sandbox/morph / Doc / whatsnew / 2.7.rst

What's New in Python 2.7

Author: A.M. Kuchling (amk at amk.ca)
Release:|release|
Date: |today|

This article explains the new features in Python 2.7. Python 2.7 was released on July 3, 2010.

Numeric handling has been improved in many ways, for both floating-point numbers and for the :class:`~decimal.Decimal` class. There are some useful additions to the standard library, such as a greatly enhanced :mod:`unittest` module, the :mod:`argparse` module for parsing command-line options, convenient :class:`~collections.OrderedDict` and :class:`~collections.Counter` classes in the :mod:`collections` module, and many other improvements.

Python 2.7 is planned to be the last of the 2.x releases, so we worked on making it a good release for the long term. To help with porting to Python 3, several new features from the Python 3.x series have been included in 2.7.

This article doesn't attempt to provide a complete specification of the new features, but instead provides a convenient overview. For full details, you should refer to the documentation for Python 2.7 at http://docs.python.org. If you want to understand the rationale for the design and implementation, refer to the PEP for a particular new feature or the issue on http://bugs.python.org in which a change was discussed. Whenever possible, "What's New in Python" links to the bug/patch item for each change.

The Future for Python 2.x

Python 2.7 is intended to be the last major release in the 2.x series. The Python maintainers are planning to focus their future efforts on the Python 3.x series.

This means that 2.7 will remain in place for a long time, running production systems that have not been ported to Python 3.x. Two consequences of the long-term significance of 2.7 are:

  • It's very likely the 2.7 release will have a longer period of maintenance compared to earlier 2.x versions. Python 2.7 will continue to be maintained while the transition to 3.x continues, and the developers are planning to support Python 2.7 with bug-fix releases beyond the typical two years.

  • A policy decision was made to silence warnings only of interest to developers. :exc:`DeprecationWarning` and its descendants are now ignored unless otherwise requested, preventing users from seeing warnings triggered by an application. This change was also made in the branch that will become Python 3.2. (Discussed on stdlib-sig and carried out in :issue:`7319`.)

    In previous releases, :exc:`DeprecationWarning` messages were enabled by default, providing Python developers with a clear indication of where their code may break in a future major version of Python.

    However, there are increasingly many users of Python-based applications who are not directly involved in the development of those applications. :exc:`DeprecationWarning` messages are irrelevant to such users, making them worry about an application that's actually working correctly and burdening application developers with responding to these concerns.

    You can re-enable display of :exc:`DeprecationWarning` messages by running Python with the :option:`-Wdefault <-W>` (short form: :option:`-Wd <-W>`) switch, or by setting the :envvar:`PYTHONWARNINGS` environment variable to "default" (or "d") before running Python. Python code can also re-enable them by calling warnings.simplefilter('default').

Python 3.1 Features

Much as Python 2.6 incorporated features from Python 3.0, version 2.7 incorporates some of the new features in Python 3.1. The 2.x series continues to provide tools for migrating to the 3.x series.

A partial list of 3.1 features that were backported to 2.7:

  • The syntax for set literals ({1,2,3} is a mutable set).
  • Dictionary and set comprehensions ({i: i*2 for i in range(3)}).
  • Multiple context managers in a single :keyword:`with` statement.
  • A new version of the :mod:`io` library, rewritten in C for performance.
  • The ordered-dictionary type described in :ref:`pep-0372`.
  • The new "," format specifier described in :ref:`pep-0378`.
  • The :class:`memoryview` object.
  • A small subset of the :mod:`importlib` module, described below.
  • The :func:`repr` of a float x is shorter in many cases: it's now based on the shortest decimal string that's guaranteed to round back to x. As in previous versions of Python, it's guaranteed that float(repr(x)) recovers x.
  • Float-to-string and string-to-float conversions are correctly rounded. The :func:`round` function is also now correctly rounded.
  • The :c:type:`PyCapsule` type, used to provide a C API for extension modules.
  • The :c:func:`PyLong_AsLongAndOverflow` C API function.

Other new Python3-mode warnings include:

PEP 372: Adding an Ordered Dictionary to collections

Regular Python dictionaries iterate over key/value pairs in arbitrary order. Over the years, a number of authors have written alternative implementations that remember the order that the keys were originally inserted. Based on the experiences from those implementations, 2.7 introduces a new :class:`~collections.OrderedDict` class in the :mod:`collections` module.

The :class:`~collections.OrderedDict` API provides the same interface as regular dictionaries but iterates over keys and values in a guaranteed order depending on when a key was first inserted:

>>> from collections import OrderedDict
>>> d = OrderedDict([('first', 1),
...                  ('second', 2),
...                  ('third', 3)])
>>> d.items()
[('first', 1), ('second', 2), ('third', 3)]

If a new entry overwrites an existing entry, the original insertion position is left unchanged:

>>> d['second'] = 4
>>> d.items()
[('first', 1), ('second', 4), ('third', 3)]

Deleting an entry and reinserting it will move it to the end:

>>> del d['second']
>>> d['second'] = 5
>>> d.items()
[('first', 1), ('third', 3), ('second', 5)]

The :meth:`~collections.OrderedDict.popitem` method has an optional last argument that defaults to True. If last is True, the most recently added key is returned and removed; if it's False, the oldest key is selected:

>>> od = OrderedDict([(x,0) for x in range(20)])
>>> od.popitem()
(19, 0)
>>> od.popitem()
(18, 0)
>>> od.popitem(last=False)
(0, 0)
>>> od.popitem(last=False)
(1, 0)

Comparing two ordered dictionaries checks both the keys and values, and requires that the insertion order was the same:

>>> od1 = OrderedDict([('first', 1),
...                    ('second', 2),
...                    ('third', 3)])
>>> od2 = OrderedDict([('third', 3),
...                    ('first', 1),
...                    ('second', 2)])
>>> od1 == od2
False
>>> # Move 'third' key to the end
>>> del od2['third']; od2['third'] = 3
>>> od1 == od2
True

Comparing an :class:`~collections.OrderedDict` with a regular dictionary ignores the insertion order and just compares the keys and values.

How does the :class:`~collections.OrderedDict` work? It maintains a doubly-linked list of keys, appending new keys to the list as they're inserted. A secondary dictionary maps keys to their corresponding list node, so deletion doesn't have to traverse the entire linked list and therefore remains O(1).

The standard library now supports use of ordered dictionaries in several modules.

PEP 378: Format Specifier for Thousands Separator

To make program output more readable, it can be useful to add separators to large numbers, rendering them as 18,446,744,073,709,551,616 instead of 18446744073709551616.

The fully general solution for doing this is the :mod:`locale` module, which can use different separators ("," in North America, "." in Europe) and different grouping sizes, but :mod:`locale` is complicated to use and unsuitable for multi-threaded applications where different threads are producing output for different locales.

