1 ****************************
2 What's New in Python 2.6
3 ****************************
5 .. XXX add trademark info for Apple, Microsoft, SourceForge.
7 :Author: A.M. Kuchling (amk at amk.ca)
12 Rules for maintenance:
14 * Anyone can add text to this document. Do not spend very much time
15 on the wording of your changes, because your text will probably
16 get rewritten to some degree.
18 * The maintainer will go through Misc/NEWS periodically and add
19 changes; it's therefore more important to add your changes to
20 Misc/NEWS than to this file.
22 * This is not a complete list of every single change; completeness
23 is the purpose of Misc/NEWS. Some changes I consider too small
24 or esoteric to include. If such a change is added to the text,
25 I'll just remove it. (This is another reason you shouldn't spend
26 too much time on writing your addition.)
28 * If you want to draw your new text to the attention of the
29 maintainer, add 'XXX' to the beginning of the paragraph or
32 * It's OK to just add a fragmentary note about a change. For
33 example: "XXX Describe the transmogrify() function added to the
34 socket module." The maintainer will research the change and
35 write the necessary text.
37 * You can comment out your additions if you like, but it's not
38 necessary (especially when a final release is some months away).
40 * Credit the author of a patch or bugfix. Just the name is
41 sufficient; the e-mail address isn't necessary.
43 * It's helpful to add the bug/patch number in a parenthetical comment.
45 XXX Describe the transmogrify() function added to the socket
47 (Contributed by P.Y. Developer; :issue:`12345`.)
49 This saves the maintainer some effort going through the SVN logs
50 when researching a change.
52 This article explains the new features in Python 2.6, released on October 1
53 2008. The release schedule is described in :pep:`361`.
55 The major theme of Python 2.6 is preparing the migration path to
56 Python 3.0, a major redesign of the language. Whenever possible,
57 Python 2.6 incorporates new features and syntax from 3.0 while
58 remaining compatible with existing code by not removing older features
59 or syntax. When it's not possible to do that, Python 2.6 tries to do
60 what it can, adding compatibility functions in a
61 :mod:`future_builtins` module and a :option:`-3` switch to warn about
62 usages that will become unsupported in 3.0.
64 Some significant new packages have been added to the standard library,
65 such as the :mod:`multiprocessing` and :mod:`json` modules, but
66 there aren't many new features that aren't related to Python 3.0 in
69 Python 2.6 also sees a number of improvements and bugfixes throughout
70 the source. A search through the change logs finds there were 259
71 patches applied and 612 bugs fixed between Python 2.5 and 2.6. Both
72 figures are likely to be underestimates.
74 This article doesn't attempt to provide a complete specification of
75 the new features, but instead provides a convenient overview. For
76 full details, you should refer to the documentation for Python 2.6. If
77 you want to understand the rationale for the design and
78 implementation, refer to the PEP for a particular new feature.
79 Whenever possible, "What's New in Python" links to the bug/patch item
82 .. Compare with previous release in 2 - 3 sentences here.
83 add hyperlink when the documentation becomes available online.
85 .. ========================================================================
86 .. Large, PEP-level features and changes should be described here.
87 .. ========================================================================
92 The development cycle for Python versions 2.6 and 3.0 was
93 synchronized, with the alpha and beta releases for both versions being
94 made on the same days. The development of 3.0 has influenced many
97 Python 3.0 is a far-ranging redesign of Python that breaks
98 compatibility with the 2.x series. This means that existing Python
99 code will need some conversion in order to run on
100 Python 3.0. However, not all the changes in 3.0 necessarily break
101 compatibility. In cases where new features won't cause existing code
102 to break, they've been backported to 2.6 and are described in this
103 document in the appropriate place. Some of the 3.0-derived features
106 * A :meth:`__complex__` method for converting objects to a complex number.
107 * Alternate syntax for catching exceptions: ``except TypeError as exc``.
108 * The addition of :func:`functools.reduce` as a synonym for the built-in
109 :func:`reduce` function.
111 Python 3.0 adds several new built-in functions and changes the
112 semantics of some existing built-ins. Functions that are new in 3.0
113 such as :func:`bin` have simply been added to Python 2.6, but existing
114 built-ins haven't been changed; instead, the :mod:`future_builtins`
115 module has versions with the new 3.0 semantics. Code written to be
116 compatible with 3.0 can do ``from future_builtins import hex, map`` as
119 A new command-line switch, :option:`-3`, enables warnings
120 about features that will be removed in Python 3.0. You can run code
121 with this switch to see how much work will be necessary to port
122 code to 3.0. The value of this switch is available
123 to Python code as the boolean variable :data:`sys.py3kwarning`,
124 and to C extension code as :cdata:`Py_Py3kWarningFlag`.
128 The 3xxx series of PEPs, which contains proposals for Python 3.0.
129 :pep:`3000` describes the development process for Python 3.0.
130 Start with :pep:`3100` that describes the general goals for Python
131 3.0, and then explore the higher-numbered PEPS that propose
135 Changes to the Development Process
136 ==================================================
138 While 2.6 was being developed, the Python development process
139 underwent two significant changes: we switched from SourceForge's
140 issue tracker to a customized Roundup installation, and the
141 documentation was converted from LaTeX to reStructuredText.
144 New Issue Tracker: Roundup
145 --------------------------------------------------
147 For a long time, the Python developers had been growing increasingly
148 annoyed by SourceForge's bug tracker. SourceForge's hosted solution
149 doesn't permit much customization; for example, it wasn't possible to
150 customize the life cycle of issues.
152 The infrastructure committee of the Python Software Foundation
153 therefore posted a call for issue trackers, asking volunteers to set
154 up different products and import some of the bugs and patches from
155 SourceForge. Four different trackers were examined: `Jira
156 <http://www.atlassian.com/software/jira/>`__,
157 `Launchpad <http://www.launchpad.net>`__,
158 `Roundup <http://roundup.sourceforge.net/>`__, and
159 `Trac <http://trac.edgewall.org/>`__.
160 The committee eventually settled on Jira
161 and Roundup as the two candidates. Jira is a commercial product that
162 offers no-cost hosted instances to free-software projects; Roundup
163 is an open-source project that requires volunteers
164 to administer it and a server to host it.
166 After posting a call for volunteers, a new Roundup installation was
167 set up at http://bugs.python.org. One installation of Roundup can
168 host multiple trackers, and this server now also hosts issue trackers
169 for Jython and for the Python web site. It will surely find
170 other uses in the future. Where possible,
171 this edition of "What's New in Python" links to the bug/patch
172 item for each change.
174 Hosting of the Python bug tracker is kindly provided by
175 `Upfront Systems <http://www.upfrontsystems.co.za/>`__
176 of Stellenbosch, South Africa. Martin von Loewis put a
177 lot of effort into importing existing bugs and patches from
178 SourceForge; his scripts for this import operation are at
179 http://svn.python.org/view/tracker/importer/ and may be useful to
180 other projects wishing to move from SourceForge to Roundup.
184 http://bugs.python.org
185 The Python bug tracker.
187 http://bugs.jython.org:
188 The Jython bug tracker.
190 http://roundup.sourceforge.net/
191 Roundup downloads and documentation.
193 http://svn.python.org/view/tracker/importer/
194 Martin von Loewis's conversion scripts.
196 New Documentation Format: reStructuredText Using Sphinx
197 -----------------------------------------------------------
199 The Python documentation was written using LaTeX since the project
200 started around 1989. In the 1980s and early 1990s, most documentation
201 was printed out for later study, not viewed online. LaTeX was widely
202 used because it provided attractive printed output while remaining
203 straightforward to write once the basic rules of the markup were
206 Today LaTeX is still used for writing publications destined for
207 printing, but the landscape for programming tools has shifted. We no
208 longer print out reams of documentation; instead, we browse through it
209 online and HTML has become the most important format to support.
210 Unfortunately, converting LaTeX to HTML is fairly complicated and Fred
211 L. Drake Jr., the long-time Python documentation editor, spent a lot
212 of time maintaining the conversion process. Occasionally people would
213 suggest converting the documentation into SGML and later XML, but
214 performing a good conversion is a major task and no one ever committed
215 the time required to finish the job.
217 During the 2.6 development cycle, Georg Brandl put a lot of effort
218 into building a new toolchain for processing the documentation. The
219 resulting package is called Sphinx, and is available from
220 http://sphinx.pocoo.org/.
222 Sphinx concentrates on HTML output, producing attractively styled and
223 modern HTML; printed output is still supported through conversion to
224 LaTeX. The input format is reStructuredText, a markup syntax
225 supporting custom extensions and directives that is commonly used in
226 the Python community.
228 Sphinx is a standalone package that can be used for writing, and
229 almost two dozen other projects
230 (`listed on the Sphinx web site <http://sphinx.pocoo.org/examples.html>`__)
231 have adopted Sphinx as their documentation tool.
235 :ref:`documenting-index`
236 Describes how to write for Python's documentation.
238 `Sphinx <http://sphinx.pocoo.org/>`__
239 Documentation and code for the Sphinx toolchain.
241 `Docutils <http://docutils.sf.net>`__
242 The underlying reStructuredText parser and toolset.
245 PEP 343: The 'with' statement
246 =============================
248 The previous version, Python 2.5, added the ':keyword:`with`'
249 statement as an optional feature, to be enabled by a ``from __future__
250 import with_statement`` directive. In 2.6 the statement no longer needs to
251 be specially enabled; this means that :keyword:`with` is now always a
252 keyword. The rest of this section is a copy of the corresponding
253 section from the "What's New in Python 2.5" document; if you're
254 familiar with the ':keyword:`with`' statement
255 from Python 2.5, you can skip this section.
257 The ':keyword:`with`' statement clarifies code that previously would use
258 ``try...finally`` blocks to ensure that clean-up code is executed. In this
259 section, I'll discuss the statement as it will commonly be used. In the next
260 section, I'll examine the implementation details and show how to write objects
261 for use with this statement.
263 The ':keyword:`with`' statement is a control-flow structure whose basic
266 with expression [as variable]:
269 The expression is evaluated, and it should result in an object that supports the
270 context management protocol (that is, has :meth:`__enter__` and :meth:`__exit__`
273 The object's :meth:`__enter__` is called before *with-block* is executed and
274 therefore can run set-up code. It also may return a value that is bound to the
275 name *variable*, if given. (Note carefully that *variable* is *not* assigned
276 the result of *expression*.)
