bugbear.py

import ast
import builtins
import itertools
import logging
import math
import re
import sys
from collections import namedtuple
from contextlib import suppress
from functools import lru_cache, partial
from keyword import iskeyword

import attr
import pycodestyle

__version__ = "23.1.20"

LOG = logging.getLogger("flake8.bugbear")
CONTEXTFUL_NODES = (
    ast.Module,
    ast.ClassDef,
    ast.AsyncFunctionDef,
    ast.FunctionDef,
    ast.Lambda,
    ast.ListComp,
    ast.SetComp,
    ast.DictComp,
    ast.GeneratorExp,
)
FUNCTION_NODES = (ast.AsyncFunctionDef, ast.FunctionDef, ast.Lambda)

Context = namedtuple("Context", ["node", "stack"])


@attr.s(hash=False)
class BugBearChecker:
    name = "flake8-bugbear"
    version = __version__

    tree = attr.ib(default=None)
    filename = attr.ib(default="(none)")
    lines = attr.ib(default=None)
    max_line_length = attr.ib(default=79)
    visitor = attr.ib(init=False, default=attr.Factory(lambda: BugBearVisitor))
    options = attr.ib(default=None)

    def run(self):
        if not self.tree or not self.lines:
            self.load_file()

        if self.options and hasattr(self.options, "extend_immutable_calls"):
            b008_extend_immutable_calls = set(self.options.extend_immutable_calls)
        else:
            b008_extend_immutable_calls = set()

        visitor = self.visitor(
            filename=self.filename,
            lines=self.lines,
            b008_extend_immutable_calls=b008_extend_immutable_calls,
        )
        visitor.visit(self.tree)
        for e in itertools.chain(visitor.errors, self.gen_line_based_checks()):
            if self.should_warn(e.message[:4]):
                yield self.adapt_error(e)

    def gen_line_based_checks(self):
        """gen_line_based_checks() -> (error, error, error, ...)

        The following simple checks are based on the raw lines, not the AST.
        """
        for lineno, line in enumerate(self.lines, start=1):
            # Special case: ignore long shebang (following pycodestyle).
            if lineno == 1 and line.startswith("#!"):
                continue

            length = len(line) - 1
            if length > 1.1 * self.max_line_length and line.strip():
                # Special case long URLS and paths to follow pycodestyle.
                # Would use the `pycodestyle.maximum_line_length` directly but
                # need to supply it arguments that are not available so chose
                # to replicate instead.
                chunks = line.split()

                is_line_comment_url_path = len(chunks) == 2 and chunks[0] == "#"

                just_long_url_path = len(chunks) == 1

                num_leading_whitespaces = len(line) - len(chunks[-1])
                too_many_leading_white_spaces = (
                    num_leading_whitespaces >= self.max_line_length - 7
                )

                skip = is_line_comment_url_path or just_long_url_path
                if skip and not too_many_leading_white_spaces:
                    continue

                yield B950(lineno, length, vars=(length, self.max_line_length))

    @classmethod
    def adapt_error(cls, e):
        """Adapts the extended error namedtuple to be compatible with Flake8."""
        return e._replace(message=e.message.format(*e.vars))[:4]

    def load_file(self):
        """Loads the file in a way that auto-detects source encoding and deals
        with broken terminal encodings for stdin.

        Stolen from flake8_import_order because it's good.
        """

        if self.filename in ("stdin", "-", None):
            self.filename = "stdin"
            self.lines = pycodestyle.stdin_get_value().splitlines(True)
        else:
            self.lines = pycodestyle.readlines(self.filename)

        if not self.tree:
            self.tree = ast.parse("".join(self.lines))

    @staticmethod
    def add_options(optmanager):
        """Informs flake8 to ignore B9xx by default."""
        optmanager.extend_default_ignore(disabled_by_default)
        optmanager.add_option(
            "--extend-immutable-calls",
            comma_separated_list=True,
            parse_from_config=True,
            default=[],
            help="Skip B008 test for additional immutable calls.",
        )

    @lru_cache()  # noqa: B019
    def should_warn(self, code):
        """Returns `True` if Bugbear should emit a particular warning.

        flake8 overrides default ignores when the user specifies
        `ignore = ` in configuration.  This is problematic because it means
        specifying anything in `ignore = ` implicitly enables all optional
        warnings.  This function is a workaround for this behavior.

        As documented in the README, the user is expected to explicitly select
        the warnings.

        NOTE: This method is deprecated and will be removed in a future release. It is
        recommended to use `extend-ignore` and `extend-select` in your flake8
        configuration to avoid implicitly altering selected and ignored codes.
        """
        if code[:2] != "B9":
            # Normal warnings are safe for emission.
            return True

        if self.options is None:
            # Without options configured, Bugbear will emit B9 but flake8 will ignore
            LOG.info(
                "Options not provided to Bugbear, optional warning %s selected.", code
            )
            return True

        for i in range(2, len(code) + 1):
            if self.options.select and code[:i] in self.options.select:
                return True

            # flake8 >=4.0: Also check for codes in extend_select
            if (
                hasattr(self.options, "extend_select")
                and self.options.extend_select
                and code[:i] in self.options.extend_select
            ):
                return True

        LOG.info(
            (
                "Optional warning %s not present in selected warnings: %r. Not "
                "firing it at all."
            ),
            code,
            self.options.select,
        )
        return False


def _is_identifier(arg):
    # Return True if arg is a valid identifier, per
    # https://docs.python.org/2/reference/lexical_analysis.html#identifiers

    if not isinstance(arg, ast.Str):
        return False

    return re.match(r"^[A-Za-z_][A-Za-z0-9_]*$", arg.s) is not None


def _to_name_str(node):
    # Turn Name and Attribute nodes to strings, e.g "ValueError" or
    # "pkg.mod.error", handling any depth of attribute accesses.
    if isinstance(node, ast.Name):
        return node.id
    if isinstance(node, ast.Call):
        return _to_name_str(node.func)
    try:
        return _to_name_str(node.value) + "." + node.attr
    except AttributeError:
        return _to_name_str(node.value)


def names_from_assignments(assign_target):
    if isinstance(assign_target, ast.Name):
        yield assign_target.id
    elif isinstance(assign_target, ast.Starred):
        yield from names_from_assignments(assign_target.value)
    elif isinstance(assign_target, (ast.List, ast.Tuple)):
        for child in assign_target.elts:
            yield from names_from_assignments(child)


def children_in_scope(node):
    yield node
    if not isinstance(node, FUNCTION_NODES):
        for child in ast.iter_child_nodes(node):
            yield from children_in_scope(child)


def _typesafe_issubclass(cls, class_or_tuple):
    try:
        return issubclass(cls, class_or_tuple)
    except TypeError:
        # User code specifies a type that is not a type in our current run. Might be
        # their error, might be a difference in our environments. We don't know so we
        # ignore this
        return False