Therefore, a simple comma-grouping mechanism has been added to the mini-language used by the :meth:`str.format` method. When formatting a floating-point number, simply include a comma between the width and the precision:

>>> '{:20,.2f}'.format(18446744073709551616.0)
'18,446,744,073,709,551,616.00'

When formatting an integer, include the comma after the width:

>>> '{:20,d}'.format(18446744073709551616)
'18,446,744,073,709,551,616'

This mechanism is not adaptable at all; commas are always used as the separator and the grouping is always into three-digit groups. The comma-formatting mechanism isn't as general as the :mod:`locale` module, but it's easier to use.

PEP 389: The argparse Module for Parsing Command Lines

The :mod:`argparse` module for parsing command-line arguments was added as a more powerful replacement for the :mod:`optparse` module.

This means Python now supports three different modules for parsing command-line arguments: :mod:`getopt`, :mod:`optparse`, and :mod:`argparse`. The :mod:`getopt` module closely resembles the C library's :c:func:`getopt` function, so it remains useful if you're writing a Python prototype that will eventually be rewritten in C. :mod:`optparse` becomes redundant, but there are no plans to remove it because there are many scripts still using it, and there's no automated way to update these scripts. (Making the :mod:`argparse` API consistent with :mod:`optparse`'s interface was discussed but rejected as too messy and difficult.)

In short, if you're writing a new script and don't need to worry about compatibility with earlier versions of Python, use :mod:`argparse` instead of :mod:`optparse`.

Here's an example:

import argparse

parser = argparse.ArgumentParser(description='Command-line example.')

# Add optional switches
parser.add_argument('-v', action='store_true', dest='is_verbose',
                    help='produce verbose output')
parser.add_argument('-o', action='store', dest='output',
                    metavar='FILE',
                    help='direct output to FILE instead of stdout')
parser.add_argument('-C', action='store', type=int, dest='context',
                    metavar='NUM', default=0,
                    help='display NUM lines of added context')

# Allow any number of additional arguments.
parser.add_argument(nargs='*', action='store', dest='inputs',
                    help='input filenames (default is stdin)')

args = parser.parse_args()
print args.__dict__

Unless you override it, :option:`-h` and :option:`--help` switches are automatically added, and produce neatly formatted output:

-> ./python.exe argparse-example.py --help
usage: argparse-example.py [-h] [-v] [-o FILE] [-C NUM] [inputs [inputs ...]]

Command-line example.

positional arguments:
  inputs      input filenames (default is stdin)

optional arguments:
  -h, --help  show this help message and exit
  -v          produce verbose output
  -o FILE     direct output to FILE instead of stdout
  -C NUM      display NUM lines of added context

As with :mod:`optparse`, the command-line switches and arguments are returned as an object with attributes named by the dest parameters:

-> ./python.exe argparse-example.py -v
{'output': None,
 'is_verbose': True,
 'context': 0,
 'inputs': []}

-> ./python.exe argparse-example.py -v -o /tmp/output -C 4 file1 file2
{'output': '/tmp/output',
 'is_verbose': True,
 'context': 4,
 'inputs': ['file1', 'file2']}

:mod:`argparse` has much fancier validation than :mod:`optparse`; you can specify an exact number of arguments as an integer, 0 or more arguments by passing '*', 1 or more by passing '+', or an optional argument with '?'. A top-level parser can contain sub-parsers to define subcommands that have different sets of switches, as in svn commit, svn checkout, etc. You can specify an argument's type as :class:`~argparse.FileType`, which will automatically open files for you and understands that '-' means standard input or output.

PEP 391: Dictionary-Based Configuration For Logging

The :mod:`logging` module is very flexible; applications can define a tree of logging subsystems, and each logger in this tree can filter out certain messages, format them differently, and direct messages to a varying number of handlers.

All this flexibility can require a lot of configuration. You can write Python statements to create objects and set their properties, but a complex set-up requires verbose but boring code. :mod:`logging` also supports a :func:`~logging.fileConfig` function that parses a file, but the file format doesn't support configuring filters, and it's messier to generate programmatically.

Python 2.7 adds a :func:`~logging.dictConfig` function that uses a dictionary to configure logging. There are many ways to produce a dictionary from different sources: construct one with code; parse a file containing JSON; or use a YAML parsing library if one is installed. For more information see :ref:`logging-config-api`.

The following example configures two loggers, the root logger and a logger named "network". Messages sent to the root logger will be sent to the system log using the syslog protocol, and messages to the "network" logger will be written to a :file:`network.log` file that will be rotated once the log reaches 1MB.

import logging
import logging.config

configdict = {
 'version': 1,    # Configuration schema in use; must be 1 for now
 'formatters': {
     'standard': {
         'format': ('%(asctime)s %(name)-15s '
                    '%(levelname)-8s %(message)s')}},

 'handlers': {'netlog': {'backupCount': 10,
                     'class': 'logging.handlers.RotatingFileHandler',
                     'filename': '/logs/network.log',
                     'formatter': 'standard',
                     'level': 'INFO',
                     'maxBytes': 1000000},
              'syslog': {'class': 'logging.handlers.SysLogHandler',
                         'formatter': 'standard',
                         'level': 'ERROR'}},

 # Specify all the subordinate loggers
 'loggers': {
             'network': {
                         'handlers': ['netlog']
             }
 },
 # Specify properties of the root logger
 'root': {
          'handlers': ['syslog']
 },
}

# Set up configuration
logging.config.dictConfig(configdict)

# As an example, log two error messages
logger = logging.getLogger('/')
logger.error('Database not found')

netlogger = logging.getLogger('network')
netlogger.error('Connection failed')

Three smaller enhancements to the :mod:`logging` module, all implemented by Vinay Sajip, are:

PEP 3106: Dictionary Views

The dictionary methods :meth:`~dict.keys`, :meth:`~dict.values`, and :meth:`~dict.items` are different in Python 3.x. They return an object called a :dfn:`view` instead of a fully materialized list.

It's not possible to change the return values of :meth:`~dict.keys`, :meth:`~dict.values`, and :meth:`~dict.items` in Python 2.7 because too much code would break. Instead the 3.x versions were added under the new names :meth:`~dict.viewkeys`, :meth:`~dict.viewvalues`, and :meth:`~dict.viewitems`.