278 After execution of the *with-block* is finished, the object's :meth:`__exit__`
279 method is called, even if the block raised an exception, and can therefore run
282 Some standard Python objects now support the context management protocol and can
283 be used with the ':keyword:`with`' statement. File objects are one example::
285 with open('/etc/passwd', 'r') as f:
288 ... more processing code ...
290 After this statement has executed, the file object in *f* will have been
291 automatically closed, even if the :keyword:`for` loop raised an exception part-
292 way through the block.
296 In this case, *f* is the same object created by :func:`open`, because
297 :meth:`file.__enter__` returns *self*.
299 The :mod:`threading` module's locks and condition variables also support the
300 ':keyword:`with`' statement::
302 lock = threading.Lock()
304 # Critical section of code
307 The lock is acquired before the block is executed and always released once the
310 The :func:`localcontext` function in the :mod:`decimal` module makes it easy
311 to save and restore the current decimal context, which encapsulates the desired
312 precision and rounding characteristics for computations::
314 from decimal import Decimal, Context, localcontext
316 # Displays with default precision of 28 digits
320 with localcontext(Context(prec=16)):
321 # All code in this block uses a precision of 16 digits.
322 # The original context is restored on exiting the block.
326 .. _new-26-context-managers:
328 Writing Context Managers
329 ------------------------
331 Under the hood, the ':keyword:`with`' statement is fairly complicated. Most
332 people will only use ':keyword:`with`' in company with existing objects and
333 don't need to know these details, so you can skip the rest of this section if
334 you like. Authors of new objects will need to understand the details of the
335 underlying implementation and should keep reading.
337 A high-level explanation of the context management protocol is:
339 * The expression is evaluated and should result in an object called a "context
340 manager". The context manager must have :meth:`__enter__` and :meth:`__exit__`
343 * The context manager's :meth:`__enter__` method is called. The value returned
344 is assigned to *VAR*. If no ``as VAR`` clause is present, the value is simply
347 * The code in *BLOCK* is executed.
349 * If *BLOCK* raises an exception, the :meth:`__exit__(type, value, traceback)`
350 is called with the exception details, the same values returned by
351 :func:`sys.exc_info`. The method's return value controls whether the exception
352 is re-raised: any false value re-raises the exception, and ``True`` will result
353 in suppressing it. You'll only rarely want to suppress the exception, because
354 if you do the author of the code containing the ':keyword:`with`' statement will
355 never realize anything went wrong.
357 * If *BLOCK* didn't raise an exception, the :meth:`__exit__` method is still
358 called, but *type*, *value*, and *traceback* are all ``None``.
360 Let's think through an example. I won't present detailed code but will only
361 sketch the methods necessary for a database that supports transactions.
363 (For people unfamiliar with database terminology: a set of changes to the
364 database are grouped into a transaction. Transactions can be either committed,
365 meaning that all the changes are written into the database, or rolled back,
366 meaning that the changes are all discarded and the database is unchanged. See
367 any database textbook for more information.)
369 Let's assume there's an object representing a database connection. Our goal will
370 be to let the user write code like this::
372 db_connection = DatabaseConnection()
373 with db_connection as cursor:
374 cursor.execute('insert into ...')
375 cursor.execute('delete from ...')
376 # ... more operations ...
378 The transaction should be committed if the code in the block runs flawlessly or
379 rolled back if there's an exception. Here's the basic interface for
380 :class:`DatabaseConnection` that I'll assume::
382 class DatabaseConnection:
385 "Returns a cursor object and starts a new transaction"
387 "Commits current transaction"
389 "Rolls back current transaction"
391 The :meth:`__enter__` method is pretty easy, having only to start a new
392 transaction. For this application the resulting cursor object would be a useful
393 result, so the method will return it. The user can then add ``as cursor`` to
394 their ':keyword:`with`' statement to bind the cursor to a variable name. ::
396 class DatabaseConnection:
399 # Code to start a new transaction
400 cursor = self.cursor()
403 The :meth:`__exit__` method is the most complicated because it's where most of
404 the work has to be done. The method has to check if an exception occurred. If
405 there was no exception, the transaction is committed. The transaction is rolled
406 back if there was an exception.
408 In the code below, execution will just fall off the end of the function,
409 returning the default value of ``None``. ``None`` is false, so the exception
410 will be re-raised automatically. If you wished, you could be more explicit and
411 add a :keyword:`return` statement at the marked location. ::
413 class DatabaseConnection:
415 def __exit__(self, type, value, tb):
417 # No exception, so commit
420 # Exception occurred, so rollback.
425 .. _module-contextlib:
427 The contextlib module
428 ---------------------
430 The :mod:`contextlib` module provides some functions and a decorator that
431 are useful when writing objects for use with the ':keyword:`with`' statement.
433 The decorator is called :func:`contextmanager`, and lets you write a single
434 generator function instead of defining a new class. The generator should yield
435 exactly one value. The code up to the :keyword:`yield` will be executed as the
436 :meth:`__enter__` method, and the value yielded will be the method's return
437 value that will get bound to the variable in the ':keyword:`with`' statement's
438 :keyword:`as` clause, if any. The code after the :keyword:`yield` will be
439 executed in the :meth:`__exit__` method. Any exception raised in the block will
440 be raised by the :keyword:`yield` statement.
442 Using this decorator, our database example from the previous section
443 could be written as::
445 from contextlib import contextmanager
448 def db_transaction(connection):
449 cursor = connection.cursor()
453 connection.rollback()
458 db = DatabaseConnection()
459 with db_transaction(db) as cursor:
462 The :mod:`contextlib` module also has a :func:`nested(mgr1, mgr2, ...)` function
463 that combines a number of context managers so you don't need to write nested
464 ':keyword:`with`' statements. In this example, the single ':keyword:`with`'
465 statement both starts a database transaction and acquires a thread lock::
467 lock = threading.Lock()
468 with nested (db_transaction(db), lock) as (cursor, locked):
471 Finally, the :func:`closing(object)` function returns *object* so that it can be
472 bound to a variable, and calls ``object.close`` at the end of the block. ::
475 from contextlib import closing
477 with closing(urllib.urlopen('http://www.yahoo.com')) as f:
479 sys.stdout.write(line)
484 :pep:`343` - The "with" statement
485 PEP written by Guido van Rossum and Nick Coghlan; implemented by Mike Bland,
486 Guido van Rossum, and Neal Norwitz. The PEP shows the code generated for a
487 ':keyword:`with`' statement, which can be helpful in learning how the statement
490 The documentation for the :mod:`contextlib` module.
492 .. ======================================================================
496 PEP 366: Explicit Relative Imports From a Main Module
497 ============================================================
499 Python's :option:`-m` switch allows running a module as a script.
500 When you ran a module that was located inside a package, relative
501 imports didn't work correctly.
503 The fix for Python 2.6 adds a :attr:`__package__` attribute to
504 modules. When this attribute is present, relative imports will be
505 relative to the value of this attribute instead of the
506 :attr:`__name__` attribute.
508 PEP 302-style importers can then set :attr:`__package__` as necessary.
509 The :mod:`runpy` module that implements the :option:`-m` switch now
510 does this, so relative imports will now work correctly in scripts
511 running from inside a package.
513 .. ======================================================================
517 PEP 370: Per-user ``site-packages`` Directory
518 =====================================================
520 When you run Python, the module search path ``sys.path`` usually
521 includes a directory whose path ends in ``"site-packages"``. This
522 directory is intended to hold locally-installed packages available to
523 all users using a machine or a particular site installation.
525 Python 2.6 introduces a convention for user-specific site directories.
526 The directory varies depending on the platform:
528 * Unix and Mac OS X: :file:`~/.local/`
529 * Windows: :file:`%APPDATA%/Python`
531 Within this directory, there will be version-specific subdirectories,
532 such as :file:`lib/python2.6/site-packages` on Unix/Mac OS and
533 :file:`Python26/site-packages` on Windows.
535 If you don't like the default directory, it can be overridden by an
536 environment variable. :envvar:`PYTHONUSERBASE` sets the root
537 directory used for all Python versions supporting this feature. On
538 Windows, the directory for application-specific data can be changed by
539 setting the :envvar:`APPDATA` environment variable. You can also
540 modify the :file:`site.py` file for your Python installation.
542 The feature can be disabled entirely by running Python with the
543 :option:`-s` option or setting the :envvar:`PYTHONNOUSERSITE`
544 environment variable.
548 :pep:`370` - Per-user ``site-packages`` Directory
549 PEP written and implemented by Christian Heimes.
552 .. ======================================================================
556 PEP 371: The ``multiprocessing`` Package
557 =====================================================
559 The new :mod:`multiprocessing` package lets Python programs create new
560 processes that will perform a computation and return a result to the
561 parent. The parent and child processes can communicate using queues
562 and pipes, synchronize their operations using locks and semaphores,
563 and can share simple arrays of data.
565 The :mod:`multiprocessing` module started out as an exact emulation of
566 the :mod:`threading` module using processes instead of threads. That
567 goal was discarded along the path to Python 2.6, but the general
568 approach of the module is still similar. The fundamental class
569 is the :class:`Process`, which is passed a callable object and
570 a collection of arguments. The :meth:`start` method
571 sets the callable running in a subprocess, after which you can call
572 the :meth:`is_alive` method to check whether the subprocess is still running
573 and the :meth:`join` method to wait for the process to exit.
575 Here's a simple example where the subprocess will calculate a
576 factorial. The function doing the calculation is written strangely so
577 that it takes significantly longer when the input argument is a
583 from multiprocessing import Process, Queue
586 def factorial(queue, N):
587 "Compute a factorial."
588 # If N is a multiple of 4, this function will take much longer.
590 time.sleep(.05 * N/4)
592 # Calculate the result
594 for i in range(1, N+1):
597 # Put the result on the queue
600 if __name__ == '__main__':
605 p = Process(target=factorial, args=(queue, N))
610 print 'Factorial', N, '=', result
612 A :class:`Queue` is used to communicate the input parameter *N* and
613 the result. The :class:`Queue` object is stored in a global variable.