@attr.s
class BugBearVisitor(ast.NodeVisitor):
    filename = attr.ib()
    lines = attr.ib()
    b008_extend_immutable_calls = attr.ib(default=attr.Factory(set))
    node_stack = attr.ib(default=attr.Factory(list))
    node_window = attr.ib(default=attr.Factory(list))
    errors = attr.ib(default=attr.Factory(list))
    futures = attr.ib(default=attr.Factory(set))
    contexts = attr.ib(default=attr.Factory(list))

    NODE_WINDOW_SIZE = 4
    _b023_seen = attr.ib(factory=set, init=False)

    if False:
        # Useful for tracing what the hell is going on.

        def __getattr__(self, name):
            print(name)
            return self.__getattribute__(name)

    @property
    def node_stack(self):
        if len(self.contexts) == 0:
            return []

        context, stack = self.contexts[-1]
        return stack

    def visit(self, node):
        is_contextful = isinstance(node, CONTEXTFUL_NODES)

        if is_contextful:
            context = Context(node, [])
            self.contexts.append(context)

        self.node_stack.append(node)
        self.node_window.append(node)
        self.node_window = self.node_window[-self.NODE_WINDOW_SIZE :]
        super().visit(node)
        self.node_stack.pop()

        if is_contextful:
            self.contexts.pop()

    def visit_ExceptHandler(self, node):
        if node.type is None:
            self.errors.append(
                B001(node.lineno, node.col_offset, vars=("bare `except:`",))
            )
        elif isinstance(node.type, ast.Tuple):
            names = [_to_name_str(e) for e in node.type.elts]
            as_ = " as " + node.name if node.name is not None else ""
            if len(names) == 0:
                vs = (f"`except (){as_}:`",)
                self.errors.append(B001(node.lineno, node.col_offset, vars=vs))
            elif len(names) == 1:
                self.errors.append(B013(node.lineno, node.col_offset, vars=names))
            else:
                # See if any of the given exception names could be removed, e.g. from:
                #    (MyError, MyError)  # duplicate names
                #    (MyError, BaseException)  # everything derives from the Base
                #    (Exception, TypeError)  # builtins where one subclasses another
                #    (IOError, OSError)  # IOError is an alias of OSError since Python3.3
                # but note that other cases are impractical to handle from the AST.
                # We expect this is mostly useful for users who do not have the
                # builtin exception hierarchy memorised, and include a 'shadowed'
                # subtype without realising that it's redundant.
                good = sorted(set(names), key=names.index)
                if "BaseException" in good:
                    good = ["BaseException"]
                # Remove redundant exceptions that the automatic system either handles
                # poorly (usually aliases) or can't be checked (e.g. it's not an
                # built-in exception).
                for primary, equivalents in B014.redundant_exceptions.items():
                    if primary in good:
                        good = [g for g in good if g not in equivalents]

                for name, other in itertools.permutations(tuple(good), 2):
                    if _typesafe_issubclass(
                        getattr(builtins, name, type), getattr(builtins, other, ())
                    ):
                        if name in good:
                            good.remove(name)
                if good != names:
                    desc = good[0] if len(good) == 1 else "({})".format(", ".join(good))
                    self.errors.append(
                        B014(
                            node.lineno,
                            node.col_offset,
                            vars=(", ".join(names), as_, desc),
                        )
                    )
        self.generic_visit(node)

    def visit_UAdd(self, node):
        trailing_nodes = list(map(type, self.node_window[-4:]))
        if trailing_nodes == [ast.UnaryOp, ast.UAdd, ast.UnaryOp, ast.UAdd]:
            originator = self.node_window[-4]
            self.errors.append(B002(originator.lineno, originator.col_offset))
        self.generic_visit(node)

    def visit_Call(self, node):
        if isinstance(node.func, ast.Attribute):
            self.check_for_b005(node)
        else:
            with suppress(AttributeError, IndexError):
                if (
                    node.func.id in ("getattr", "hasattr")
                    and node.args[1].s == "__call__"
                ):
                    self.errors.append(B004(node.lineno, node.col_offset))
                if (
                    node.func.id == "getattr"
                    and len(node.args) == 2
                    and _is_identifier(node.args[1])
                    and not iskeyword(node.args[1].s)
                ):
                    self.errors.append(B009(node.lineno, node.col_offset))
                elif (
                    not any(isinstance(n, ast.Lambda) for n in self.node_stack)
                    and node.func.id == "setattr"
                    and len(node.args) == 3
                    and _is_identifier(node.args[1])
                    and not iskeyword(node.args[1].s)
                ):
                    self.errors.append(B010(node.lineno, node.col_offset))

            self.check_for_b026(node)

        self.check_for_b905(node)
        self.generic_visit(node)

    def visit_Module(self, node):
        self.check_for_b018(node)
        self.generic_visit(node)

    def visit_Assign(self, node):
        if len(node.targets) == 1:
            t = node.targets[0]
            if isinstance(t, ast.Attribute) and isinstance(t.value, ast.Name):
                if (t.value.id, t.attr) == ("os", "environ"):
                    self.errors.append(B003(node.lineno, node.col_offset))
        self.generic_visit(node)

    def visit_For(self, node):
        self.check_for_b007(node)
        self.check_for_b020(node)
        self.check_for_b023(node)
        self.generic_visit(node)

    def visit_AsyncFor(self, node):
        self.check_for_b023(node)
        self.generic_visit(node)

    def visit_While(self, node):
        self.check_for_b023(node)
        self.generic_visit(node)