>>> d = dict((i*10, chr(65+i)) for i in range(26))
>>> d
{0: 'A', 130: 'N', 10: 'B', 140: 'O', 20: ..., 250: 'Z'}
>>> d.viewkeys()
dict_keys([0, 130, 10, 140, 20, 150, 30, ..., 250])

Views can be iterated over, but the key and item views also behave like sets. The & operator performs intersection, and | performs a union:

>>> d1 = dict((i*10, chr(65+i)) for i in range(26))
>>> d2 = dict((i**.5, i) for i in range(1000))
>>> d1.viewkeys() & d2.viewkeys()
set([0.0, 10.0, 20.0, 30.0])
>>> d1.viewkeys() | range(0, 30)
set([0, 1, 130, 3, 4, 5, 6, ..., 120, 250])

The view keeps track of the dictionary and its contents change as the dictionary is modified:

>>> vk = d.viewkeys()
>>> vk
dict_keys([0, 130, 10, ..., 250])
>>> d[260] = '&'
>>> vk
dict_keys([0, 130, 260, 10, ..., 250])

However, note that you can't add or remove keys while you're iterating over the view:

>>> for k in vk:
...     d[k*2] = k
...
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
RuntimeError: dictionary changed size during iteration

You can use the view methods in Python 2.x code, and the 2to3 converter will change them to the standard :meth:`~dict.keys`, :meth:`~dict.values`, and :meth:`~dict.items` methods.

PEP 3137: The memoryview Object

The :class:`memoryview` object provides a view of another object's memory content that matches the :class:`bytes` type's interface.

>>> import string
>>> m = memoryview(string.letters)
>>> m
<memory at 0x37f850>
>>> len(m)           # Returns length of underlying object
52
>>> m[0], m[25], m[26]   # Indexing returns one byte
('a', 'z', 'A')
>>> m2 = m[0:26]         # Slicing returns another memoryview
>>> m2
<memory at 0x37f080>

The content of the view can be converted to a string of bytes or a list of integers:

>>> m2.tobytes()
'abcdefghijklmnopqrstuvwxyz'
>>> m2.tolist()
[97, 98, 99, 100, 101, 102, 103, ... 121, 122]
>>>

:class:`memoryview` objects allow modifying the underlying object if it's a mutable object.

>>> m2[0] = 75
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: cannot modify read-only memory
>>> b = bytearray(string.letters)  # Creating a mutable object
>>> b
bytearray(b'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ')
>>> mb = memoryview(b)
>>> mb[0] = '*'         # Assign to view, changing the bytearray.
>>> b[0:5]              # The bytearray has been changed.
bytearray(b'*bcde')
>>>

Other Language Changes

Some smaller changes made to the core Python language are:

  • The syntax for set literals has been backported from Python 3.x. Curly brackets are used to surround the contents of the resulting mutable set; set literals are distinguished from dictionaries by not containing colons and values. {} continues to represent an empty dictionary; use set() for an empty set.

    >>> {1, 2, 3, 4, 5}
    set([1, 2, 3, 4, 5])
    >>> set() # empty set
    set([])
    >>> {}    # empty dict
    {}
    

    Backported by Alexandre Vassalotti; :issue:`2335`.

  • Dictionary and set comprehensions are another feature backported from 3.x, generalizing list/generator comprehensions to use the literal syntax for sets and dictionaries.

    >>> {x: x*x for x in range(6)}
    {0: 0, 1: 1, 2: 4, 3: 9, 4: 16, 5: 25}
    >>> {('a'*x) for x in range(6)}
    set(['', 'a', 'aa', 'aaa', 'aaaa', 'aaaaa'])
    

    Backported by Alexandre Vassalotti; :issue:`2333`.

  • The :keyword:`with` statement can now use multiple context managers in one statement. Context managers are processed from left to right and each one is treated as beginning a new :keyword:`with` statement. This means that:

    with A() as a, B() as b:
        ... suite of statements ...
    

    is equivalent to:

    with A() as a:
        with B() as b:
            ... suite of statements ...
    

    The :func:`contextlib.nested` function provides a very similar function, so it's no longer necessary and has been deprecated.

    (Proposed in http://codereview.appspot.com/53094; implemented by Georg Brandl.)

  • Conversions between floating-point numbers and strings are now correctly rounded on most platforms. These conversions occur in many different places: :func:`str` on floats and complex numbers; the :class:`float` and :class:`complex` constructors; numeric formatting; serializing and deserializing floats and complex numbers using the :mod:`marshal`, :mod:`pickle` and :mod:`json` modules; parsing of float and imaginary literals in Python code; and :class:`~decimal.Decimal`-to-float conversion.

    Related to this, the :func:`repr` of a floating-point number x now returns a result based on the shortest decimal string that's guaranteed to round back to x under correct rounding (with round-half-to-even rounding mode). Previously it gave a string based on rounding x to 17 decimal digits.

    The rounding library responsible for this improvement works on Windows and on Unix platforms using the gcc, icc, or suncc compilers. There may be a small number of platforms where correct operation of this code cannot be guaranteed, so the code is not used on such systems. You can find out which code is being used by checking :data:`sys.float_repr_style`, which will be short if the new code is in use and legacy if it isn't.

    Implemented by Eric Smith and Mark Dickinson, using David Gay's :file:`dtoa.c` library; :issue:`7117`.

  • Conversions from long integers and regular integers to floating point now round differently, returning the floating-point number closest to the number. This doesn't matter for small integers that can be converted exactly, but for large numbers that will unavoidably lose precision, Python 2.7 now approximates more closely. For example, Python 2.6 computed the following:

    >>> n = 295147905179352891391
    >>> float(n)
    2.9514790517935283e+20
    >>> n - long(float(n))
    65535L
    

    Python 2.7's floating-point result is larger, but much closer to the true value:

    >>> n = 295147905179352891391
    >>> float(n)
    2.9514790517935289e+20
    >>> n - long(float(n))
    -1L
    

    (Implemented by Mark Dickinson; :issue:`3166`.)

    Integer division is also more accurate in its rounding behaviours. (Also implemented by Mark Dickinson; :issue:`1811`.)

  • Implicit coercion for complex numbers has been removed; the interpreter will no longer ever attempt to call a :meth:`__coerce__` method on complex objects. (Removed by Meador Inge and Mark Dickinson; :issue:`5211`.)

  • The :meth:`str.format` method now supports automatic numbering of the replacement fields. This makes using :meth:`str.format` more closely resemble using %s formatting:

    >>> '{}:{}:{}'.format(2009, 04, 'Sunday')
    '2009:4:Sunday'
    >>> '{}:{}:{day}'.format(2009, 4, day='Sunday')
    '2009:4:Sunday'
    

    The auto-numbering takes the fields from left to right, so the first {...} specifier will use the first argument to :meth:`str.format`, the next specifier will use the next argument, and so on. You can't mix auto-numbering and explicit numbering -- either number all of your specifier fields or none of them -- but you can mix auto-numbering and named fields, as in the second example above. (Contributed by Eric Smith; :issue:`5237`.)

    Complex numbers now correctly support usage with :func:`format`, and default to being right-aligned. Specifying a precision or comma-separation applies to both the real and imaginary parts of the number, but a specified field width and alignment is applied to the whole of the resulting 1.5+3j output. (Contributed by Eric Smith; :issue:`1588` and :issue:`7988`.)

    The 'F' format code now always formats its output using uppercase characters, so it will now produce 'INF' and 'NAN'. (Contributed by Eric Smith; :issue:`3382`.)