614 The child process will use the value of the variable when the child
615 was created; because it's a :class:`Queue`, parent and child can use
616 the object to communicate. (If the parent were to change the value of
617 the global variable, the child's value would be unaffected, and vice
620 Two other classes, :class:`Pool` and :class:`Manager`, provide
621 higher-level interfaces. :class:`Pool` will create a fixed number of
622 worker processes, and requests can then be distributed to the workers
623 by calling :meth:`apply` or :meth:`apply_async` to add a single request,
624 and :meth:`map` or :meth:`map_async` to add a number of
625 requests. The following code uses a :class:`Pool` to spread requests
626 across 5 worker processes and retrieve a list of results::
628 from multiprocessing import Pool
630 def factorial(N, dictionary):
631 "Compute a factorial."
634 result = p.map(factorial, range(1, 1000, 10))
638 This produces the following output::
643 8222838654177922817725562880000000
644 33452526613163807108170062053440751665152000000000
647 The other high-level interface, the :class:`Manager` class, creates a
648 separate server process that can hold master copies of Python data
649 structures. Other processes can then access and modify these data
650 structures using proxy objects. The following example creates a
651 shared dictionary by calling the :meth:`dict` method; the worker
652 processes then insert values into the dictionary. (Locking is not
653 done for you automatically, which doesn't matter in this example.
654 :class:`Manager`'s methods also include :meth:`Lock`, :meth:`RLock`,
655 and :meth:`Semaphore` to create shared locks.)
660 from multiprocessing import Pool, Manager
662 def factorial(N, dictionary):
663 "Compute a factorial."
664 # Calculate the result
666 for i in range(1, N+1):
669 # Store result in dictionary
672 if __name__ == '__main__':
675 d = mgr.dict() # Create shared dictionary
677 # Run tasks using the pool
678 for N in range(1, 1000, 10):
679 p.apply_async(factorial, (N, d))
681 # Mark pool as closed -- no more tasks can be added.
684 # Wait for tasks to exit
688 for k, v in sorted(d.items()):
691 This will produce the output::
695 21 51090942171709440000
696 31 8222838654177922817725562880000000
697 41 33452526613163807108170062053440751665152000000000
698 51 15511187532873822802242430164693032110632597200169861120000...
702 The documentation for the :mod:`multiprocessing` module.
704 :pep:`371` - Addition of the multiprocessing package
705 PEP written by Jesse Noller and Richard Oudkerk;
706 implemented by Richard Oudkerk and Jesse Noller.
709 .. ======================================================================
713 PEP 3101: Advanced String Formatting
714 =====================================================
716 In Python 3.0, the `%` operator is supplemented by a more powerful string
717 formatting method, :meth:`format`. Support for the :meth:`str.format` method
718 has been backported to Python 2.6.
720 In 2.6, both 8-bit and Unicode strings have a `.format()` method that
721 treats the string as a template and takes the arguments to be formatted.
722 The formatting template uses curly brackets (`{`, `}`) as special characters::
724 >>> # Substitute positional argument 0 into the string.
725 >>> "User ID: {0}".format("root")
727 >>> # Use the named keyword arguments
728 >>> "User ID: {uid} Last seen: {last_login}".format(
730 ... last_login = "5 Mar 2008 07:20")
731 'User ID: root Last seen: 5 Mar 2008 07:20'
733 Curly brackets can be escaped by doubling them::
735 >>> "Empty dict: {{}}".format()
738 Field names can be integers indicating positional arguments, such as
739 ``{0}``, ``{1}``, etc. or names of keyword arguments. You can also
740 supply compound field names that read attributes or access dictionary keys::
743 >>> print 'Platform: {0.platform}\nPython version: {0.version}'.format(sys)
745 Python version: 2.6a1+ (trunk:61261M, Mar 5 2008, 20:29:41)
746 [GCC 4.0.1 (Apple Computer, Inc. build 5367)]'
749 >>> 'Content-type: {0[.mp4]}'.format(mimetypes.types_map)
750 'Content-type: video/mp4'
752 Note that when using dictionary-style notation such as ``[.mp4]``, you
753 don't need to put any quotation marks around the string; it will look
754 up the value using ``.mp4`` as the key. Strings beginning with a
755 number will be converted to an integer. You can't write more
756 complicated expressions inside a format string.
758 So far we've shown how to specify which field to substitute into the
759 resulting string. The precise formatting used is also controllable by
760 adding a colon followed by a format specifier. For example::
762 >>> # Field 0: left justify, pad to 15 characters
763 >>> # Field 1: right justify, pad to 6 characters
764 >>> fmt = '{0:15} ${1:>6}'
765 >>> fmt.format('Registration', 35)
767 >>> fmt.format('Tutorial', 50)
769 >>> fmt.format('Banquet', 125)
772 Format specifiers can reference other fields through nesting::
776 >>> fmt.format('Invoice #1234', width)
779 >>> fmt.format('Invoice #1234', width)
782 The alignment of a field within the desired width can be specified:
784 ================ ============================================
786 ================ ============================================
787 < (default) Left-align
790 = (For numeric types only) Pad after the sign.
791 ================ ============================================
793 Format specifiers can also include a presentation type, which
794 controls how the value is formatted. For example, floating-point numbers
795 can be formatted as a general number or in exponential notation::
797 >>> '{0:g}'.format(3.75)
799 >>> '{0:e}'.format(3.75)
802 A variety of presentation types are available. Consult the 2.6
803 documentation for a :ref:`complete list <formatstrings>`; here's a sample:
805 ===== ========================================================================
806 ``b`` Binary. Outputs the number in base 2.
807 ``c`` Character. Converts the integer to the corresponding Unicode character
809 ``d`` Decimal Integer. Outputs the number in base 10.
810 ``o`` Octal format. Outputs the number in base 8.
811 ``x`` Hex format. Outputs the number in base 16, using lower-case letters for
813 ``e`` Exponent notation. Prints the number in scientific notation using the
814 letter 'e' to indicate the exponent.
815 ``g`` General format. This prints the number as a fixed-point number, unless
816 the number is too large, in which case it switches to 'e' exponent
818 ``n`` Number. This is the same as 'g' (for floats) or 'd' (for integers),
819 except that it uses the current locale setting to insert the appropriate
820 number separator characters.
821 ``%`` Percentage. Multiplies the number by 100 and displays in fixed ('f')
822 format, followed by a percent sign.
823 ===== ========================================================================
825 Classes and types can define a :meth:`__format__` method to control how they're
826 formatted. It receives a single argument, the format specifier::
828 def __format__(self, format_spec):
829 if isinstance(format_spec, unicode):
830 return unicode(str(self))
834 There's also a :func:`format` built-in that will format a single
835 value. It calls the type's :meth:`__format__` method with the
838 >>> format(75.6564, '.2f')
845 The reference documentation for format fields.
847 :pep:`3101` - Advanced String Formatting
848 PEP written by Talin. Implemented by Eric Smith.
850 .. ======================================================================
854 PEP 3105: ``print`` As a Function
855 =====================================================
857 The ``print`` statement becomes the :func:`print` function in Python 3.0.
858 Making :func:`print` a function makes it possible to replace the function
859 by doing ``def print(...)`` or importing a new function from somewhere else.
861 Python 2.6 has a ``__future__`` import that removes ``print`` as language
862 syntax, letting you use the functional form instead. For example::
864 >>> from __future__ import print_function
865 >>> print('# of entries', len(dictionary), file=sys.stderr)
867 The signature of the new function is::
869 def print(*args, sep=' ', end='\n', file=None)
874 * *args*: positional arguments whose values will be printed out.
875 * *sep*: the separator, which will be printed between arguments.
876 * *end*: the ending text, which will be printed after all of the
877 arguments have been output.
878 * *file*: the file object to which the output will be sent.
882 :pep:`3105` - Make print a function
883 PEP written by Georg Brandl.
885 .. ======================================================================
889 PEP 3110: Exception-Handling Changes
890 =====================================================
892 One error that Python programmers occasionally make
893 is writing the following code::
897 except TypeError, ValueError: # Wrong!
900 The author is probably trying to catch both :exc:`TypeError` and
901 :exc:`ValueError` exceptions, but this code actually does something
902 different: it will catch :exc:`TypeError` and bind the resulting
903 exception object to the local name ``"ValueError"``. The
904 :exc:`ValueError` exception will not be caught at all. The correct
905 code specifies a tuple of exceptions::
909 except (TypeError, ValueError):
912 This error happens because the use of the comma here is ambiguous:
913 does it indicate two different nodes in the parse tree, or a single
916 Python 3.0 makes this unambiguous by replacing the comma with the word
917 "as". To catch an exception and store the exception object in the
918 variable ``exc``, you must write::
922 except TypeError as exc:
925 Python 3.0 will only support the use of "as", and therefore interprets
926 the first example as catching two different exceptions. Python 2.6
927 supports both the comma and "as", so existing code will continue to
928 work. We therefore suggest using "as" when writing new Python code
929 that will only be executed with 2.6.
933 :pep:`3110` - Catching Exceptions in Python 3000
934 PEP written and implemented by Collin Winter.
936 .. ======================================================================
940 PEP 3112: Byte Literals
941 =====================================================
943 Python 3.0 adopts Unicode as the language's fundamental string type and
944 denotes 8-bit literals differently, either as ``b'string'``
945 or using a :class:`bytes` constructor. For future compatibility,
946 Python 2.6 adds :class:`bytes` as a synonym for the :class:`str` type,
947 and it also supports the ``b''`` notation.
950 The 2.6 :class:`str` differs from 3.0's :class:`bytes` type in various
951 ways; most notably, the constructor is completely different. In 3.0,
952 ``bytes([65, 66, 67])`` is 3 elements long, containing the bytes
953 representing ``ABC``; in 2.6, ``bytes([65, 66, 67])`` returns the
954 12-byte string representing the :func:`str` of the list.
956 The primary use of :class:`bytes` in 2.6 will be to write tests of
957 object type such as ``isinstance(x, bytes)``. This will help the 2to3
958 converter, which can't tell whether 2.x code intends strings to
959 contain either characters or 8-bit bytes; you can now
960 use either :class:`bytes` or :class:`str` to represent your intention
961 exactly, and the resulting code will also be correct in Python 3.0.