    def visit_ListComp(self, node):
        self.check_for_b023(node)
        self.generic_visit(node)

    def visit_SetComp(self, node):
        self.check_for_b023(node)
        self.generic_visit(node)

    def visit_DictComp(self, node):
        self.check_for_b023(node)
        self.generic_visit(node)

    def visit_GeneratorExp(self, node):
        self.check_for_b023(node)
        self.generic_visit(node)

    def visit_Assert(self, node):
        self.check_for_b011(node)
        self.generic_visit(node)

    def visit_AsyncFunctionDef(self, node):
        self.check_for_b902(node)
        self.check_for_b006_and_b008(node)
        self.generic_visit(node)

    def visit_FunctionDef(self, node):
        self.check_for_b901(node)
        self.check_for_b902(node)
        self.check_for_b006_and_b008(node)
        self.check_for_b018(node)
        self.check_for_b019(node)
        self.check_for_b021(node)
        self.check_for_b906(node)
        self.generic_visit(node)

    def visit_ClassDef(self, node):
        self.check_for_b903(node)
        self.check_for_b018(node)
        self.check_for_b021(node)
        self.check_for_b024_and_b027(node)
        self.generic_visit(node)

    def visit_Try(self, node):
        self.check_for_b012(node)
        self.check_for_b025(node)
        self.generic_visit(node)

    def visit_Compare(self, node):
        self.check_for_b015(node)
        self.generic_visit(node)

    def visit_Raise(self, node):
        self.check_for_b016(node)
        self.check_for_b904(node)
        self.generic_visit(node)

    def visit_With(self, node):
        self.check_for_b017(node)
        self.check_for_b022(node)
        self.generic_visit(node)

    def visit_JoinedStr(self, node):
        self.check_for_b907(node)
        self.generic_visit(node)

    def check_for_b005(self, node):
        if node.func.attr not in B005.methods:
            return  # method name doesn't match

        if len(node.args) != 1 or not isinstance(node.args[0], ast.Str):
            return  # used arguments don't match the builtin strip

        call_path = ".".join(compose_call_path(node.func.value))
        if call_path in B005.valid_paths:
            return  # path is exempt

        s = node.args[0].s
        if len(s) == 1:
            return  # stripping just one character

        if len(s) == len(set(s)):
            return  # no characters appear more than once

        self.errors.append(B005(node.lineno, node.col_offset))

    def check_for_b006_and_b008(self, node):
        visitor = FuntionDefDefaultsVisitor(self.b008_extend_immutable_calls)
        visitor.visit(node.args.defaults + node.args.kw_defaults)
        self.errors.extend(visitor.errors)

    def check_for_b007(self, node):
        targets = NameFinder()
        targets.visit(node.target)
        ctrl_names = set(filter(lambda s: not s.startswith("_"), targets.names))
        body = NameFinder()
        for expr in node.body:
            body.visit(expr)
        used_names = set(body.names)
        for name in sorted(ctrl_names - used_names):
            n = targets.names[name][0]
            self.errors.append(B007(n.lineno, n.col_offset, vars=(name,)))

    def check_for_b011(self, node):
        if isinstance(node.test, ast.NameConstant) and node.test.value is False:
            self.errors.append(B011(node.lineno, node.col_offset))

    def check_for_b012(self, node):
        def _loop(node, bad_node_types):
            if isinstance(node, (ast.AsyncFunctionDef, ast.FunctionDef)):
                return

            if isinstance(node, (ast.While, ast.For)):
                bad_node_types = (ast.Return,)

            elif isinstance(node, bad_node_types):
                self.errors.append(B012(node.lineno, node.col_offset))

            for child in ast.iter_child_nodes(node):
                _loop(child, bad_node_types)

        for child in node.finalbody:
            _loop(child, (ast.Return, ast.Continue, ast.Break))

    def check_for_b015(self, node):
        if isinstance(self.node_stack[-2], ast.Expr):
            self.errors.append(B015(node.lineno, node.col_offset))

    def check_for_b016(self, node):
        if isinstance(node.exc, (ast.NameConstant, ast.Num, ast.Str, ast.JoinedStr)):
            self.errors.append(B016(node.lineno, node.col_offset))

    def check_for_b017(self, node):
        """Checks for use of the evil syntax 'with assertRaises(Exception):'
        or 'with pytest.raises(Exception)'.

        This form of assertRaises will catch everything that subclasses
        Exception, which happens to be the vast majority of Python internal
        errors, including the ones raised when a non-existing method/function
        is called, or a function is called with an invalid dictionary key
        lookup.
        """
        item = node.items[0]
        item_context = item.context_expr

        if (
            hasattr(item_context, "func")
            and isinstance(item_context.func, ast.Attribute)
            and (
                item_context.func.attr == "assertRaises"
                or (
                    item_context.func.attr == "raises"
                    and isinstance(item_context.func.value, ast.Name)
                    and item_context.func.value.id == "pytest"
                    and "match" not in [kwd.arg for kwd in item_context.keywords]
                )
            )
            and len(item_context.args) == 1
            and isinstance(item_context.args[0], ast.Name)
            and item_context.args[0].id == "Exception"
            and not item.optional_vars
        ):
            self.errors.append(B017(node.lineno, node.col_offset))

    def check_for_b019(self, node):
        if (
            len(node.decorator_list) == 0
            or len(self.contexts) < 2
            or not isinstance(self.contexts[-2].node, ast.ClassDef)
        ):
            return

        # Preserve decorator order so we can get the lineno from the decorator node
        # rather than the function node (this location definition changes in Python 3.8)
        resolved_decorators = (
            ".".join(compose_call_path(decorator)) for decorator in node.decorator_list
        )
        for idx, decorator in enumerate(resolved_decorators):
            if decorator in {"classmethod", "staticmethod"}:
                return

            if decorator in B019.caches:
                self.errors.append(
                    B019(
                        node.decorator_list[idx].lineno,
                        node.decorator_list[idx].col_offset,
                    )
                )
                return

    def check_for_b020(self, node):
        targets = NameFinder()
        targets.visit(node.target)
        ctrl_names = set(targets.names)

        iterset = B020NameFinder()
        iterset.visit(node.iter)
        iterset_names = set(iterset.names)

        for name in sorted(ctrl_names):
            if name in iterset_names:
                n = targets.names[name][0]
                self.errors.append(B020(n.lineno, n.col_offset, vars=(name,)))

    def check_for_b023(self, loop_node):  # noqa: C901
        """Check that functions (including lambdas) do not use loop variables.

        https://docs.python-guide.org/writing/gotchas/#late-binding-closures from
        functions - usually but not always lambdas - defined inside a loop are a
        classic source of bugs.