    A low-level change: the :meth:`object.__format__` method now triggers a :exc:`PendingDeprecationWarning` if it's passed a format string, because the :meth:`__format__` method for :class:`object` converts the object to a string representation and formats that. Previously the method silently applied the format string to the string representation, but that could hide mistakes in Python code. If you're supplying formatting information such as an alignment or precision, presumably you're expecting the formatting to be applied in some object-specific way. (Fixed by Eric Smith; :issue:`7994`.)

  • The :func:`int` and :func:`long` types gained a bit_length method that returns the number of bits necessary to represent its argument in binary:

    >>> n = 37
    >>> bin(n)
    '0b100101'
    >>> n.bit_length()
    6
    >>> n = 2**123-1
    >>> n.bit_length()
    123
    >>> (n+1).bit_length()
    124
    

    (Contributed by Fredrik Johansson and Victor Stinner; :issue:`3439`.)

  • The :keyword:`import` statement will no longer try an absolute import if a relative import (e.g. from .os import sep) fails. This fixes a bug, but could possibly break certain :keyword:`import` statements that were only working by accident. (Fixed by Meador Inge; :issue:`7902`.)

  • It's now possible for a subclass of the built-in :class:`unicode` type to override the :meth:`__unicode__` method. (Implemented by Victor Stinner; :issue:`1583863`.)

  • The :class:`bytearray` type's :meth:`~bytearray.translate` method now accepts None as its first argument. (Fixed by Georg Brandl; :issue:`4759`.)

  • When using @classmethod and @staticmethod to wrap methods as class or static methods, the wrapper object now exposes the wrapped function as their :attr:`__func__` attribute. (Contributed by Amaury Forgeot d'Arc, after a suggestion by George Sakkis; :issue:`5982`.)

  • When a restricted set of attributes were set using __slots__, deleting an unset attribute would not raise :exc:`AttributeError` as you would expect. Fixed by Benjamin Peterson; :issue:`7604`.)

  • Two new encodings are now supported: "cp720", used primarily for Arabic text; and "cp858", a variant of CP 850 that adds the euro symbol. (CP720 contributed by Alexander Belchenko and Amaury Forgeot d'Arc in :issue:`1616979`; CP858 contributed by Tim Hatch in :issue:`8016`.)

  • The :class:`file` object will now set the :attr:`filename` attribute on the :exc:`IOError` exception when trying to open a directory on POSIX platforms (noted by Jan Kaliszewski; :issue:`4764`), and now explicitly checks for and forbids writing to read-only file objects instead of trusting the C library to catch and report the error (fixed by Stefan Krah; :issue:`5677`).

  • The Python tokenizer now translates line endings itself, so the :func:`compile` built-in function now accepts code using any line-ending convention. Additionally, it no longer requires that the code end in a newline.

  • Extra parentheses in function definitions are illegal in Python 3.x, meaning that you get a syntax error from def f((x)): pass. In Python3-warning mode, Python 2.7 will now warn about this odd usage. (Noted by James Lingard; :issue:`7362`.)

  • It's now possible to create weak references to old-style class objects. New-style classes were always weak-referenceable. (Fixed by Antoine Pitrou; :issue:`8268`.)

  • When a module object is garbage-collected, the module's dictionary is now only cleared if no one else is holding a reference to the dictionary (:issue:`7140`).

Interpreter Changes

A new environment variable, :envvar:`PYTHONWARNINGS`, allows controlling warnings. It should be set to a string containing warning settings, equivalent to those used with the :option:`-W` switch, separated by commas. (Contributed by Brian Curtin; :issue:`7301`.)

For example, the following setting will print warnings every time they occur, but turn warnings from the :mod:`Cookie` module into an error. (The exact syntax for setting an environment variable varies across operating systems and shells.)

export PYTHONWARNINGS=all,error:::Cookie:0

Optimizations

Several performance enhancements have been added:

  • A new opcode was added to perform the initial setup for :keyword:`with` statements, looking up the :meth:`__enter__` and :meth:`__exit__` methods. (Contributed by Benjamin Peterson.)

  • The garbage collector now performs better for one common usage pattern: when many objects are being allocated without deallocating any of them. This would previously take quadratic time for garbage collection, but now the number of full garbage collections is reduced as the number of objects on the heap grows. The new logic only performs a full garbage collection pass when the middle generation has been collected 10 times and when the number of survivor objects from the middle generation exceeds 10% of the number of objects in the oldest generation. (Suggested by Martin von Löwis and implemented by Antoine Pitrou; :issue:`4074`.)

  • The garbage collector tries to avoid tracking simple containers which can't be part of a cycle. In Python 2.7, this is now true for tuples and dicts containing atomic types (such as ints, strings, etc.). Transitively, a dict containing tuples of atomic types won't be tracked either. This helps reduce the cost of each garbage collection by decreasing the number of objects to be considered and traversed by the collector. (Contributed by Antoine Pitrou; :issue:`4688`.)

  • Long integers are now stored internally either in base 2**15 or in base 2**30, the base being determined at build time. Previously, they were always stored in base 2**15. Using base 2**30 gives significant performance improvements on 64-bit machines, but benchmark results on 32-bit machines have been mixed. Therefore, the default is to use base 2**30 on 64-bit machines and base 2**15 on 32-bit machines; on Unix, there's a new configure option :option:`--enable-big-digits` that can be used to override this default.

    Apart from the performance improvements this change should be invisible to end users, with one exception: for testing and debugging purposes there's a new structseq :data:`sys.long_info` that provides information about the internal format, giving the number of bits per digit and the size in bytes of the C type used to store each digit:

    >>> import sys
    >>> sys.long_info
    sys.long_info(bits_per_digit=30, sizeof_digit=4)
    

    (Contributed by Mark Dickinson; :issue:`4258`.)

    Another set of changes made long objects a few bytes smaller: 2 bytes smaller on 32-bit systems and 6 bytes on 64-bit. (Contributed by Mark Dickinson; :issue:`5260`.)

  • The division algorithm for long integers has been made faster by tightening the inner loop, doing shifts instead of multiplications, and fixing an unnecessary extra iteration. Various benchmarks show speedups of between 50% and 150% for long integer divisions and modulo operations. (Contributed by Mark Dickinson; :issue:`5512`.) Bitwise operations are also significantly faster (initial patch by Gregory Smith; :issue:`1087418`).

  • The implementation of % checks for the left-side operand being a Python string and special-cases it; this results in a 1-3% performance increase for applications that frequently use % with strings, such as templating libraries. (Implemented by Collin Winter; :issue:`5176`.)

  • List comprehensions with an if condition are compiled into faster bytecode. (Patch by Antoine Pitrou, back-ported to 2.7 by Jeffrey Yasskin; :issue:`4715`.)

  • Converting an integer or long integer to a decimal string was made faster by special-casing base 10 instead of using a generalized conversion function that supports arbitrary bases. (Patch by Gawain Bolton; :issue:`6713`.)

