What’s New In Pylint 1.8

Release:1.8.0
Date:|TBA|

Summary – Release highlights

  • None so far

New checkers

  • 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
        return
    

    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:
            try:
                yield next(g)
            except StopIteration:
                return
    

    The check about raising a StopIteration inside a generator is also valid if the exception raised inherit from StopIteration. Close #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)
    

Other Changes

  • 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):
        @property
        def foo(self):
            """My Sphinx style docstring description.
    
            :type: int
            """
            return 10
    
    class Foo(object):
        @property
        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-postive).

    foo = None
    if 'bar' in (foo or {}):
      pass
    
  • missing-param-doc and missing-type-doc are no longer emitted when Args and Keyword Args are mixed in Google docstring.