963 There's also a ``__future__`` import that causes all string literals
964 to become Unicode strings. This means that ``\u`` escape sequences
965 can be used to include Unicode characters::
968 from __future__ import unicode_literals
970 s = ('\u751f\u3080\u304e\u3000\u751f\u3054'
971 '\u3081\u3000\u751f\u305f\u307e\u3054')
973 print len(s) # 12 Unicode characters
975 At the C level, Python 3.0 will rename the existing 8-bit
976 string type, called :ctype:`PyStringObject` in Python 2.x,
977 to :ctype:`PyBytesObject`. Python 2.6 uses ``#define``
978 to support using the names :cfunc:`PyBytesObject`,
979 :cfunc:`PyBytes_Check`, :cfunc:`PyBytes_FromStringAndSize`,
980 and all the other functions and macros used with strings.
982 Instances of the :class:`bytes` type are immutable just
983 as strings are. A new :class:`bytearray` type stores a mutable
986 >>> bytearray([65, 66, 67])
988 >>> b = bytearray(u'\u21ef\u3244', 'utf-8')
990 bytearray(b'\xe2\x87\xaf\xe3\x89\x84')
993 bytearray(b'\xe3\x87\xaf\xe3\x89\x84')
994 >>> unicode(str(b), 'utf-8')
997 Byte arrays support most of the methods of string types, such as
998 :meth:`startswith`/:meth:`endswith`, :meth:`find`/:meth:`rfind`,
999 and some of the methods of lists, such as :meth:`append`,
1000 :meth:`pop`, and :meth:`reverse`.
1004 >>> b = bytearray('ABC')
1006 >>> b.append(ord('e'))
1010 There's also a corresponding C API, with
1011 :cfunc:`PyByteArray_FromObject`,
1012 :cfunc:`PyByteArray_FromStringAndSize`,
1013 and various other functions.
1017 :pep:`3112` - Bytes literals in Python 3000
1018 PEP written by Jason Orendorff; backported to 2.6 by Christian Heimes.
1020 .. ======================================================================
1024 PEP 3116: New I/O Library
1025 =====================================================
1027 Python's built-in file objects support a number of methods, but
1028 file-like objects don't necessarily support all of them. Objects that
1029 imitate files usually support :meth:`read` and :meth:`write`, but they
1030 may not support :meth:`readline`, for example. Python 3.0 introduces
1031 a layered I/O library in the :mod:`io` module that separates buffering
1032 and text-handling features from the fundamental read and write
1035 There are three levels of abstract base classes provided by
1036 the :mod:`io` module:
1038 * :class:`RawIOBase` defines raw I/O operations: :meth:`read`,
1040 :meth:`write`, :meth:`seek`, :meth:`tell`, :meth:`truncate`,
1042 Most of the methods of this class will often map to a single system call.
1043 There are also :meth:`readable`, :meth:`writable`, and :meth:`seekable`
1044 methods for determining what operations a given object will allow.
1046 Python 3.0 has concrete implementations of this class for files and
1047 sockets, but Python 2.6 hasn't restructured its file and socket objects
1050 .. XXX should 2.6 register them in io.py?
1052 * :class:`BufferedIOBase` is an abstract base class that
1053 buffers data in memory to reduce the number of
1054 system calls used, making I/O processing more efficient.
1055 It supports all of the methods of :class:`RawIOBase`,
1056 and adds a :attr:`raw` attribute holding the underlying raw object.
1058 There are five concrete classes implementing this ABC.
1059 :class:`BufferedWriter` and :class:`BufferedReader` are for objects
1060 that support write-only or read-only usage that have a :meth:`seek`
1061 method for random access. :class:`BufferedRandom` objects support
1062 read and write access upon the same underlying stream, and
1063 :class:`BufferedRWPair` is for objects such as TTYs that have both
1064 read and write operations acting upon unconnected streams of data.
1065 The :class:`BytesIO` class supports reading, writing, and seeking
1066 over an in-memory buffer.
1068 * :class:`TextIOBase`: Provides functions for reading and writing
1069 strings (remember, strings will be Unicode in Python 3.0),
1070 and supporting universal newlines. :class:`TextIOBase` defines
1071 the :meth:`readline` method and supports iteration upon
1074 There are two concrete implementations. :class:`TextIOWrapper`
1075 wraps a buffered I/O object, supporting all of the methods for
1076 text I/O and adding a :attr:`buffer` attribute for access
1077 to the underlying object. :class:`StringIO` simply buffers
1078 everything in memory without ever writing anything to disk.
1080 (In Python 2.6, :class:`io.StringIO` is implemented in
1081 pure Python, so it's pretty slow. You should therefore stick with the
1082 existing :mod:`StringIO` module or :mod:`cStringIO` for now. At some
1083 point Python 3.0's :mod:`io` module will be rewritten into C for speed,
1084 and perhaps the C implementation will be backported to the 2.x releases.)
1086 In Python 2.6, the underlying implementations haven't been
1087 restructured to build on top of the :mod:`io` module's classes. The
1088 module is being provided to make it easier to write code that's
1089 forward-compatible with 3.0, and to save developers the effort of writing
1090 their own implementations of buffering and text I/O.
1094 :pep:`3116` - New I/O
1095 PEP written by Daniel Stutzbach, Mike Verdone, and Guido van Rossum.
1096 Code by Guido van Rossum, Georg Brandl, Walter Doerwald,
1097 Jeremy Hylton, Martin von Loewis, Tony Lownds, and others.
1099 .. ======================================================================
1103 PEP 3118: Revised Buffer Protocol
1104 =====================================================
1106 The buffer protocol is a C-level API that lets Python types
1107 exchange pointers into their internal representations. A
1108 memory-mapped file can be viewed as a buffer of characters, for
1109 example, and this lets another module such as :mod:`re`
1110 treat memory-mapped files as a string of characters to be searched.
1112 The primary users of the buffer protocol are numeric-processing
1113 packages such as NumPy, which expose the internal representation
1114 of arrays so that callers can write data directly into an array instead
1115 of going through a slower API. This PEP updates the buffer protocol in light of experience
1116 from NumPy development, adding a number of new features
1117 such as indicating the shape of an array or locking a memory region.
1119 The most important new C API function is
1120 ``PyObject_GetBuffer(PyObject *obj, Py_buffer *view, int flags)``, which
1121 takes an object and a set of flags, and fills in the
1122 ``Py_buffer`` structure with information
1123 about the object's memory representation. Objects
1124 can use this operation to lock memory in place
1125 while an external caller could be modifying the contents,
1126 so there's a corresponding ``PyBuffer_Release(Py_buffer *view)`` to
1127 indicate that the external caller is done.
1129 .. XXX PyObject_GetBuffer not documented in c-api
1131 The *flags* argument to :cfunc:`PyObject_GetBuffer` specifies
1132 constraints upon the memory returned. Some examples are:
1134 * :const:`PyBUF_WRITABLE` indicates that the memory must be writable.
1136 * :const:`PyBUF_LOCK` requests a read-only or exclusive lock on the memory.
1138 * :const:`PyBUF_C_CONTIGUOUS` and :const:`PyBUF_F_CONTIGUOUS`
1139 requests a C-contiguous (last dimension varies the fastest) or
1140 Fortran-contiguous (first dimension varies the fastest) array layout.
1142 Two new argument codes for :cfunc:`PyArg_ParseTuple`,
1143 ``s*`` and ``z*``, return locked buffer objects for a parameter.
1147 :pep:`3118` - Revising the buffer protocol
1148 PEP written by Travis Oliphant and Carl Banks; implemented by
1152 .. ======================================================================
1156 PEP 3119: Abstract Base Classes
1157 =====================================================
1159 Some object-oriented languages such as Java support interfaces,
1160 declaring that a class has a given set of methods or supports a given
1161 access protocol. Abstract Base Classes (or ABCs) are an equivalent
1162 feature for Python. The ABC support consists of an :mod:`abc` module
1163 containing a metaclass called :class:`ABCMeta`, special handling of
1164 this metaclass by the :func:`isinstance` and :func:`issubclass`
1165 built-ins, and a collection of basic ABCs that the Python developers
1166 think will be widely useful. Future versions of Python will probably
1169 Let's say you have a particular class and wish to know whether it supports
1170 dictionary-style access. The phrase "dictionary-style" is vague, however.
1171 It probably means that accessing items with ``obj[1]`` works.
1172 Does it imply that setting items with ``obj[2] = value`` works?
1173 Or that the object will have :meth:`keys`, :meth:`values`, and :meth:`items`
1174 methods? What about the iterative variants such as :meth:`iterkeys`? :meth:`copy`
1175 and :meth:`update`? Iterating over the object with :func:`iter`?
1177 The Python 2.6 :mod:`collections` module includes a number of
1178 different ABCs that represent these distinctions. :class:`Iterable`
1179 indicates that a class defines :meth:`__iter__`, and
1180 :class:`Container` means the class defines a :meth:`__contains__`
1181 method and therefore supports ``x in y`` expressions. The basic
1182 dictionary interface of getting items, setting items, and
1183 :meth:`keys`, :meth:`values`, and :meth:`items`, is defined by the
1184 :class:`MutableMapping` ABC.
1186 You can derive your own classes from a particular ABC
1187 to indicate they support that ABC's interface::
1191 class Storage(collections.MutableMapping):
1195 Alternatively, you could write the class without deriving from
1196 the desired ABC and instead register the class by
1197 calling the ABC's :meth:`register` method::
1204 collections.MutableMapping.register(Storage)
1206 For classes that you write, deriving from the ABC is probably clearer.
1207 The :meth:`register` method is useful when you've written a new
1208 ABC that can describe an existing type or class, or if you want
1209 to declare that some third-party class implements an ABC.
1210 For example, if you defined a :class:`PrintableType` ABC,
1213 # Register Python's types
1214 PrintableType.register(int)
1215 PrintableType.register(float)
1216 PrintableType.register(str)
1218 Classes should obey the semantics specified by an ABC, but
1219 Python can't check this; it's up to the class author to
1220 understand the ABC's requirements and to implement the code accordingly.