  • The :meth:`split`, :meth:`replace`, :meth:`rindex`, :meth:`rpartition`, and :meth:`rsplit` methods of string-like types (strings, Unicode strings, and :class:`bytearray` objects) now use a fast reverse-search algorithm instead of a character-by-character scan. This is sometimes faster by a factor of 10. (Added by Florent Xicluna; :issue:`7462` and :issue:`7622`.)

  • The :mod:`pickle` and :mod:`cPickle` modules now automatically intern the strings used for attribute names, reducing memory usage of the objects resulting from unpickling. (Contributed by Jake McGuire; :issue:`5084`.)

  • The :mod:`cPickle` module now special-cases dictionaries, nearly halving the time required to pickle them. (Contributed by Collin Winter; :issue:`5670`.)

New and Improved Modules

As in every release, Python's standard library received a number of enhancements and bug fixes. Here's a partial list of the most notable changes, sorted alphabetically by module name. Consult the :file:`Misc/NEWS` file in the source tree for a more complete list of changes, or look through the Subversion logs for all the details.

New module: importlib

Python 3.1 includes the :mod:`importlib` package, a re-implementation of the logic underlying Python's :keyword:`import` statement. :mod:`importlib` is useful for implementors of Python interpreters and to users who wish to write new importers that can participate in the import process. Python 2.7 doesn't contain the complete :mod:`importlib` package, but instead has a tiny subset that contains a single function, :func:`~importlib.import_module`.

import_module(name, package=None) imports a module. name is a string containing the module or package's name. It's possible to do relative imports by providing a string that begins with a . character, such as ..utils.errors. For relative imports, the package argument must be provided and is the name of the package that will be used as the anchor for the relative import. :func:`~importlib.import_module` both inserts the imported module into sys.modules and returns the module object.

Here are some examples:

>>> from importlib import import_module
>>> anydbm = import_module('anydbm')  # Standard absolute import
>>> anydbm
<module 'anydbm' from '/p/python/Lib/anydbm.py'>
>>> # Relative import
>>> file_util = import_module('..file_util', 'distutils.command')
>>> file_util
<module 'distutils.file_util' from '/python/Lib/distutils/file_util.pyc'>

:mod:`importlib` was implemented by Brett Cannon and introduced in Python 3.1.

New module: sysconfig

The :mod:`sysconfig` module has been pulled out of the Distutils package, becoming a new top-level module in its own right. :mod:`sysconfig` provides functions for getting information about Python's build process: compiler switches, installation paths, the platform name, and whether Python is running from its source directory.

Some of the functions in the module are:

Consult the :mod:`sysconfig` documentation for more details and for a complete list of functions.

The Distutils package and :mod:`sysconfig` are now maintained by Tarek Ziadé, who has also started a Distutils2 package (source repository at http://hg.python.org/distutils2/) for developing a next-generation version of Distutils.

ttk: Themed Widgets for Tk

Tcl/Tk 8.5 includes a set of themed widgets that re-implement basic Tk widgets but have a more customizable appearance and can therefore more closely resemble the native platform's widgets. This widget set was originally called Tile, but was renamed to Ttk (for "themed Tk") on being added to Tcl/Tck release 8.5.

To learn more, read the :mod:`ttk` module documentation. You may also wish to read the Tcl/Tk manual page describing the Ttk theme engine, available at http://www.tcl.tk/man/tcl8.5/TkCmd/ttk_intro.htm. Some screenshots of the Python/Ttk code in use are at http://code.google.com/p/python-ttk/wiki/Screenshots.

The :mod:`ttk` module was written by Guilherme Polo and added in :issue:`2983`. An alternate version called Tile.py, written by Martin Franklin and maintained by Kevin Walzer, was proposed for inclusion in :issue:`2618`, but the authors argued that Guilherme Polo's work was more comprehensive.

Updated module: unittest

The :mod:`unittest` module was greatly enhanced; many new features were added. Most of these features were implemented by Michael Foord, unless otherwise noted. The enhanced version of the module is downloadable separately for use with Python versions 2.4 to 2.6, packaged as the :mod:`unittest2` package, from http://pypi.python.org/pypi/unittest2.

When used from the command line, the module can automatically discover tests. It's not as fancy as py.test or nose, but provides a simple way to run tests kept within a set of package directories. For example, the following command will search the :file:`test/` subdirectory for any importable test files named test*.py:

python -m unittest discover -s test

Consult the :mod:`unittest` module documentation for more details. (Developed in :issue:`6001`.)

The :func:`~unittest.main` function supports some other new options:

  • :option:`-b` or :option:`--buffer` will buffer the standard output and standard error streams during each test. If the test passes, any resulting output will be discarded; on failure, the buffered output will be displayed.

  • :option:`-c` or :option:`--catch` will cause the control-C interrupt to be handled more gracefully. Instead of interrupting the test process immediately, the currently running test will be completed and then the partial results up to the interruption will be reported. If you're impatient, a second press of control-C will cause an immediate interruption.

    This control-C handler tries to avoid causing problems when the code being tested or the tests being run have defined a signal handler of their own, by noticing that a signal handler was already set and calling it. If this doesn't work for you, there's a :func:`~unittest.removeHandler` decorator that can be used to mark tests that should have the control-C handling disabled.

  • :option:`-f` or :option:`--failfast` makes test execution stop immediately when a test fails instead of continuing to execute further tests. (Suggested by Cliff Dyer and implemented by Michael Foord; :issue:`8074`.)

The progress messages now show 'x' for expected failures and 'u' for unexpected successes when run in verbose mode. (Contributed by Benjamin Peterson.)

Test cases can raise the :exc:`~unittest.SkipTest` exception to skip a test (:issue:`1034053`).

The error messages for :meth:`~unittest.TestCase.assertEqual`, :meth:`~unittest.TestCase.assertTrue`, and :meth:`~unittest.TestCase.assertFalse` failures now provide more information. If you set the :attr:`~unittest.TestCase.longMessage` attribute of your :class:`~unittest.TestCase` classes to True, both the standard error message and any additional message you provide will be printed for failures. (Added by Michael Foord; :issue:`5663`.)

The :meth:`~unittest.TestCase.assertRaises` method now returns a context handler when called without providing a callable object to run. For example, you can write this:

with self.assertRaises(KeyError):
    {}['foo']

(Implemented by Antoine Pitrou; :issue:`4444`.)

Module- and class-level setup and teardown fixtures are now supported. Modules can contain :func:`~unittest.setUpModule` and :func:`~unittest.tearDownModule` functions. Classes can have :meth:`~unittest.TestCase.setUpClass` and :meth:`~unittest.TestCase.tearDownClass` methods that must be defined as class methods (using @classmethod or equivalent). These functions and methods are invoked when the test runner switches to a test case in a different module or class.