        For each use of a variable inside a function defined inside a loop, we
        emit a warning if that variable is reassigned on each loop iteration
        (outside the function).  This includes but is not limited to explicit
        loop variables like the `x` in `for x in range(3):`.
        """
        # Because most loops don't contain functions, it's most efficient to
        # implement this "backwards": first we find all the candidate variable
        # uses, and then if there are any we check for assignment of those names
        # inside the loop body.
        safe_functions = []
        suspicious_variables = []
        for node in ast.walk(loop_node):
            # check if function is immediately consumed to avoid false alarm
            if isinstance(node, ast.Call):
                # check for filter&reduce
                if (
                    isinstance(node.func, ast.Name)
                    and node.func.id in ("filter", "reduce", "map")
                ) or (
                    isinstance(node.func, ast.Attribute)
                    and node.func.attr == "reduce"
                    and isinstance(node.func.value, ast.Name)
                    and node.func.value.id == "functools"
                ):
                    for arg in node.args:
                        if isinstance(arg, FUNCTION_NODES):
                            safe_functions.append(arg)

                # check for key=
                for keyword in node.keywords:
                    if keyword.arg == "key" and isinstance(
                        keyword.value, FUNCTION_NODES
                    ):
                        safe_functions.append(keyword.value)

            # mark `return lambda: x` as safe
            # does not (currently) check inner lambdas in a returned expression
            # e.g. `return (lambda: x, )
            if isinstance(node, ast.Return):
                if isinstance(node.value, FUNCTION_NODES):
                    safe_functions.append(node.value)

            # find unsafe functions
            if isinstance(node, FUNCTION_NODES) and node not in safe_functions:
                argnames = {
                    arg.arg for arg in ast.walk(node.args) if isinstance(arg, ast.arg)
                }
                if isinstance(node, ast.Lambda):
                    body_nodes = ast.walk(node.body)
                else:
                    body_nodes = itertools.chain.from_iterable(map(ast.walk, node.body))
                errors = []
                for name in body_nodes:
                    if isinstance(name, ast.Name) and name.id not in argnames:
                        if isinstance(name.ctx, ast.Load):
                            errors.append(
                                B023(name.lineno, name.col_offset, vars=(name.id,))
                            )
                        elif isinstance(name.ctx, ast.Store):
                            argnames.add(name.id)
                for err in errors:
                    if err.vars[0] not in argnames and err not in self._b023_seen:
                        self._b023_seen.add(err)  # dedupe across nested loops
                        suspicious_variables.append(err)

        if suspicious_variables:
            reassigned_in_loop = set(self._get_assigned_names(loop_node))

        for err in sorted(suspicious_variables):
            if reassigned_in_loop.issuperset(err.vars):
                self.errors.append(err)

    def check_for_b024_and_b027(self, node: ast.ClassDef):  # noqa: C901
        """Check for inheritance from abstract classes in abc and lack of
        any methods decorated with abstract*"""

        def is_abc_class(value, name="ABC"):
            # class foo(metaclass = [abc.]ABCMeta)
            if isinstance(value, ast.keyword):
                return value.arg == "metaclass" and is_abc_class(value.value, "ABCMeta")
            # class foo(ABC)
            # class foo(abc.ABC)
            return (isinstance(value, ast.Name) and value.id == name) or (
                isinstance(value, ast.Attribute)
                and value.attr == name
                and isinstance(value.value, ast.Name)
                and value.value.id == "abc"
            )

        def is_abstract_decorator(expr):
            return (isinstance(expr, ast.Name) and expr.id[:8] == "abstract") or (
                isinstance(expr, ast.Attribute) and expr.attr[:8] == "abstract"
            )

        def is_overload(expr):
            return (isinstance(expr, ast.Name) and expr.id == "overload") or (
                isinstance(expr, ast.Attribute) and expr.attr == "overload"
            )

        def empty_body(body) -> bool:
            def is_str_or_ellipsis(node):
                # ast.Ellipsis and ast.Str used in python<3.8
                return isinstance(node, (ast.Ellipsis, ast.Str)) or (
                    isinstance(node, ast.Constant)
                    and (node.value is Ellipsis or isinstance(node.value, str))
                )

            # Function body consist solely of `pass`, `...`, and/or (doc)string literals
            return all(
                isinstance(stmt, ast.Pass)
                or (isinstance(stmt, ast.Expr) and is_str_or_ellipsis(stmt.value))
                for stmt in body
            )

        # don't check multiple inheritance
        # https://github.com/PyCQA/flake8-bugbear/issues/277
        if len(node.bases) + len(node.keywords) > 1:
            return

        # only check abstract classes
        if not any(map(is_abc_class, (*node.bases, *node.keywords))):
            return

        has_method = False
        has_abstract_method = False

        for stmt in node.body:
            # https://github.com/PyCQA/flake8-bugbear/issues/293
            # Ignore abc's that declares a class attribute that must be set
            if isinstance(stmt, (ast.AnnAssign, ast.Assign)):
                has_abstract_method = True
                continue