1222 To check whether an object supports a particular interface, you can
1226 if not isinstance(d, collections.MutableMapping):
1227 raise ValueError("Mapping object expected, not %r" % d)
1229 Don't feel that you must now begin writing lots of checks as in the
1230 above example. Python has a strong tradition of duck-typing, where
1231 explicit type-checking is never done and code simply calls methods on
1232 an object, trusting that those methods will be there and raising an
1233 exception if they aren't. Be judicious in checking for ABCs and only
1234 do it where it's absolutely necessary.
1236 You can write your own ABCs by using ``abc.ABCMeta`` as the
1237 metaclass in a class definition::
1239 from abc import ABCMeta, abstractmethod
1242 __metaclass__ = ABCMeta
1245 def draw(self, x, y, scale=1.0):
1248 def draw_doubled(self, x, y):
1249 self.draw(x, y, scale=2.0)
1252 class Square(Drawable):
1253 def draw(self, x, y, scale):
1257 In the :class:`Drawable` ABC above, the :meth:`draw_doubled` method
1258 renders the object at twice its size and can be implemented in terms
1259 of other methods described in :class:`Drawable`. Classes implementing
1260 this ABC therefore don't need to provide their own implementation
1261 of :meth:`draw_doubled`, though they can do so. An implementation
1262 of :meth:`draw` is necessary, though; the ABC can't provide
1263 a useful generic implementation.
1265 You can apply the ``@abstractmethod`` decorator to methods such as
1266 :meth:`draw` that must be implemented; Python will then raise an
1267 exception for classes that don't define the method.
1268 Note that the exception is only raised when you actually
1269 try to create an instance of a subclass lacking the method::
1271 >>> class Circle(Drawable):
1275 Traceback (most recent call last):
1276 File "<stdin>", line 1, in <module>
1277 TypeError: Can't instantiate abstract class Circle with abstract methods draw
1280 Abstract data attributes can be declared using the
1281 ``@abstractproperty`` decorator::
1283 from abc import abstractproperty
1290 Subclasses must then define a :meth:`readonly` property.
1294 :pep:`3119` - Introducing Abstract Base Classes
1295 PEP written by Guido van Rossum and Talin.
1296 Implemented by Guido van Rossum.
1297 Backported to 2.6 by Benjamin Aranguren, with Alex Martelli.
1299 .. ======================================================================
1303 PEP 3127: Integer Literal Support and Syntax
1304 =====================================================
1306 Python 3.0 changes the syntax for octal (base-8) integer literals,
1307 prefixing them with "0o" or "0O" instead of a leading zero, and adds
1308 support for binary (base-2) integer literals, signalled by a "0b" or
1311 Python 2.6 doesn't drop support for a leading 0 signalling
1312 an octal number, but it does add support for "0o" and "0b"::
1319 The :func:`oct` built-in still returns numbers
1320 prefixed with a leading zero, and a new :func:`bin`
1321 built-in returns the binary representation for a number::
1325 >>> future_builtins.oct(42)
1330 The :func:`int` and :func:`long` built-ins will now accept the "0o"
1331 and "0b" prefixes when base-8 or base-2 are requested, or when the
1332 *base* argument is zero (signalling that the base used should be
1333 determined from the string)::
1339 >>> int('0b1101', 2)
1341 >>> int('0b1101', 0)
1347 :pep:`3127` - Integer Literal Support and Syntax
1348 PEP written by Patrick Maupin; backported to 2.6 by
1351 .. ======================================================================
1355 PEP 3129: Class Decorators
1356 =====================================================
1358 Decorators have been extended from functions to classes. It's now legal to
1366 This is equivalent to::
1375 :pep:`3129` - Class Decorators
1376 PEP written by Collin Winter.
1378 .. ======================================================================
1382 PEP 3141: A Type Hierarchy for Numbers
1383 =====================================================
1385 Python 3.0 adds several abstract base classes for numeric types
1386 inspired by Scheme's numeric tower. These classes were backported to
1387 2.6 as the :mod:`numbers` module.
1389 The most general ABC is :class:`Number`. It defines no operations at
1390 all, and only exists to allow checking if an object is a number by
1391 doing ``isinstance(obj, Number)``.
1393 :class:`Complex` is a subclass of :class:`Number`. Complex numbers
1394 can undergo the basic operations of addition, subtraction,
1395 multiplication, division, and exponentiation, and you can retrieve the
1396 real and imaginary parts and obtain a number's conjugate. Python's built-in
1397 complex type is an implementation of :class:`Complex`.
1399 :class:`Real` further derives from :class:`Complex`, and adds
1400 operations that only work on real numbers: :func:`floor`, :func:`trunc`,
1401 rounding, taking the remainder mod N, floor division,
1404 :class:`Rational` numbers derive from :class:`Real`, have
1405 :attr:`numerator` and :attr:`denominator` properties, and can be
1406 converted to floats. Python 2.6 adds a simple rational-number class,
1407 :class:`Fraction`, in the :mod:`fractions` module. (It's called
1408 :class:`Fraction` instead of :class:`Rational` to avoid
1409 a name clash with :class:`numbers.Rational`.)
1411 :class:`Integral` numbers derive from :class:`Rational`, and
1412 can be shifted left and right with ``<<`` and ``>>``,
1413 combined using bitwise operations such as ``&`` and ``|``,
1414 and can be used as array indexes and slice boundaries.
1416 In Python 3.0, the PEP slightly redefines the existing built-ins
1417 :func:`round`, :func:`math.floor`, :func:`math.ceil`, and adds a new
1418 one, :func:`math.trunc`, that's been backported to Python 2.6.
1419 :func:`math.trunc` rounds toward zero, returning the closest
1420 :class:`Integral` that's between the function's argument and zero.
1424 :pep:`3141` - A Type Hierarchy for Numbers
1425 PEP written by Jeffrey Yasskin.
1427 `Scheme's numerical tower <http://www.gnu.org/software/guile/manual/html_node/Numerical-Tower.html#Numerical-Tower>`__, from the Guile manual.
1429 `Scheme's number datatypes <http://schemers.org/Documents/Standards/R5RS/HTML/r5rs-Z-H-9.html#%_sec_6.2>`__ from the R5RS Scheme specification.
1432 The :mod:`fractions` Module
1433 --------------------------------------------------
1435 To fill out the hierarchy of numeric types, the :mod:`fractions`
1436 module provides a rational-number class. Rational numbers store their
1437 values as a numerator and denominator forming a fraction, and can
1438 exactly represent numbers such as ``2/3`` that floating-point numbers
1439 can only approximate.
1441 The :class:`Fraction` constructor takes two :class:`Integral` values
1442 that will be the numerator and denominator of the resulting fraction. ::
1444 >>> from fractions import Fraction
1445 >>> a = Fraction(2, 3)
1446 >>> b = Fraction(2, 5)
1447 >>> float(a), float(b)
1448 (0.66666666666666663, 0.40000000000000002)
1454 For converting floating-point numbers to rationals,
1455 the float type now has an :meth:`as_integer_ratio()` method that returns
1456 the numerator and denominator for a fraction that evaluates to the same
1457 floating-point value::
1459 >>> (2.5) .as_integer_ratio()
1461 >>> (3.1415) .as_integer_ratio()
1462 (7074029114692207L, 2251799813685248L)
1463 >>> (1./3) .as_integer_ratio()
1464 (6004799503160661L, 18014398509481984L)
1466 Note that values that can only be approximated by floating-point
1467 numbers, such as 1./3, are not simplified to the number being
1468 approximated; the fraction attempts to match the floating-point value
1471 The :mod:`fractions` module is based upon an implementation by Sjoerd
1472 Mullender that was in Python's :file:`Demo/classes/` directory for a
1473 long time. This implementation was significantly updated by Jeffrey
1477 Other Language Changes
1478 ======================
1480 Some smaller changes made to the core Python language are:
1482 * The :func:`hasattr` function was catching and ignoring all errors,
1483 under the assumption that they meant a :meth:`__getattr__` method
1484 was failing somehow and the return value of :func:`hasattr` would
1485 therefore be ``False``. This logic shouldn't be applied to
1486 :exc:`KeyboardInterrupt` and :exc:`SystemExit`, however; Python 2.6
1487 will no longer discard such exceptions when :func:`hasattr`
1488 encounters them. (Fixed by Benjamin Peterson; :issue:`2196`.)
1490 * When calling a function using the ``**`` syntax to provide keyword
1491 arguments, you are no longer required to use a Python dictionary;
1492 any mapping will now work::
1495 ... print sorted(kw)
1497 >>> ud=UserDict.UserDict()
1499 >>> ud['b'] = 'string'
1503 (Contributed by Alexander Belopolsky; :issue:`1686487`.)
1505 It's also become legal to provide keyword arguments after a ``*args`` argument
1506 to a function call. ::
1508 >>> def f(*args, **kw):
1511 >>> f(1,2,3, *(4,5,6), keyword=13)
1512 (1, 2, 3, 4, 5, 6) {'keyword': 13}
1514 Previously this would have been a syntax error.
1515 (Contributed by Amaury Forgeot d'Arc; :issue:`3473`.)
1517 * A new built-in, ``next(iterator, [default])`` returns the next item
1518 from the specified iterator. If the *default* argument is supplied,
1519 it will be returned if *iterator* has been exhausted; otherwise,
1520 the :exc:`StopIteration` exception will be raised. (Backported
1523 * Tuples now have :meth:`index` and :meth:`count` methods matching the
1524 list type's :meth:`index` and :meth:`count` methods::
1526 >>> t = (0,1,2,3,4,0,1,2)
1532 (Contributed by Raymond Hettinger)
1534 * The built-in types now have improved support for extended slicing syntax,
1535 accepting various combinations of ``(start, stop, step)``.
1536 Previously, the support was partial and certain corner cases wouldn't work.
1537 (Implemented by Thomas Wouters.)
1541 * Properties now have three attributes, :attr:`getter`, :attr:`setter`
1542 and :attr:`deleter`, that are decorators providing useful shortcuts
1543 for adding a getter, setter or deleter function to an existing
1544 property. You would use them like this::
1568 * Several methods of the built-in set types now accept multiple iterables:
1569 :meth:`intersection`,
1570 :meth:`intersection_update`,
1571 :meth:`union`, :meth:`update`,
1572 :meth:`difference` and :meth:`difference_update`.