The methods :meth:`~unittest.TestCase.addCleanup` and :meth:`~unittest.TestCase.doCleanups` were added. :meth:`~unittest.TestCase.addCleanup` lets you add cleanup functions that will be called unconditionally (after :meth:`~unittest.TestCase.setUp` if :meth:`~unittest.TestCase.setUp` fails, otherwise after :meth:`~unittest.TestCase.tearDown`). This allows for much simpler resource allocation and deallocation during tests (:issue:`5679`).

A number of new methods were added that provide more specialized tests. Many of these methods were written by Google engineers for use in their test suites; Gregory P. Smith, Michael Foord, and GvR worked on merging them into Python's version of :mod:`unittest`.

:func:`unittest.main` now takes an optional exit argument. If False, :func:`~unittest.main` doesn't call :func:`sys.exit`, allowing :func:`~unittest.main` to be used from the interactive interpreter. (Contributed by J. Pablo Fernández; :issue:`3379`.)

:class:`~unittest.TestResult` has new :meth:`~unittest.TestResult.startTestRun` and :meth:`~unittest.TestResult.stopTestRun` methods that are called immediately before and after a test run. (Contributed by Robert Collins; :issue:`5728`.)

With all these changes, the :file:`unittest.py` was becoming awkwardly large, so the module was turned into a package and the code split into several files (by Benjamin Peterson). This doesn't affect how the module is imported or used.

Updated module: ElementTree 1.3

The version of the ElementTree library included with Python was updated to version 1.3. Some of the new features are:

  • The various parsing functions now take a parser keyword argument giving an :class:`~xml.etree.ElementTree.XMLParser` instance that will be used. This makes it possible to override the file's internal encoding:

    p = ET.XMLParser(encoding='utf-8')
    t = ET.XML("""<root/>""", parser=p)
    

    Errors in parsing XML now raise a :exc:`ParseError` exception, whose instances have a :attr:`position` attribute containing a (line, column) tuple giving the location of the problem.

  • ElementTree's code for converting trees to a string has been significantly reworked, making it roughly twice as fast in many cases. The :meth:`ElementTree.write() <xml.etree.ElementTree.ElementTree.write>` and :meth:`Element.write` methods now have a method parameter that can be "xml" (the default), "html", or "text". HTML mode will output empty elements as <empty></empty> instead of <empty/>, and text mode will skip over elements and only output the text chunks. If you set the :attr:`tag` attribute of an element to None but leave its children in place, the element will be omitted when the tree is written out, so you don't need to do more extensive rearrangement to remove a single element.

    Namespace handling has also been improved. All xmlns:<whatever> declarations are now output on the root element, not scattered throughout the resulting XML. You can set the default namespace for a tree by setting the :attr:`default_namespace` attribute and can register new prefixes with :meth:`~xml.etree.ElementTree.register_namespace`. In XML mode, you can use the true/false xml_declaration parameter to suppress the XML declaration.

  • New :class:`~xml.etree.ElementTree.Element` method: :meth:`~xml.etree.ElementTree.Element.extend` appends the items from a sequence to the element's children. Elements themselves behave like sequences, so it's easy to move children from one element to another:

    from xml.etree import ElementTree as ET
    
    t = ET.XML("""<list>
      <item>1</item> <item>2</item>  <item>3</item>
    </list>""")
    new = ET.XML('<root/>')
    new.extend(t)
    
    # Outputs <root><item>1</item>...</root>
    print ET.tostring(new)
    
  • New :class:`Element` method: :meth:`~xml.etree.ElementTree.Element.iter` yields the children of the element as a generator. It's also possible to write for child in elem: to loop over an element's children. The existing method :meth:`getiterator` is now deprecated, as is :meth:`getchildren` which constructs and returns a list of children.

  • New :class:`Element` method: :meth:`~xml.etree.ElementTree.Element.itertext` yields all chunks of text that are descendants of the element. For example:

    t = ET.XML("""<list>
      <item>1</item> <item>2</item>  <item>3</item>
    </list>""")
    
    # Outputs ['\n  ', '1', ' ', '2', '  ', '3', '\n']
    print list(t.itertext())
    
  • Deprecated: using an element as a Boolean (i.e., if elem:) would return true if the element had any children, or false if there were no children. This behaviour is confusing -- None is false, but so is a childless element? -- so it will now trigger a :exc:`FutureWarning`. In your code, you should be explicit: write len(elem) != 0 if you're interested in the number of children, or elem is not None.

Fredrik Lundh develops ElementTree and produced the 1.3 version; you can read his article describing 1.3 at http://effbot.org/zone/elementtree-13-intro.htm. Florent Xicluna updated the version included with Python, after discussions on python-dev and in :issue:`6472`.)

Build and C API Changes

Changes to Python's build process and to the C API include:

  • The latest release of the GNU Debugger, GDB 7, can be scripted using Python. When you begin debugging an executable program P, GDB will look for a file named P-gdb.py and automatically read it. Dave Malcolm contributed a :file:`python-gdb.py` that adds a number of commands useful when debugging Python itself. For example, py-up and py-down go up or down one Python stack frame, which usually corresponds to several C stack frames. py-print prints the value of a Python variable, and py-bt prints the Python stack trace. (Added as a result of :issue:`8032`.)

  • If you use the :file:`.gdbinit` file provided with Python, the "pyo" macro in the 2.7 version now works correctly when the thread being debugged doesn't hold the GIL; the macro now acquires it before printing. (Contributed by Victor Stinner; :issue:`3632`.)

  • :c:func:`Py_AddPendingCall` is now thread-safe, letting any worker thread submit notifications to the main Python thread. This is particularly useful for asynchronous IO operations. (Contributed by Kristján Valur Jónsson; :issue:`4293`.)

  • New function: :c:func:`PyCode_NewEmpty` creates an empty code object; only the filename, function name, and first line number are required. This is useful for extension modules that are attempting to construct a more useful traceback stack. Previously such extensions needed to call :c:func:`PyCode_New`, which had many more arguments. (Added by Jeffrey Yasskin.)

  • New function: :c:func:`PyErr_NewExceptionWithDoc` creates a new exception class, just as the existing :c:func:`PyErr_NewException` does, but takes an extra char * argument containing the docstring for the new exception class. (Added by 'lekma' on the Python bug tracker; :issue:`7033`.)

  • New function: :c:func:`PyFrame_GetLineNumber` takes a frame object and returns the line number that the frame is currently executing. Previously code would need to get the index of the bytecode instruction currently executing, and then look up the line number corresponding to that address. (Added by Jeffrey Yasskin.)

  • New functions: :c:func:`PyLong_AsLongAndOverflow` and :c:func:`PyLong_AsLongLongAndOverflow` approximates a Python long integer as a C :c:type:`long` or :c:type:`long long`. If the number is too large to fit into the output type, an overflow flag is set and returned to the caller. (Contributed by Case Van Horsen; :issue:`7528` and :issue:`7767`.)