            # only check function defs
            if not isinstance(stmt, (ast.FunctionDef, ast.AsyncFunctionDef)):
                continue
            has_method = True

            has_abstract_decorator = any(
                map(is_abstract_decorator, stmt.decorator_list)
            )

            has_abstract_method |= has_abstract_decorator

            if (
                not has_abstract_decorator
                and empty_body(stmt.body)
                and not any(map(is_overload, stmt.decorator_list))
            ):
                self.errors.append(
                    B027(stmt.lineno, stmt.col_offset, vars=(stmt.name,))
                )

        if has_method and not has_abstract_method:
            self.errors.append(B024(node.lineno, node.col_offset, vars=(node.name,)))

    def check_for_b026(self, call: ast.Call):
        if not call.keywords:
            return

        starreds = [arg for arg in call.args if isinstance(arg, ast.Starred)]
        if not starreds:
            return

        first_keyword = call.keywords[0].value
        for starred in starreds:
            if (starred.lineno, starred.col_offset) > (
                first_keyword.lineno,
                first_keyword.col_offset,
            ):
                self.errors.append(B026(starred.lineno, starred.col_offset))

    def _get_assigned_names(self, loop_node):
        loop_targets = (ast.For, ast.AsyncFor, ast.comprehension)
        for node in children_in_scope(loop_node):
            if isinstance(node, (ast.Assign)):
                for child in node.targets:
                    yield from names_from_assignments(child)
            if isinstance(node, loop_targets + (ast.AnnAssign, ast.AugAssign)):
                yield from names_from_assignments(node.target)

    def check_for_b904(self, node):
        """Checks `raise` without `from` inside an `except` clause.

        In these cases, you should use explicit exception chaining from the
        earlier error, or suppress it with `raise ... from None`.  See
        https://docs.python.org/3/tutorial/errors.html#exception-chaining
        """
        if (
            node.cause is None
            and node.exc is not None
            and not (isinstance(node.exc, ast.Name) and node.exc.id.islower())
            and any(isinstance(n, ast.ExceptHandler) for n in self.node_stack)
        ):
            self.errors.append(B904(node.lineno, node.col_offset))

    def walk_function_body(self, node):
        def _loop(parent, node):
            if isinstance(node, (ast.AsyncFunctionDef, ast.FunctionDef)):
                return
            yield parent, node
            for child in ast.iter_child_nodes(node):
                yield from _loop(node, child)

        for child in node.body:
            yield from _loop(node, child)

    def check_for_b901(self, node):
        if node.name == "__await__":
            return

        has_yield = False
        return_node = None

        for parent, x in self.walk_function_body(node):
            # Only consider yield when it is part of an Expr statement.
            if isinstance(parent, ast.Expr) and isinstance(
                x, (ast.Yield, ast.YieldFrom)
            ):
                has_yield = True

            if isinstance(x, ast.Return) and x.value is not None:
                return_node = x

            if has_yield and return_node is not None:
                self.errors.append(B901(return_node.lineno, return_node.col_offset))
                break

    def check_for_b902(self, node):
        if len(self.contexts) < 2 or not isinstance(
            self.contexts[-2].node, ast.ClassDef
        ):
            return

        cls = self.contexts[-2].node
        decorators = NameFinder()
        decorators.visit(node.decorator_list)

        if "staticmethod" in decorators.names:
            # TODO: maybe warn if the first argument is surprisingly `self` or
            # `cls`?
            return

        bases = {b.id for b in cls.bases if isinstance(b, ast.Name)}
        if "type" in bases:
            if (
                "classmethod" in decorators.names
                or node.name in B902.implicit_classmethods
            ):
                expected_first_args = B902.metacls
                kind = "metaclass class"
            else:
                expected_first_args = B902.cls
                kind = "metaclass instance"
        else:
            if (
                "classmethod" in decorators.names
                or node.name in B902.implicit_classmethods
            ):
                expected_first_args = B902.cls
                kind = "class"
            else:
                expected_first_args = B902.self
                kind = "instance"

        args = getattr(node.args, "posonlyargs", []) + node.args.args
        vararg = node.args.vararg
        kwarg = node.args.kwarg
        kwonlyargs = node.args.kwonlyargs

        if args:
            actual_first_arg = args[0].arg
            lineno = args[0].lineno
            col = args[0].col_offset
        elif vararg:
            actual_first_arg = "*" + vararg.arg
            lineno = vararg.lineno
            col = vararg.col_offset
        elif kwarg:
            actual_first_arg = "**" + kwarg.arg
            lineno = kwarg.lineno
            col = kwarg.col_offset
        elif kwonlyargs:
            actual_first_arg = "*, " + kwonlyargs[0].arg
            lineno = kwonlyargs[0].lineno
            col = kwonlyargs[0].col_offset
        else:
            actual_first_arg = "(none)"
            lineno = node.lineno
            col = node.col_offset

        if actual_first_arg not in expected_first_args:
            if not actual_first_arg.startswith(("(", "*")):
                actual_first_arg = repr(actual_first_arg)
            self.errors.append(
                B902(lineno, col, vars=(actual_first_arg, kind, expected_first_args[0]))
            )

    def check_for_b903(self, node):
        body = node.body
        if (
            body
            and isinstance(body[0], ast.Expr)
            and isinstance(body[0].value, ast.Str)
        ):
            # Ignore the docstring
            body = body[1:]

        if (
            len(body) != 1
            or not isinstance(body[0], ast.FunctionDef)
            or body[0].name != "__init__"
        ):
            # only classes with *just* an __init__ method are interesting
            return

        # all the __init__ function does is a series of assignments to attributes
        for stmt in body[0].body:
            if not isinstance(stmt, ast.Assign):
                return
            targets = stmt.targets
            if len(targets) > 1 or not isinstance(targets[0], ast.Attribute):
                return
            if not isinstance(stmt.value, ast.Name):
                return

        self.errors.append(B903(node.lineno, node.col_offset))

    def check_for_b018(self, node):
        for subnode in node.body:
            if not isinstance(subnode, ast.Expr):
                continue
            if isinstance(
                subnode.value,
                (
                    ast.Num,
                    ast.Bytes,
                    ast.NameConstant,
                    ast.List,
                    ast.Set,
                    ast.Dict,
                ),
            ):
                self.errors.append(B018(subnode.lineno, subnode.col_offset))

    def check_for_b021(self, node):
        if (
            node.body
            and isinstance(node.body[0], ast.Expr)
            and isinstance(node.body[0].value, ast.JoinedStr)
        ):
            self.errors.append(
                B021(node.body[0].value.lineno, node.body[0].value.col_offset)
            )