1576 >>> s=set('1234567890')
1577 >>> s.intersection('abc123', 'cdf246') # Intersection between all inputs
1579 >>> s.difference('246', '789')
1580 set(['1', '0', '3', '5'])
1582 (Contributed by Raymond Hettinger.)
1584 * Many floating-point features were added. The :func:`float` function
1585 will now turn the string ``nan`` into an
1586 IEEE 754 Not A Number value, and ``+inf`` and ``-inf`` into
1587 positive or negative infinity. This works on any platform with
1588 IEEE 754 semantics. (Contributed by Christian Heimes; :issue:`1635`.)
1590 Other functions in the :mod:`math` module, :func:`isinf` and
1591 :func:`isnan`, return true if their floating-point argument is
1592 infinite or Not A Number. (:issue:`1640`)
1594 Conversion functions were added to convert floating-point numbers
1595 into hexadecimal strings (:issue:`3008`). These functions
1596 convert floats to and from a string representation without
1597 introducing rounding errors from the conversion between decimal and
1598 binary. Floats have a :meth:`hex` method that returns a string
1599 representation, and the ``float.fromhex()`` method converts a string
1600 back into a number::
1604 '0x1.e000000000000p+1'
1605 >>> float.fromhex('0x1.e000000000000p+1')
1609 '0x1.5555555555555p-2'
1611 * A numerical nicety: when creating a complex number from two floats
1612 on systems that support signed zeros (-0 and +0), the
1613 :func:`complex` constructor will now preserve the sign
1614 of the zero. (Fixed by Mark T. Dickinson; :issue:`1507`.)
1616 * Classes that inherit a :meth:`__hash__` method from a parent class
1617 can set ``__hash__ = None`` to indicate that the class isn't
1618 hashable. This will make ``hash(obj)`` raise a :exc:`TypeError`
1619 and the class will not be indicated as implementing the
1620 :class:`Hashable` ABC.
1622 You should do this when you've defined a :meth:`__cmp__` or
1623 :meth:`__eq__` method that compares objects by their value rather
1624 than by identity. All objects have a default hash method that uses
1625 ``id(obj)`` as the hash value. There's no tidy way to remove the
1626 :meth:`__hash__` method inherited from a parent class, so
1627 assigning ``None`` was implemented as an override. At the
1628 C level, extensions can set ``tp_hash`` to
1629 :cfunc:`PyObject_HashNotImplemented`.
1630 (Fixed by Nick Coghlan and Amaury Forgeot d'Arc; :issue:`2235`.)
1632 * Changes to the :class:`Exception` interface
1633 as dictated by :pep:`352` continue to be made. For 2.6,
1634 the :attr:`message` attribute is being deprecated in favor of the
1635 :attr:`args` attribute.
1637 * The :exc:`GeneratorExit` exception now subclasses
1638 :exc:`BaseException` instead of :exc:`Exception`. This means
1639 that an exception handler that does ``except Exception:``
1640 will not inadvertently catch :exc:`GeneratorExit`.
1641 (Contributed by Chad Austin; :issue:`1537`.)
1643 * Generator objects now have a :attr:`gi_code` attribute that refers to
1644 the original code object backing the generator.
1645 (Contributed by Collin Winter; :issue:`1473257`.)
1647 * The :func:`compile` built-in function now accepts keyword arguments
1648 as well as positional parameters. (Contributed by Thomas Wouters;
1651 * The :func:`complex` constructor now accepts strings containing
1652 parenthesized complex numbers, meaning that ``complex(repr(cplx))``
1653 will now round-trip values. For example, ``complex('(3+4j)')``
1654 now returns the value (3+4j). (:issue:`1491866`)
1656 * The string :meth:`translate` method now accepts ``None`` as the
1657 translation table parameter, which is treated as the identity
1658 transformation. This makes it easier to carry out operations
1659 that only delete characters. (Contributed by Bengt Richter and
1660 implemented by Raymond Hettinger; :issue:`1193128`.)
1662 * The built-in :func:`dir` function now checks for a :meth:`__dir__`
1663 method on the objects it receives. This method must return a list
1664 of strings containing the names of valid attributes for the object,
1665 and lets the object control the value that :func:`dir` produces.
1666 Objects that have :meth:`__getattr__` or :meth:`__getattribute__`
1667 methods can use this to advertise pseudo-attributes they will honor.
1670 * Instance method objects have new attributes for the object and function
1671 comprising the method; the new synonym for :attr:`im_self` is
1672 :attr:`__self__`, and :attr:`im_func` is also available as :attr:`__func__`.
1673 The old names are still supported in Python 2.6, but are gone in 3.0.
1675 * An obscure change: when you use the :func:`locals` function inside a
1676 :keyword:`class` statement, the resulting dictionary no longer returns free
1677 variables. (Free variables, in this case, are variables referenced in the
1678 :keyword:`class` statement that aren't attributes of the class.)
1680 .. ======================================================================
1686 * The :mod:`warnings` module has been rewritten in C. This makes
1687 it possible to invoke warnings from the parser, and may also
1688 make the interpreter's startup faster.
1689 (Contributed by Neal Norwitz and Brett Cannon; :issue:`1631171`.)
1691 * Type objects now have a cache of methods that can reduce
1692 the work required to find the correct method implementation
1693 for a particular class; once cached, the interpreter doesn't need to
1694 traverse base classes to figure out the right method to call.
1695 The cache is cleared if a base class or the class itself is modified,
1696 so the cache should remain correct even in the face of Python's dynamic
1698 (Original optimization implemented by Armin Rigo, updated for
1699 Python 2.6 by Kevin Jacobs; :issue:`1700288`.)
1701 By default, this change is only applied to types that are included with
1702 the Python core. Extension modules may not necessarily be compatible with
1704 so they must explicitly add :cmacro:`Py_TPFLAGS_HAVE_VERSION_TAG`
1705 to the module's ``tp_flags`` field to enable the method cache.
1706 (To be compatible with the method cache, the extension module's code
1707 must not directly access and modify the ``tp_dict`` member of
1708 any of the types it implements. Most modules don't do this,
1709 but it's impossible for the Python interpreter to determine that.
1710 See :issue:`1878` for some discussion.)
1712 * Function calls that use keyword arguments are significantly faster
1713 by doing a quick pointer comparison, usually saving the time of a
1714 full string comparison. (Contributed by Raymond Hettinger, after an
1715 initial implementation by Antoine Pitrou; :issue:`1819`.)
1717 * All of the functions in the :mod:`struct` module have been rewritten in
1718 C, thanks to work at the Need For Speed sprint.
1719 (Contributed by Raymond Hettinger.)
1721 * Some of the standard built-in types now set a bit in their type
1722 objects. This speeds up checking whether an object is a subclass of
1723 one of these types. (Contributed by Neal Norwitz.)
1725 * Unicode strings now use faster code for detecting
1726 whitespace and line breaks; this speeds up the :meth:`split` method
1727 by about 25% and :meth:`splitlines` by 35%.
1728 (Contributed by Antoine Pitrou.) Memory usage is reduced
1729 by using pymalloc for the Unicode string's data.
1731 * The ``with`` statement now stores the :meth:`__exit__` method on the stack,
1732 producing a small speedup. (Implemented by Jeffrey Yasskin.)
1734 * To reduce memory usage, the garbage collector will now clear internal
1735 free lists when garbage-collecting the highest generation of objects.
1736 This may return memory to the operating system sooner.
1738 .. ======================================================================
1740 .. _new-26-interpreter:
1743 -------------------------------
1745 Two command-line options have been reserved for use by other Python
1746 implementations. The :option:`-J` switch has been reserved for use by
1747 Jython for Jython-specific options, such as switches that are passed to
1748 the underlying JVM. :option:`-X` has been reserved for options
1749 specific to a particular implementation of Python such as CPython,
1750 Jython, or IronPython. If either option is used with Python 2.6, the
1751 interpreter will report that the option isn't currently used.
1753 Python can now be prevented from writing :file:`.pyc` or :file:`.pyo`
1754 files by supplying the :option:`-B` switch to the Python interpreter,
1755 or by setting the :envvar:`PYTHONDONTWRITEBYTECODE` environment
1756 variable before running the interpreter. This setting is available to
1757 Python programs as the ``sys.dont_write_bytecode`` variable, and
1758 Python code can change the value to modify the interpreter's
1759 behaviour. (Contributed by Neal Norwitz and Georg Brandl.)
1761 The encoding used for standard input, output, and standard error can
1762 be specified by setting the :envvar:`PYTHONIOENCODING` environment
1763 variable before running the interpreter. The value should be a string
1764 in the form ``<encoding>`` or ``<encoding>:<errorhandler>``.
1765 The *encoding* part specifies the encoding's name, e.g. ``utf-8`` or
1766 ``latin-1``; the optional *errorhandler* part specifies
1767 what to do with characters that can't be handled by the encoding,
1768 and should be one of "error", "ignore", or "replace". (Contributed
1769 by Martin von Loewis.)
1771 .. ======================================================================
1773 New, Improved, and Deprecated Modules
1774 =====================================
1776 As in every release, Python's standard library received a number of
1777 enhancements and bug fixes. Here's a partial list of the most notable
1778 changes, sorted alphabetically by module name. Consult the
1779 :file:`Misc/NEWS` file in the source tree for a more complete list of
1780 changes, or look through the Subversion logs for all the details.
1782 * (3.0-warning mode) Python 3.0 will feature a reorganized standard
1783 library that will drop many outdated modules and rename others.
1784 Python 2.6 running in 3.0-warning mode will warn about these modules
1785 when they are imported.
1787 The list of deprecated modules is:
1789 :mod:`bgenlocations`,
1791 :mod:`bundlebuilder`,
1797 :mod:`gensuitemodule`,
1801 :mod:`linuxaudiodev`,
1809 :mod:`test.testall`, and
1812 * The :mod:`asyncore` and :mod:`asynchat` modules are
1813 being actively maintained again, and a number of patches and bugfixes
1814 were applied. (Maintained by Josiah Carlson; see :issue:`1736190` for
1817 * The :mod:`bsddb` module also has a new maintainer, Jesús Cea, and the package
1818 is now available as a standalone package. The web page for the package is
1819 `www.jcea.es/programacion/pybsddb.htm
1820 <http://www.jcea.es/programacion/pybsddb.htm>`__.