  • New function: stemming from the rewrite of string-to-float conversion, a new :c:func:`PyOS_string_to_double` function was added. The old :c:func:`PyOS_ascii_strtod` and :c:func:`PyOS_ascii_atof` functions are now deprecated.

  • New function: :c:func:`PySys_SetArgvEx` sets the value of sys.argv and can optionally update sys.path to include the directory containing the script named by sys.argv[0] depending on the value of an updatepath parameter.

    This function was added to close a security hole for applications that embed Python. The old function, :c:func:`PySys_SetArgv`, would always update sys.path, and sometimes it would add the current directory. This meant that, if you ran an application embedding Python in a directory controlled by someone else, attackers could put a Trojan-horse module in the directory (say, a file named :file:`os.py`) that your application would then import and run.

    If you maintain a C/C++ application that embeds Python, check whether you're calling :c:func:`PySys_SetArgv` and carefully consider whether the application should be using :c:func:`PySys_SetArgvEx` with updatepath set to false.

    Security issue reported as CVE-2008-5983; discussed in :issue:`5753`, and fixed by Antoine Pitrou.

  • New macros: the Python header files now define the following macros: :c:macro:`Py_ISALNUM`, :c:macro:`Py_ISALPHA`, :c:macro:`Py_ISDIGIT`, :c:macro:`Py_ISLOWER`, :c:macro:`Py_ISSPACE`, :c:macro:`Py_ISUPPER`, :c:macro:`Py_ISXDIGIT`, :c:macro:`Py_TOLOWER`, and :c:macro:`Py_TOUPPER`. All of these functions are analogous to the C standard macros for classifying characters, but ignore the current locale setting, because in several places Python needs to analyze characters in a locale-independent way. (Added by Eric Smith; :issue:`5793`.)

  • Removed function: :c:macro:`PyEval_CallObject` is now only available as a macro. A function version was being kept around to preserve ABI linking compatibility, but that was in 1997; it can certainly be deleted by now. (Removed by Antoine Pitrou; :issue:`8276`.)

  • New format codes: the :c:func:`PyFormat_FromString`, :c:func:`PyFormat_FromStringV`, and :c:func:`PyErr_Format` functions now accept %lld and %llu format codes for displaying C's :c:type:`long long` types. (Contributed by Mark Dickinson; :issue:`7228`.)

  • The complicated interaction between threads and process forking has been changed. Previously, the child process created by :func:`os.fork` might fail because the child is created with only a single thread running, the thread performing the :func:`os.fork`. If other threads were holding a lock, such as Python's import lock, when the fork was performed, the lock would still be marked as "held" in the new process. But in the child process nothing would ever release the lock, since the other threads weren't replicated, and the child process would no longer be able to perform imports.

    Python 2.7 acquires the import lock before performing an :func:`os.fork`, and will also clean up any locks created using the :mod:`threading` module. C extension modules that have internal locks, or that call :c:func:`fork()` themselves, will not benefit from this clean-up.

    (Fixed by Thomas Wouters; :issue:`1590864`.)

  • The :c:func:`Py_Finalize` function now calls the internal :func:`threading._shutdown` function; this prevents some exceptions from being raised when an interpreter shuts down. (Patch by Adam Olsen; :issue:`1722344`.)

  • When using the :c:type:`PyMemberDef` structure to define attributes of a type, Python will no longer let you try to delete or set a :const:`T_STRING_INPLACE` attribute.

  • Global symbols defined by the :mod:`ctypes` module are now prefixed with Py, or with _ctypes. (Implemented by Thomas Heller; :issue:`3102`.)

  • New configure option: the :option:`--with-system-expat` switch allows building the :mod:`pyexpat` module to use the system Expat library. (Contributed by Arfrever Frehtes Taifersar Arahesis; :issue:`7609`.)

  • New configure option: the :option:`--with-valgrind` option will now disable the pymalloc allocator, which is difficult for the Valgrind memory-error detector to analyze correctly. Valgrind will therefore be better at detecting memory leaks and overruns. (Contributed by James Henstridge; :issue:`2422`.)

  • New configure option: you can now supply an empty string to :option:`--with-dbmliborder=` in order to disable all of the various DBM modules. (Added by Arfrever Frehtes Taifersar Arahesis; :issue:`6491`.)

  • The :program:`configure` script now checks for floating-point rounding bugs on certain 32-bit Intel chips and defines a :c:macro:`X87_DOUBLE_ROUNDING` preprocessor definition. No code currently uses this definition, but it's available if anyone wishes to use it. (Added by Mark Dickinson; :issue:`2937`.)

    :program:`configure` also now sets a :envvar:`LDCXXSHARED` Makefile variable for supporting C++ linking. (Contributed by Arfrever Frehtes Taifersar Arahesis; :issue:`1222585`.)

  • The build process now creates the necessary files for pkg-config support. (Contributed by Clinton Roy; :issue:`3585`.)

  • The build process now supports Subversion 1.7. (Contributed by Arfrever Frehtes Taifersar Arahesis; :issue:`6094`.)

Capsules

Python 3.1 adds a new C datatype, :c:type:`PyCapsule`, for providing a C API to an extension module. A capsule is essentially the holder of a C void * pointer, and is made available as a module attribute; for example, the :mod:`socket` module's API is exposed as socket.CAPI, and :mod:`unicodedata` exposes ucnhash_CAPI. Other extensions can import the module, access its dictionary to get the capsule object, and then get the void * pointer, which will usually point to an array of pointers to the module's various API functions.

There is an existing data type already used for this, :c:type:`PyCObject`, but it doesn't provide type safety. Evil code written in pure Python could cause a segmentation fault by taking a :c:type:`PyCObject` from module A and somehow substituting it for the :c:type:`PyCObject` in module B. Capsules know their own name, and getting the pointer requires providing the name:

void *vtable;

if (!PyCapsule_IsValid(capsule, "mymodule.CAPI") {
        PyErr_SetString(PyExc_ValueError, "argument type invalid");
        return NULL;
}

vtable = PyCapsule_GetPointer(capsule, "mymodule.CAPI");

You are assured that vtable points to whatever you're expecting. If a different capsule was passed in, :c:func:`PyCapsule_IsValid` would detect the mismatched name and return false. Refer to :ref:`using-capsules` for more information on using these objects.

Python 2.7 now uses capsules internally to provide various extension-module APIs, but the :c:func:`PyCObject_AsVoidPtr` was modified to handle capsules, preserving compile-time compatibility with the :c:type:`CObject` interface. Use of :c:func:`PyCObject_AsVoidPtr` will signal a :exc:`PendingDeprecationWarning`, which is silent by default.

Implemented in Python 3.1 and backported to 2.7 by Larry Hastings; discussed in :issue:`5630`.