    def check_for_b022(self, node):
        item = node.items[0]
        item_context = item.context_expr
        if (
            hasattr(item_context, "func")
            and hasattr(item_context.func, "value")
            and hasattr(item_context.func.value, "id")
            and item_context.func.value.id == "contextlib"
            and hasattr(item_context.func, "attr")
            and item_context.func.attr == "suppress"
            and len(item_context.args) == 0
        ):
            self.errors.append(B022(node.lineno, node.col_offset))

    def check_for_b025(self, node):
        seen = []
        for handler in node.handlers:
            if isinstance(handler.type, (ast.Name, ast.Attribute)):
                name = ".".join(compose_call_path(handler.type))
                seen.append(name)
            elif isinstance(handler.type, ast.Tuple):
                # to avoid checking the same as B014, remove duplicates per except
                uniques = set()
                for entry in handler.type.elts:
                    name = ".".join(compose_call_path(entry))
                    uniques.add(name)
                seen.extend(uniques)
        # sort to have a deterministic output
        duplicates = sorted({x for x in seen if seen.count(x) > 1})
        for duplicate in duplicates:
            self.errors.append(B025(node.lineno, node.col_offset, vars=(duplicate,)))

    def check_for_b905(self, node):
        if (
            isinstance(node.func, ast.Name)
            and node.func.id == "zip"
            and not any(kw.arg == "strict" for kw in node.keywords)
        ):
            self.errors.append(B905(node.lineno, node.col_offset))

    def check_for_b906(self, node: ast.FunctionDef):
        if not node.name.startswith("visit_"):
            return

        # extract what's visited
        class_name = node.name[len("visit_") :]
        class_type = getattr(ast, class_name, None)

        if (
            # not a valid ast subclass
            class_type is None
            # doesn't have a non-empty '_fields' attribute - which is what's
            # iterated over in ast.NodeVisitor.generic_visit
            or not getattr(class_type, "_fields", None)
            # or can't contain any ast subnodes that could be visited
            # See https://docs.python.org/3/library/ast.html#abstract-grammar
            or class_type.__name__
            in (
                "alias",
                "Constant",
                "Global",
                "MatchSingleton",
                "MatchStar",
                "Nonlocal",
                "TypeIgnore",
                # These ast nodes are deprecated, but some codebases may still use them
                # for backwards-compatibility with Python 3.7
                "Bytes",
                "Num",
                "Str",
            )
        ):
            return

        for n in itertools.chain.from_iterable(ast.walk(nn) for nn in node.body):
            if isinstance(n, ast.Call) and (
                (isinstance(n.func, ast.Attribute) and "visit" in n.func.attr)
                or (isinstance(n.func, ast.Name) and "visit" in n.func.id)
            ):
                break
        else:
            self.errors.append(B906(node.lineno, node.col_offset))

    def check_for_b907(self, node: ast.JoinedStr):  # noqa: C901
        # AST structure of strings in f-strings in 3.7 is different enough this
        # implementation doesn't work
        if sys.version_info <= (3, 7):
            return  # pragma: no cover

        def myunparse(node: ast.AST) -> str:  # pragma: no cover
            if sys.version_info >= (3, 9):
                return ast.unparse(node)
            if isinstance(node, ast.Name):
                return node.id
            if isinstance(node, ast.Attribute):
                return myunparse(node.value) + "." + node.attr
            if isinstance(node, ast.Constant):
                return repr(node.value)
            if isinstance(node, ast.Call):
                return myunparse(node.func) + "()"  # don't bother with arguments

            # as a last resort, just give the type name
            return type(node).__name__

        quote_marks = "'\""
        current_mark = None
        variable = None
        for value in node.values:
            # check for quote mark after pre-marked variable
            if (
                current_mark is not None
                and variable is not None
                and isinstance(value, ast.Constant)
                and isinstance(value.value, str)
                and value.value[0] == current_mark
            ):
                self.errors.append(
                    B907(
                        variable.lineno,
                        variable.col_offset,
                        vars=(myunparse(variable.value),),
                    )
                )
                current_mark = variable = None
                # don't continue with length>1, so we can detect a new pre-mark
                # in the same string as a post-mark, e.g. `"{foo}" "{bar}"`
                if len(value.value) == 1:
                    continue

            # detect pre-mark
            if (
                isinstance(value, ast.Constant)
                and isinstance(value.value, str)
                and value.value[-1] in quote_marks
            ):
                current_mark = value.value[-1]
                variable = None
                continue

            # detect variable, if there's a pre-mark
            if (
                current_mark is not None
                and variable is None
                and isinstance(value, ast.FormattedValue)
                and value.conversion != ord("r")
            ):
                # check if the format spec shows that this is numeric
                # or otherwise hard/impossible to convert to `!r`
                if (
                    isinstance(value.format_spec, ast.JoinedStr)
                    and value.format_spec.values  # empty format spec is fine
                ):
                    if (
                        # if there's variables in the format_spec, skip
                        len(value.format_spec.values) > 1
                        or not isinstance(value.format_spec.values[0], ast.Constant)
                    ):
                        current_mark = variable = None
                        continue
                    format_specifier = value.format_spec.values[0].value

                    # if second character is an align, first character is a fill
                    # char - strip it
                    if len(format_specifier) > 1 and format_specifier[1] in "<>=^":
                        format_specifier = format_specifier[1:]

                    # strip out precision digits, so the only remaining ones are
                    # width digits, which will misplace the quotes
                    format_specifier = re.sub(r"\.\d*", "", format_specifier)

                    # skip if any invalid characters in format spec
                    invalid_characters = "".join(
                        (
                            "=",  # align character only valid for numerics
                            "+- ",  # sign
                            "0123456789",  # width digits
                            "z",  # coerce negative zero floating point to positive
                            "#",  # alternate form
                            "_,",  # thousands grouping
                            "bcdeEfFgGnoxX%",  # various number specifiers
                        )
                    )
                    if set(format_specifier).intersection(invalid_characters):
                        current_mark = variable = None
                        continue

                # otherwise save value as variable
                variable = value
                continue

            # if no pre-mark or variable detected, reset state
            current_mark = variable = None


def compose_call_path(node):
    if isinstance(node, ast.Attribute):
        yield from compose_call_path(node.value)
        yield node.attr
    elif isinstance(node, ast.Call):
        yield from compose_call_path(node.func)
    elif isinstance(node, ast.Name):
        yield node.id


@attr.s
class NameFinder(ast.NodeVisitor):
    """Finds a name within a tree of nodes.