1821 The plan is to remove the package from the standard library
1822 in Python 3.0, because its pace of releases is much more frequent than
1825 The :mod:`bsddb.dbshelve` module now uses the highest pickling protocol
1826 available, instead of restricting itself to protocol 1.
1827 (Contributed by W. Barnes; :issue:`1551443`.)
1829 * The :mod:`cgi` module will now read variables from the query string
1830 of an HTTP POST request. This makes it possible to use form actions
1831 with URLs that include query strings such as
1832 "/cgi-bin/add.py?category=1". (Contributed by Alexandre Fiori and
1833 Nubis; :issue:`1817`.)
1835 The :func:`parse_qs` and :func:`parse_qsl` functions have been
1836 relocated from the :mod:`cgi` module to the :mod:`urlparse` module.
1837 The versions still available in the :mod:`cgi` module will
1838 trigger :exc:`PendingDeprecationWarning` messages in 2.6
1841 * The :mod:`cmath` module underwent extensive revision,
1842 contributed by Mark Dickinson and Christian Heimes.
1843 Five new functions were added:
1845 * :func:`polar` converts a complex number to polar form, returning
1846 the modulus and argument of the complex number.
1848 * :func:`rect` does the opposite, turning a modulus, argument pair
1849 back into the corresponding complex number.
1851 * :func:`phase` returns the argument (also called the angle) of a complex
1854 * :func:`isnan` returns True if either
1855 the real or imaginary part of its argument is a NaN.
1857 * :func:`isinf` returns True if either the real or imaginary part of
1858 its argument is infinite.
1860 The revisions also improved the numerical soundness of the
1861 :mod:`cmath` module. For all functions, the real and imaginary
1862 parts of the results are accurate to within a few units of least
1863 precision (ulps) whenever possible. See :issue:`1381` for the
1864 details. The branch cuts for :func:`asinh`, :func:`atanh`: and
1865 :func:`atan` have also been corrected.
1867 The tests for the module have been greatly expanded; nearly 2000 new
1868 test cases exercise the algebraic functions.
1870 On IEEE 754 platforms, the :mod:`cmath` module now handles IEEE 754
1871 special values and floating-point exceptions in a manner consistent
1872 with Annex 'G' of the C99 standard.
1874 * A new data type in the :mod:`collections` module: :class:`namedtuple(typename,
1875 fieldnames)` is a factory function that creates subclasses of the standard tuple
1876 whose fields are accessible by name as well as index. For example::
1878 >>> var_type = collections.namedtuple('variable',
1879 ... 'id name type size')
1880 >>> # Names are separated by spaces or commas.
1881 >>> # 'id, name, type, size' would also work.
1882 >>> var_type._fields
1883 ('id', 'name', 'type', 'size')
1885 >>> var = var_type(1, 'frequency', 'int', 4)
1886 >>> print var[0], var.id # Equivalent
1888 >>> print var[2], var.type # Equivalent
1891 {'size': 4, 'type': 'int', 'id': 1, 'name': 'frequency'}
1892 >>> v2 = var._replace(name='amplitude')
1894 variable(id=1, name='amplitude', type='int', size=4)
1896 Several places in the standard library that returned tuples have
1897 been modified to return :class:`namedtuple` instances. For example,
1898 the :meth:`Decimal.as_tuple` method now returns a named tuple with
1899 :attr:`sign`, :attr:`digits`, and :attr:`exponent` fields.
1901 (Contributed by Raymond Hettinger.)
1903 * Another change to the :mod:`collections` module is that the
1904 :class:`deque` type now supports an optional *maxlen* parameter;
1905 if supplied, the deque's size will be restricted to no more
1906 than *maxlen* items. Adding more items to a full deque causes
1907 old items to be discarded.
1911 >>> from collections import deque
1912 >>> dq=deque(maxlen=3)
1915 >>> dq.append(1) ; dq.append(2) ; dq.append(3)
1917 deque([1, 2, 3], maxlen=3)
1920 deque([2, 3, 4], maxlen=3)
1922 (Contributed by Raymond Hettinger.)
1924 * The :mod:`Cookie` module's :class:`Morsel` objects now support an
1925 :attr:`httponly` attribute. In some browsers. cookies with this attribute
1926 set cannot be accessed or manipulated by JavaScript code.
1927 (Contributed by Arvin Schnell; :issue:`1638033`.)
1929 * A new window method in the :mod:`curses` module,
1930 :meth:`chgat`, changes the display attributes for a certain number of
1931 characters on a single line. (Contributed by Fabian Kreutz.)
1935 # Boldface text starting at y=0,x=21
1936 # and affecting the rest of the line.
1937 stdscr.chgat(0, 21, curses.A_BOLD)
1939 The :class:`Textbox` class in the :mod:`curses.textpad` module
1940 now supports editing in insert mode as well as overwrite mode.
1941 Insert mode is enabled by supplying a true value for the *insert_mode*
1942 parameter when creating the :class:`Textbox` instance.
1944 * The :mod:`datetime` module's :meth:`strftime` methods now support a
1945 ``%f`` format code that expands to the number of microseconds in the
1946 object, zero-padded on
1947 the left to six places. (Contributed by Skip Montanaro; :issue:`1158`.)
1949 * The :mod:`decimal` module was updated to version 1.66 of
1950 `the General Decimal Specification <http://www2.hursley.ibm.com/decimal/decarith.html>`__. New features
1951 include some methods for some basic mathematical functions such as
1952 :meth:`exp` and :meth:`log10`::
1954 >>> Decimal(1).exp()
1955 Decimal("2.718281828459045235360287471")
1956 >>> Decimal("2.7182818").ln()
1957 Decimal("0.9999999895305022877376682436")
1958 >>> Decimal(1000).log10()
1961 The :meth:`as_tuple` method of :class:`Decimal` objects now returns a
1962 named tuple with :attr:`sign`, :attr:`digits`, and :attr:`exponent` fields.
1964 (Implemented by Facundo Batista and Mark Dickinson. Named tuple
1965 support added by Raymond Hettinger.)
1967 * The :mod:`difflib` module's :class:`SequenceMatcher` class
1968 now returns named tuples representing matches,
1969 with :attr:`a`, :attr:`b`, and :attr:`size` attributes.
1970 (Contributed by Raymond Hettinger.)
1972 * An optional ``timeout`` parameter, specifying a timeout measured in
1973 seconds, was added to the :class:`ftplib.FTP` class constructor as
1974 well as the :meth:`connect` method. (Added by Facundo Batista.)
1975 Also, the :class:`FTP` class's :meth:`storbinary` and
1976 :meth:`storlines` now take an optional *callback* parameter that
1977 will be called with each block of data after the data has been sent.
1978 (Contributed by Phil Schwartz; :issue:`1221598`.)
1980 * The :func:`reduce` built-in function is also available in the
1981 :mod:`functools` module. In Python 3.0, the built-in has been
1982 dropped and :func:`reduce` is only available from :mod:`functools`;
1983 currently there are no plans to drop the built-in in the 2.x series.
1984 (Patched by Christian Heimes; :issue:`1739906`.)
1986 * When possible, the :mod:`getpass` module will now use
1987 :file:`/dev/tty` to print a prompt message and read the password,
1988 falling back to standard error and standard input. If the
1989 password may be echoed to the terminal, a warning is printed before
1990 the prompt is displayed. (Contributed by Gregory P. Smith.)
1992 * The :func:`glob.glob` function can now return Unicode filenames if
1993 a Unicode path was used and Unicode filenames are matched within the
1994 directory. (:issue:`1001604`)
1996 * The :mod:`gopherlib` module has been removed.
1998 * A new function in the :mod:`heapq` module, ``merge(iter1, iter2, ...)``,
1999 takes any number of iterables returning data in sorted
2000 order, and returns a new generator that returns the contents of all
2001 the iterators, also in sorted order. For example::
2003 >>> list(heapq.merge([1, 3, 5, 9], [2, 8, 16]))
2004 [1, 2, 3, 5, 8, 9, 16]
2006 Another new function, ``heappushpop(heap, item)``,
2007 pushes *item* onto *heap*, then pops off and returns the smallest item.
2008 This is more efficient than making a call to :func:`heappush` and then
2011 :mod:`heapq` is now implemented to only use less-than comparison,
2012 instead of the less-than-or-equal comparison it previously used.
2013 This makes :mod:`heapq`'s usage of a type match the
2014 :meth:`list.sort` method.
2015 (Contributed by Raymond Hettinger.)
2017 * An optional ``timeout`` parameter, specifying a timeout measured in
2018 seconds, was added to the :class:`httplib.HTTPConnection` and
2019 :class:`HTTPSConnection` class constructors. (Added by Facundo
2022 * Most of the :mod:`inspect` module's functions, such as
2023 :func:`getmoduleinfo` and :func:`getargs`, now return named tuples.
2024 In addition to behaving like tuples, the elements of the return value
2025 can also be accessed as attributes.
2026 (Contributed by Raymond Hettinger.)
2028 Some new functions in the module include
2029 :func:`isgenerator`, :func:`isgeneratorfunction`,
2030 and :func:`isabstract`.
2032 * The :mod:`itertools` module gained several new functions.