Port-Specific Changes: Windows

Port-Specific Changes: Mac OS X

  • The path /Library/Python/2.7/site-packages is now appended to sys.path, in order to share added packages between the system installation and a user-installed copy of the same version. (Changed by Ronald Oussoren; :issue:`4865`.)

Port-Specific Changes: FreeBSD

Other Changes and Fixes

  • Two benchmark scripts, :file:`iobench` and :file:`ccbench`, were added to the :file:`Tools` directory. :file:`iobench` measures the speed of the built-in file I/O objects returned by :func:`open` while performing various operations, and :file:`ccbench` is a concurrency benchmark that tries to measure computing throughput, thread switching latency, and IO processing bandwidth when performing several tasks using a varying number of threads.
  • The :file:`Tools/i18n/msgfmt.py` script now understands plural forms in :file:`.po` files. (Fixed by Martin von Löwis; :issue:`5464`.)
  • When importing a module from a :file:`.pyc` or :file:`.pyo` file with an existing :file:`.py` counterpart, the :attr:`co_filename` attributes of the resulting code objects are overwritten when the original filename is obsolete. This can happen if the file has been renamed, moved, or is accessed through different paths. (Patch by Ziga Seilnacht and Jean-Paul Calderone; :issue:`1180193`.)
  • The :file:`regrtest.py` script now takes a :option:`--randseed=` switch that takes an integer that will be used as the random seed for the :option:`-r` option that executes tests in random order. The :option:`-r` option also reports the seed that was used (Added by Collin Winter.)
  • Another :file:`regrtest.py` switch is :option:`-j`, which takes an integer specifying how many tests run in parallel. This allows reducing the total runtime on multi-core machines. This option is compatible with several other options, including the :option:`-R` switch which is known to produce long runtimes. (Added by Antoine Pitrou, :issue:`6152`.) This can also be used with a new :option:`-F` switch that runs selected tests in a loop until they fail. (Added by Antoine Pitrou; :issue:`7312`.)
  • When executed as a script, the :file:`py_compile.py` module now accepts '-' as an argument, which will read standard input for the list of filenames to be compiled. (Contributed by Piotr Ożarowski; :issue:`8233`.)

Porting to Python 2.7

This section lists previously described changes and other bugfixes that may require changes to your code:

  • The :func:`range` function processes its arguments more consistently; it will now call :meth:`__int__` on non-float, non-integer arguments that are supplied to it. (Fixed by Alexander Belopolsky; :issue:`1533`.)
  • The string :meth:`format` method changed the default precision used for floating-point and complex numbers from 6 decimal places to 12, which matches the precision used by :func:`str`. (Changed by Eric Smith; :issue:`5920`.)
  • Because of an optimization for the :keyword:`with` statement, the special methods :meth:`__enter__` and :meth:`__exit__` must belong to the object's type, and cannot be directly attached to the object's instance. This affects new-style classes (derived from :class:`object`) and C extension types. (:issue:`6101`.)
  • Due to a bug in Python 2.6, the exc_value parameter to :meth:`__exit__` methods was often the string representation of the exception, not an instance. This was fixed in 2.7, so exc_value will be an instance as expected. (Fixed by Florent Xicluna; :issue:`7853`.)
  • When a restricted set of attributes were set using __slots__, deleting an unset attribute would not raise :exc:`AttributeError` as you would expect. Fixed by Benjamin Peterson; :issue:`7604`.)

In the standard library:

  • Operations with :class:`~datetime.datetime` instances that resulted in a year falling outside the supported range didn't always raise :exc:`OverflowError`. Such errors are now checked more carefully and will now raise the exception. (Reported by Mark Leander, patch by Anand B. Pillai and Alexander Belopolsky; :issue:`7150`.)

  • When using :class:`~decimal.Decimal` instances with a string's :meth:`format` method, the default alignment was previously left-alignment. This has been changed to right-alignment, which might change the output of your programs. (Changed by Mark Dickinson; :issue:`6857`.)

    Comparisons involving a signaling NaN value (or sNAN) now signal :const:`~decimal.InvalidOperation` instead of silently returning a true or false value depending on the comparison operator. Quiet NaN values (or NaN) are now hashable. (Fixed by Mark Dickinson; :issue:`7279`.)

  • The ElementTree library, :mod:`xml.etree`, no longer escapes ampersands and angle brackets when outputting an XML processing instruction (which looks like <?xml-stylesheet href="#style1"?>) or comment (which looks like <!-- comment -->). (Patch by Neil Muller; :issue:`2746`.)

  • The :meth:`~StringIO.StringIO.readline` method of :class:`~StringIO.StringIO` objects now does nothing when a negative length is requested, as other file-like objects do. (:issue:`7348`).

  • The :mod:`syslog` module will now use the value of sys.argv[0] as the identifier instead of the previous default value of 'python'. (Changed by Sean Reifschneider; :issue:`8451`.)

  • The :mod:`tarfile` module's default error handling has changed, to no longer suppress fatal errors. The default error level was previously 0, which meant that errors would only result in a message being written to the debug log, but because the debug log is not activated by default, these errors go unnoticed. The default error level is now 1, which raises an exception if there's an error. (Changed by Lars Gustäbel; :issue:`7357`.)

  • The :mod:`urlparse` module's :func:`~urlparse.urlsplit` now handles unknown URL schemes in a fashion compliant with RFC 3986: if the URL is of the form "<something>://...", the text before the :// is treated as the scheme, even if it's a made-up scheme that the module doesn't know about. This change may break code that worked around the old behaviour. For example, Python 2.6.4 or 2.5 will return the following:

    >>> import urlparse
    >>> urlparse.urlsplit('invented://host/filename?query')
    ('invented', '', '//host/filename?query', '', '')
    

    Python 2.7 (and Python 2.6.5) will return:

    >>> import urlparse
    >>> urlparse.urlsplit('invented://host/filename?query')
    ('invented', 'host', '/filename?query', '', '')
    

    (Python 2.7 actually produces slightly different output, since it returns a named tuple instead of a standard tuple.)

For C extensions:

For applications that embed Python:

Acknowledgements

The author would like to thank the following people for offering suggestions, corrections and assistance with various drafts of this article: Nick Coghlan, Philip Jenvey, Ryan Lovett, R. David Murray, Hugh Secker-Walker.

Tip: Filter by directory path e.g. /media app.js to search for public/media/app.js.
Tip: Use camelCasing e.g. ProjME to search for ProjectModifiedEvent.java.
Tip: Filter by extension type e.g. /repo .js to search for all .js files in the /repo directory.
Tip: Separate your search with spaces e.g. /ssh pom.xml to search for src/ssh/pom.xml.
Tip: Use ↑ and ↓ arrow keys to navigate and return to view the file.
Tip: You can also navigate files with Ctrl+j (next) and Ctrl+k (previous) and view the file with Ctrl+o.
Tip: You can also navigate files with Alt+j (next) and Alt+k (previous) and view the file with Alt+o.