    After `.visit(node)` is called, `found` is a dict with all name nodes inside,
    key is name string, value is the node (useful for location purposes).
    """

    names = attr.ib(default=attr.Factory(dict))

    def visit_Name(self, node):  # noqa: B906 # names don't contain other names
        self.names.setdefault(node.id, []).append(node)

    def visit(self, node):
        """Like super-visit but supports iteration over lists."""
        if not isinstance(node, list):
            return super().visit(node)

        for elem in node:
            super().visit(elem)
        return node


class FuntionDefDefaultsVisitor(ast.NodeVisitor):
    def __init__(self, b008_extend_immutable_calls=None):
        self.b008_extend_immutable_calls = b008_extend_immutable_calls or set()
        for node in B006.mutable_literals + B006.mutable_comprehensions:
            setattr(self, f"visit_{node}", self.visit_mutable_literal_or_comprehension)
        self.errors = []
        self.arg_depth = 0
        super().__init__()

    def visit_mutable_literal_or_comprehension(self, node):
        # Flag B006 iff mutable literal/comprehension is not nested.
        # We only flag these at the top level of the expression as we
        # cannot easily guarantee that nested mutable structures are not
        # made immutable by outer operations, so we prefer no false positives.
        # e.g.
        # >>> def this_is_fine(a=frozenset({"a", "b", "c"})): ...
        #
        # >>> def this_is_not_fine_but_hard_to_detect(a=(lambda x: x)([1, 2, 3]))
        #
        # We do still search for cases of B008 within mutable structures though.
        if self.arg_depth == 1:
            self.errors.append(B006(node.lineno, node.col_offset))
        # Check for nested functions.
        self.generic_visit(node)

    def visit_Call(self, node):
        call_path = ".".join(compose_call_path(node.func))
        if call_path in B006.mutable_calls:
            self.errors.append(B006(node.lineno, node.col_offset))
            self.generic_visit(node)
            return

        if call_path in B008.immutable_calls | self.b008_extend_immutable_calls:
            self.generic_visit(node)
            return

        # Check if function call is actually a float infinity/NaN literal
        if call_path == "float" and len(node.args) == 1:
            try:
                value = float(ast.literal_eval(node.args[0]))
            except Exception:
                pass
            else:
                if math.isfinite(value):
                    self.errors.append(B008(node.lineno, node.col_offset))
        else:
            self.errors.append(B008(node.lineno, node.col_offset))

        # Check for nested functions.
        self.generic_visit(node)

    def visit_Lambda(self, node):  # noqa: B906
        # Don't recurse into lambda expressions
        # as they are evaluated at call time.
        pass

    def visit(self, node):
        """Like super-visit but supports iteration over lists."""
        self.arg_depth += 1
        if isinstance(node, list):
            for elem in node:
                if elem is not None:
                    super().visit(elem)
        else:
            super().visit(node)
        self.arg_depth -= 1


class B020NameFinder(NameFinder):
    """Ignore names defined within the local scope of a comprehension."""

    def visit_GeneratorExp(self, node):
        self.visit(node.generators)

    def visit_ListComp(self, node):
        self.visit(node.generators)

    def visit_DictComp(self, node):
        self.visit(node.generators)

    def visit_comprehension(self, node):
        self.visit(node.iter)

    def visit_Lambda(self, node):
        self.visit(node.body)
        for lambda_arg in node.args.args:
            self.names.pop(lambda_arg.arg, None)


error = namedtuple("error", "lineno col message type vars")
Error = partial(partial, error, type=BugBearChecker, vars=())

B001 = Error(
    message=(
        "B001 Do not use {}, it also catches unexpected "
        "events like memory errors, interrupts, system exit, and so on.  "
        "Prefer `except Exception:`.  If you're sure what you're doing, "
        "be explicit and write `except BaseException:`."
    )
)

B002 = Error(
    message=(
        "B002 Python does not support the unary prefix increment. Writing "
        "++n is equivalent to +(+(n)), which equals n. You meant n += 1."
    )
)

B003 = Error(
    message=(
        "B003 Assigning to `os.environ` doesn't clear the environment. "
        "Subprocesses are going to see outdated variables, in disagreement "
        "with the current process. Use `os.environ.clear()` or the `env=` "
        "argument to Popen."
    )
)

B004 = Error(
    message=(
        "B004 Using `hasattr(x, '__call__')` to test if `x` is callable "
        "is unreliable. If `x` implements custom `__getattr__` or its "
        "`__call__` is itself not callable, you might get misleading "
        "results. Use `callable(x)` for consistent results."
    )
)

B005 = Error(
    message=(
        "B005 Using .strip() with multi-character strings is misleading "
        "the reader. It looks like stripping a substring. Move your "
        "character set to a constant if this is deliberate. Use "
        ".replace(), .removeprefix(), .removesuffix(), or regular "
        "expressions to remove string fragments."
    )
)
B005.methods = {"lstrip", "rstrip", "strip"}
B005.valid_paths = {}

