Summary -- Release highlights

  • None so far

New checkers

  • A new check was added, shallow-copy-environ.

    This warning message is emitted when shallow copy of os.environ is created. Shallow copy of os.environ doesn't work as people may expect. os.environ is not a dict object but rather a proxy object, so any changes made on copy may have unexpected effects on os.environ

    Instead of copy.copy(os.environ) method os.environ.copy() should be used.

    See for details.

    import copy
    import os
    wrong_env_copy = copy.copy(os.environ)  # will emit pylint warning
    wrong_env_copy['ENV_VAR'] = 'new_value'  # changes os.environ
    assert os.environ['ENV_VAR'] == 'new_value'
    good_env_copy = dict(os.environ)  # the right way
    good_env_copy['ENV_VAR'] = 'different_value'  # doesn't change os.environ
    assert os.environ['ENV_VAR'] == 'new_value'
  • A new check was added, keyword-arg-before-vararg.

    This warning message is emitted when a function is defined with a keyword argument appearing before variable-length positional arguments (*args). This may lead to args list getting modified if keyword argument's value is not provided in the function call assuming it will take default value provided in the definition.

    def foo(a, b=3, *args):
        print(a, b, args)
    # Case1: a=0, b=2, args=(4,5)
    foo(0,2,4,5) # 0 2 (4,5) ==> Observed values are same as expected values
    # Case2: a=0, b=<default_value>, args=(4,5)
    foo(0,4,5) # 0 4 (5,) ==> args list got modified as well as the observed value of b
    # Case3: Syntax Error if tried as follows
    foo(0,b=2,4,5) # syntax error
  • A new check was added, simplify-boolean-expression.

    This message is emitted when consider-using-ternary check would emit not equivalent code, due to truthy element being falsy in boolean context.

    value = condition and False or other_value

    This flawed construct may be simplified to:

    value = other_value
  • A new check was added, bad-thread-instantiation.

    This message is emitted when the threading.Thread class does not receive the target argument, but receives just one argument, which is by default the group parameter.

    In the following example, the instantiation will fail, which is definitely not desired:

    import threading
    threading.Thread(lambda: print(1)) # Oups, this is the group parameter
  • A new Python 3 checker was added to warn about accessing functions that have been removed from the itertools module izip, imap, iflter, izip_longest, and ifilterfalse.

    from itertools import izip
    print(list(izip([1, 2], [3])))

    Instead use six.moves to import a Python 2 and Python 3 compatible function:

    from six.moves import zip
    print(list(zip([1, 2], [3])))
  • A new Python 3 checker was added to warn about accessing deprecated fields from the types module like ListType or IntType

    from types import ListType
    print(isinstance([], ListType))

    Instead use the declarations in the builtin namespace:

    print(isinstance([], list))
  • A new Python 3 checker was added to warn about declaring a next method that would have implemented the Iterator protocol in Python 2 but is now a normal method in Python 3.

    class Foo(object):
        def next(self):
            return 42

    Instead implement a __next__ method and use six.Iterator as a base class or alias next to __next__:

    class Foo(object):
        def __next__(self):
            return 42
        next = __next__
  • Three new Python 3 checkers were added to warn about using dictionary methods in non-iterating contexts.

    For example, the following are returning iterators in Python 3:

    .. code-block:: python

    d = {} d.keys()[0] d.items()[0] d.values() + d.keys()

  • A new Python 3 porting check was added, non-ascii-bytes-literals

    This message is emitted whenever we detect that a bytes string contain non-ASCII characters, which results in a SyntaxError on Python 3.

  • A new warning, raising-format-tuple, will catch situations where the intent was likely raising an exception with a formatted message string, but the actual code did omit the formatting and instead passes template string and value parameters as separate arguments to the exception constructor. So it detects things like

    raise SomeError('message about %s', foo)
    raise SomeError('message about {}', foo)

    which likely were meant instead as

    raise SomeError('message about %s' % foo)
    raise SomeError('message about {}'.format(foo))

    This warning can be ignored on projects which deliberately use lazy formatting of messages in all user-facing exception handlers.

