.as_string():
return
# If % applied to another type than str, it's modulo and can't be replaced by formatting
if not hasattr(node.parent.left, "value") or not isinstance(
node.parent.left.value, str
):
return
inferred_right = utils.safe_infer(node.parent.right)
# If dicts or lists of length > 1 are used
if isinstance(inferred_right, nodes.Dict):
if len(inferred_right.items) > 1:
return
elif isinstance(inferred_right, nodes.List):
if len(inferred_right.elts) > 1:
return
# If all tests pass, then raise message
self.add_message(
"consider-using-f-string",
node=node,
line=node.lineno,
col_offset=node.col_offset,
)
Hacked By AnonymousFox1.0, Coded By AnonymousFox