B006 = Error(
    message=(
        "B006 Do not use mutable data structures for argument defaults.  They "
        "are created during function definition time. All calls to the function "
        "reuse this one instance of that data structure, persisting changes "
        "between them."
    )
)
B006.mutable_literals = ("Dict", "List", "Set")
B006.mutable_comprehensions = ("ListComp", "DictComp", "SetComp")
B006.mutable_calls = {
    "Counter",
    "OrderedDict",
    "collections.Counter",
    "collections.OrderedDict",
    "collections.defaultdict",
    "collections.deque",
    "defaultdict",
    "deque",
    "dict",
    "list",
    "set",
}
B007 = Error(
    message=(
        "B007 Loop control variable {!r} not used within the loop body. "
        "If this is intended, start the name with an underscore."
    )
)
B008 = Error(
    message=(
        "B008 Do not perform function calls in argument defaults.  The call is "
        "performed only once at function definition time. All calls to your "
        "function will reuse the result of that definition-time function call.  If "
        "this is intended, assign the function call to a module-level variable and "
        "use that variable as a default value."
    )
)
B008.immutable_calls = {
    "tuple",
    "frozenset",
    "types.MappingProxyType",
    "MappingProxyType",
    "re.compile",
    "operator.attrgetter",
    "operator.itemgetter",
    "operator.methodcaller",
    "attrgetter",
    "itemgetter",
    "methodcaller",
}
B009 = Error(
    message=(
        "B009 Do not call getattr with a constant attribute value, "
        "it is not any safer than normal property access."
    )
)
B010 = Error(
    message=(
        "B010 Do not call setattr with a constant attribute value, "
        "it is not any safer than normal property access."
    )
)
B011 = Error(
    message=(
        "B011 Do not call assert False since python -O removes these calls. "
        "Instead callers should raise AssertionError()."
    )
)
B012 = Error(
    message=(
        "B012 return/continue/break inside finally blocks cause exceptions "
        "to be silenced. Exceptions should be silenced in except blocks. Control "
        "statements can be moved outside the finally block."
    )
)
B013 = Error(
    message=(
        "B013 A length-one tuple literal is redundant.  "
        "Write `except {0}:` instead of `except ({0},):`."
    )
)
B014 = Error(
    message=(
        "B014 Redundant exception types in `except ({0}){1}:`.  "
        "Write `except {2}{1}:`, which catches exactly the same exceptions."
    )
)
B014.redundant_exceptions = {
    "OSError": {
        # All of these are actually aliases of OSError since Python 3.3
        "IOError",
        "EnvironmentError",
        "WindowsError",
        "mmap.error",
        "socket.error",
        "select.error",
    },
    "ValueError": {
        "binascii.Error",
    },
}
B015 = Error(
    message=(
        "B015 Result of comparison is not used. This line doesn't do "
        "anything. Did you intend to prepend it with assert?"
    )
)
B016 = Error(
    message=(
        "B016 Cannot raise a literal. Did you intend to return it or raise "
        "an Exception?"
    )
)
B017 = Error(
    message=(
        "B017 `assertRaises(Exception)` and `pytest.raises(Exception)` should "
        "be considered evil. They can lead to your test passing even if the "
        "code being tested is never executed due to a typo. Assert for a more "
        "specific exception (builtin or custom), or use `assertRaisesRegex` "
        "(if using `assertRaises`), or add the `match` keyword argument (if "
        "using `pytest.raises`), or use the context manager form with a target."
    )
)
B018 = Error(
    message=(
        "B018 Found useless expression. Consider either assigning it to a "
        "variable or removing it."
    )
)
B019 = Error(
    message=(
        "B019 Use of `functools.lru_cache` or `functools.cache` on methods "
        "can lead to memory leaks. The cache may retain instance references, "
        "preventing garbage collection."
    )
)
B019.caches = {
    "functools.cache",
    "functools.lru_cache",
    "cache",
    "lru_cache",
}
B020 = Error(
    message=(
        "B020 Found for loop that reassigns the iterable it is iterating "
        + "with each iterable value."
    )
)
B021 = Error(
    message=(
        "B021 f-string used as docstring. "
        "This will be interpreted by python as a joined string rather than a docstring."
    )
)
B022 = Error(
    message=(
        "B022 No arguments passed to `contextlib.suppress`. "
        "No exceptions will be suppressed and therefore this "
        "context manager is redundant."
    )
)

B023 = Error(message="B023 Function definition does not bind loop variable {!r}.")
B024 = Error(
    message=(
        "B024 {} is an abstract base class, but none of the methods it defines are"
        " abstract. This is not necessarily an error, but you might have forgotten to"
        " add the @abstractmethod decorator, potentially in conjunction with"
        " @classmethod, @property and/or @staticmethod."
    )
)
B025 = Error(
    message=(
        "B025 Exception `{0}` has been caught multiple times. Only the first except"
        " will be considered and all other except catches can be safely removed."
    )
)
B026 = Error(
    message=(
        "B026 Star-arg unpacking after a keyword argument is strongly discouraged, "
        "because it only works when the keyword parameter is declared after all "
        "parameters supplied by the unpacked sequence, and this change of ordering can "
        "surprise and mislead readers."
    )
)
B027 = Error(
    message=(
        "B027 {} is an empty method in an abstract base class, but has no abstract"
        " decorator. Consider adding @abstractmethod."
    )
)

# Warnings disabled by default.
B901 = Error(
    message=(
        "B901 Using `yield` together with `return x`. Use native "
        "`async def` coroutines or put a `# noqa` comment on this "
        "line if this was intentional."
    )
)
B902 = Error(
    message=(
        "B902 Invalid first argument {} used for {} method. Use the "
        "canonical first argument name in methods, i.e. {}."
    )
)
B902.implicit_classmethods = {"__new__", "__init_subclass__", "__class_getitem__"}
B902.self = ["self"]  # it's a list because the first is preferred
B902.cls = ["cls", "klass"]  # ditto.
B902.metacls = ["metacls", "metaclass", "typ", "mcs"]  # ditto.

B903 = Error(
    message=(
        "B903 Data class should either be immutable or use __slots__ to "
        "save memory. Use collections.namedtuple to generate an immutable "
        "class, or enumerate the attributes in a __slot__ declaration in "
        "the class to leave attributes mutable."
    )
)

B904 = Error(
    message=(
        "B904 Within an `except` clause, raise exceptions with `raise ... from err` or"
        " `raise ... from None` to distinguish them from errors in exception handling. "
        " See https://docs.python.org/3/tutorial/errors.html#exception-chaining for"
        " details."
    )
)

B905 = Error(message="B905 `zip()` without an explicit `strict=` parameter.")

B906 = Error(
    message=(
        "B906 `visit_` function with no further calls to a visit function, which might"
        " prevent the `ast` visitor from properly visiting all nodes."
        " Consider adding a call to `self.generic_visit(node)`."
    )
)

B907 = Error(
    message=(
        "B907 {!r} is manually surrounded by quotes, consider using the `!r` conversion"
        " flag."
    )
)

B950 = Error(message="B950 line too long ({} > {} characters)")

disabled_by_default = ["B901", "B902", "B903", "B904", "B905", "B906", "B950"]