  • Following the recommendations of PEP479 ,a new Python 3.0 checker was added to warn about raising a StopIteration inside a generator. Raising a StopIteration inside a generator may be due a direct call to raise StopIteration:

    def gen_stopiter():
        yield 1
        yield 2
        yield 3
        raise StopIteration

    Instead use a simple return statement

    def gen_stopiter():
        yield 1
        yield 2
        yield 3

    Raising a StopIteration may also be due to the call to next function with a generator as argument:

    def gen_next_raises_stopiter():
        g = gen_ok()
        while True:
            yield next(g)

    In this case, surround the call to next with a try/except block:

    def gen_next_raises_stopiter():
        g = gen_ok()
        while True:
                yield next(g)
            except StopIteration:
  • The check about raising a StopIteration inside a generator is also valid if the exception raised inherit from StopIteration.

    Closes #1385

  • A new Python checker was added to warn about using a + operator inside call of logging methods when one of the operands is a literal string:

    import logging
    var = "123"
    logging.log(logging.INFO, "Var: " + var)

    Instead use formatted string and positional arguments :

    import logging
    var = "123"
    logging.log(logging.INFO, "Var: %s", var)
  • A new Python checker was added to warn about inconsistent-return-statements. A function or a method has inconsistent return statements if it returns both explicit and implicit values :

    def mix_implicit_explicit_returns(arg):
        if arg < 10:
            return True
        elif arg < 20:

    According to PEP8, if any return statement returns an expression, any return statements where no value is returned should explicitly state this as return None, and an explicit return statement should be present at the end of the function (if reachable). Thus, the previous function should be written:

    def mix_implicit_explicit_returns(arg):
        if arg < 10:
            return True
        elif arg < 20:
            return None

    Closes #1267

Other Changes

  • Fixing u'' string in superfluous-parens message.

  • Configuration options of invalid name checker are significantly redesigned. Predefined rules for common naming styles were introduced. For typical setups, user friendly options like --function-naming-style=camelCase may be used in place of hand-written regular expressions. Default linter config enforce PEP8-compatible naming style. See documentation for details.

  • Raise meaningful exception in case of invalid reporter class (output format) being selected.

  • The docparams extension now allows a property docstring to document both the property and the setter. Therefore setters can also have no docstring.

  • The docparams extension now understands property type syntax.

    class Foo(object):
        def foo(self):
            """My Sphinx style docstring description.
            :type: int
            return 10
    class Foo(object):
        def foo(self):
            """int: My Numpy and Google docstring style description."""
            return 10
  • In case of --output-format=json, the dictionary returned holds a new key-value pair. The key is message-id and the value the message id.

  • Spelling checker has a new configuration parameter max-spelling-suggestions, which affects maximum count of suggestions included in emitted message.

  • The invalid-name check contains the name of the template that caused the failure.

    For the given code, pylint used to emit invalid-name in the form Invalid constant name var, without offering any context why var is not such a good name.

    With this change, it is now more clear what should be improved for a name to be accepted according to its corresponding template.

  • New configuration flag, suggestion-mode was introduced. When enabled, pylint would attempt to emit user-friendly suggestions instead of spurious errors for some known false-positive scenarios. Flag is enabled by default.

  • superfluous-parens is no longer wrongly emitted for logical statements involving in operator (see example below for what used to be false-positive).

    foo = None
    if 'bar' in (foo or {}):
  • Redefinition of dummy function is now possible. function-redefined message won't be emitted anymore when dummy functions are redefined.

  • missing-param-doc and missing-type-doc are no longer emitted when Args and Keyword Args are mixed in Google docstring.

  • Fix of false positive useless-super-delegation message when parameters default values are different from those used in the base class.

  • Fix of false positive useless-else-on-loop message when break statements are deeply nested inside loop.

  • The Python 3 porting checker no longer emits multiple no-absolute-import per file.

  • The Python 3 porting checker respects disabled checkers found in the config file.

  • Modules, classes, or methods consist of compound statements that exceed the docstring-min-length are now correctly emitting missing-docstring

  • Fix no wrong-import-order message emitted on ordering of first and third party libraries. With this fix, pylint distinguishes first and third party modules when checking import order.

  • Fix the ignored pylint disable=fixme directives for comments following the last statement in a file.

  • Fix line-too-long message deactivated by wrong disable directive. The directive disable=fixme doesn't deactivate anymore the emission of line-too-long message for long commented lines.

  • If the rcfile specified on the command line doesn't exist, then an IOError exception is raised.

  • Fix the wrong scope of disable= directive after a commented line. For example when a disable=line-too-long directive is at the end of a long commented line, it no longer disables the emission of line-too-long message for lines that follow.