2034 ``izip_longest(iter1, iter2, ...[, fillvalue])`` makes tuples from
2035 each of the elements; if some of the iterables are shorter than
2036 others, the missing values are set to *fillvalue*. For example::
2038 >>> tuple(itertools.izip_longest([1,2,3], [1,2,3,4,5]))
2039 ((1, 1), (2, 2), (3, 3), (None, 4), (None, 5))
2041 ``product(iter1, iter2, ..., [repeat=N])`` returns the Cartesian product
2042 of the supplied iterables, a set of tuples containing
2043 every possible combination of the elements returned from each iterable. ::
2045 >>> list(itertools.product([1,2,3], [4,5,6]))
2046 [(1, 4), (1, 5), (1, 6),
2047 (2, 4), (2, 5), (2, 6),
2048 (3, 4), (3, 5), (3, 6)]
2050 The optional *repeat* keyword argument is used for taking the
2051 product of an iterable or a set of iterables with themselves,
2052 repeated *N* times. With a single iterable argument, *N*-tuples
2055 >>> list(itertools.product([1,2], repeat=3))
2056 [(1, 1, 1), (1, 1, 2), (1, 2, 1), (1, 2, 2),
2057 (2, 1, 1), (2, 1, 2), (2, 2, 1), (2, 2, 2)]
2059 With two iterables, *2N*-tuples are returned. ::
2061 >>> list(itertools.product([1,2], [3,4], repeat=2))
2062 [(1, 3, 1, 3), (1, 3, 1, 4), (1, 3, 2, 3), (1, 3, 2, 4),
2063 (1, 4, 1, 3), (1, 4, 1, 4), (1, 4, 2, 3), (1, 4, 2, 4),
2064 (2, 3, 1, 3), (2, 3, 1, 4), (2, 3, 2, 3), (2, 3, 2, 4),
2065 (2, 4, 1, 3), (2, 4, 1, 4), (2, 4, 2, 3), (2, 4, 2, 4)]
2067 ``combinations(iterable, r)`` returns sub-sequences of length *r* from
2068 the elements of *iterable*. ::
2070 >>> list(itertools.combinations('123', 2))
2071 [('1', '2'), ('1', '3'), ('2', '3')]
2072 >>> list(itertools.combinations('123', 3))
2074 >>> list(itertools.combinations('1234', 3))
2075 [('1', '2', '3'), ('1', '2', '4'),
2076 ('1', '3', '4'), ('2', '3', '4')]
2078 ``permutations(iter[, r])`` returns all the permutations of length *r* of
2079 the iterable's elements. If *r* is not specified, it will default to the
2080 number of elements produced by the iterable. ::
2082 >>> list(itertools.permutations([1,2,3,4], 2))
2083 [(1, 2), (1, 3), (1, 4),
2084 (2, 1), (2, 3), (2, 4),
2085 (3, 1), (3, 2), (3, 4),
2086 (4, 1), (4, 2), (4, 3)]
2088 ``itertools.chain(*iterables)`` is an existing function in
2089 :mod:`itertools` that gained a new constructor in Python 2.6.
2090 ``itertools.chain.from_iterable(iterable)`` takes a single
2091 iterable that should return other iterables. :func:`chain` will
2092 then return all the elements of the first iterable, then
2093 all the elements of the second, and so on. ::
2095 >>> list(itertools.chain.from_iterable([[1,2,3], [4,5,6]]))
2098 (All contributed by Raymond Hettinger.)
2100 * The :mod:`logging` module's :class:`FileHandler` class
2101 and its subclasses :class:`WatchedFileHandler`, :class:`RotatingFileHandler`,
2102 and :class:`TimedRotatingFileHandler` now
2103 have an optional *delay* parameter to their constructors. If *delay*
2104 is true, opening of the log file is deferred until the first
2105 :meth:`emit` call is made. (Contributed by Vinay Sajip.)
2107 :class:`TimedRotatingFileHandler` also has a *utc* constructor
2108 parameter. If the argument is true, UTC time will be used
2109 in determining when midnight occurs and in generating filenames;
2110 otherwise local time will be used.
2112 * Several new functions were added to the :mod:`math` module:
2114 * :func:`~math.isinf` and :func:`~math.isnan` determine whether a given float
2115 is a (positive or negative) infinity or a NaN (Not a Number), respectively.
2117 * :func:`~math.copysign` copies the sign bit of an IEEE 754 number,
2118 returning the absolute value of *x* combined with the sign bit of
2119 *y*. For example, ``math.copysign(1, -0.0)`` returns -1.0.
2120 (Contributed by Christian Heimes.)
2122 * :func:`~math.factorial` computes the factorial of a number.
2123 (Contributed by Raymond Hettinger; :issue:`2138`.)
2125 * :func:`~math.fsum` adds up the stream of numbers from an iterable,
2126 and is careful to avoid loss of precision through using partial sums.
2127 (Contributed by Jean Brouwers, Raymond Hettinger, and Mark Dickinson;
2130 * :func:`~math.acosh`, :func:`~math.asinh`
2131 and :func:`~math.atanh` compute the inverse hyperbolic functions.
2133 * :func:`~math.log1p` returns the natural logarithm of *1+x*
2136 * :func:`trunc` rounds a number toward zero, returning the closest
2137 :class:`Integral` that's between the function's argument and zero.
2138 Added as part of the backport of
2139 `PEP 3141's type hierarchy for numbers <#pep-3141>`__.
2141 * The :mod:`math` module has been improved to give more consistent
2142 behaviour across platforms, especially with respect to handling of
2143 floating-point exceptions and IEEE 754 special values.
2145 Whenever possible, the module follows the recommendations of the C99
2146 standard about 754's special values. For example, ``sqrt(-1.)``
2147 should now give a :exc:`ValueError` across almost all platforms,
2148 while ``sqrt(float('NaN'))`` should return a NaN on all IEEE 754
2149 platforms. Where Annex 'F' of the C99 standard recommends signaling
2150 'divide-by-zero' or 'invalid', Python will raise :exc:`ValueError`.
2151 Where Annex 'F' of the C99 standard recommends signaling 'overflow',
2152 Python will raise :exc:`OverflowError`. (See :issue:`711019` and
2155 (Contributed by Christian Heimes and Mark Dickinson.)
2157 * The :mod:`MimeWriter` module and :mod:`mimify` module
2158 have been deprecated; use the :mod:`email`
2161 * The :mod:`md5` module has been deprecated; use the :mod:`hashlib` module
2164 * :class:`mmap` objects now have a :meth:`rfind` method that searches for a
2165 substring beginning at the end of the string and searching
2166 backwards. The :meth:`find` method also gained an *end* parameter
2167 giving an index at which to stop searching.
2168 (Contributed by John Lenton.)
2170 * The :mod:`operator` module gained a
2171 :func:`methodcaller` function that takes a name and an optional
2172 set of arguments, returning a callable that will call
2173 the named function on any arguments passed to it. For example::
2175 >>> # Equivalent to lambda s: s.replace('old', 'new')
2176 >>> replacer = operator.methodcaller('replace', 'old', 'new')
2177 >>> replacer('old wine in old bottles')
2178 'new wine in new bottles'
2180 (Contributed by Georg Brandl, after a suggestion by Gregory Petrosyan.)
2182 The :func:`attrgetter` function now accepts dotted names and performs
2183 the corresponding attribute lookups::
2185 >>> inst_name = operator.attrgetter(
2186 ... '__class__.__name__')
2192 (Contributed by Georg Brandl, after a suggestion by Barry Warsaw.)
2194 * The :mod:`os` module now wraps several new system calls.
2195 ``fchmod(fd, mode)`` and ``fchown(fd, uid, gid)`` change the mode
2196 and ownership of an opened file, and ``lchmod(path, mode)`` changes
2197 the mode of a symlink. (Contributed by Georg Brandl and Christian
2200 :func:`chflags` and :func:`lchflags` are wrappers for the
2201 corresponding system calls (where they're available), changing the
2202 flags set on a file. Constants for the flag values are defined in
2203 the :mod:`stat` module; some possible values include
2204 :const:`UF_IMMUTABLE` to signal the file may not be changed and
2205 :const:`UF_APPEND` to indicate that data can only be appended to the
2206 file. (Contributed by M. Levinson.)
2208 ``os.closerange(low, high)`` efficiently closes all file descriptors
2209 from *low* to *high*, ignoring any errors and not including *high* itself.
2210 This function is now used by the :mod:`subprocess` module to make starting
2211 processes faster. (Contributed by Georg Brandl; :issue:`1663329`.)
2213 * The ``os.environ`` object's :meth:`clear` method will now unset the
2214 environment variables using :func:`os.unsetenv` in addition to clearing
2215 the object's keys. (Contributed by Martin Horcicka; :issue:`1181`.)
2217 * The :func:`os.walk` function now has a ``followlinks`` parameter. If
2218 set to True, it will follow symlinks pointing to directories and
2219 visit the directory's contents. For backward compatibility, the
2220 parameter's default value is false. Note that the function can fall
2221 into an infinite recursion if there's a symlink that points to a
2222 parent directory. (:issue:`1273829`)
2224 * In the :mod:`os.path` module, the :func:`splitext` function
2225 has been changed to not split on leading period characters.
2226 This produces better results when operating on Unix's dot-files.
2227 For example, ``os.path.splitext('.ipython')``
2228 now returns ``('.ipython', '')`` instead of ``('', '.ipython')``.
2231 A new function, ``os.path.relpath(path, start='.')``, returns a relative path
2232 from the ``start`` path, if it's supplied, or from the current
2233 working directory to the destination ``path``. (Contributed by
2234 Richard Barran; :issue:`1339796`.)
2236 On Windows, :func:`os.path.expandvars` will now expand environment variables
2237 given in the form "%var%", and "~user" will be expanded into the
2238 user's home directory path. (Contributed by Josiah Carlson;
2241 * The Python debugger provided by the :mod:`pdb` module
2242 gained a new command: "run" restarts the Python program being debugged
2243 and can optionally take new command-line arguments for the program.
2244 (Contributed by Rocky Bernstein; :issue:`1393667`.)
2246 * The :mod:`posixfile` module has been deprecated; :func:`fcntl.lockf`
2247 provides better locking.
2249 The :func:`post_mortem` function, used to begin debugging a
2250 traceback, will now use the traceback returned by :func:`sys.exc_info`
2251 if no traceback is supplied. (Contributed by Facundo Batista;
2254 * The :mod:`pickletools` module now has an :func:`optimize` function
2255 that takes a string containing a pickle and removes some unused
2256 opcodes, returning a shorter pickle that contains the same data structure.
2257 (Contributed by Raymond Hettinger.)
2259 * The :mod:`popen2` module has been deprecated; use the :mod:`subprocess`
2262 * A :func:`get_data` function was added to the :mod:`pkgutil`
2263 module that returns the contents of resource files included
2264 with an installed Python package. For example::
2267 >>> print pkgutil.get_data('test', 'exception_hierarchy.txt')
2270 +-- KeyboardInterrupt
2277 (Contributed by Paul Moore; :issue:`2439`.)
2279 * The :mod:`pyexpat` module's :class:`Parser` objects now allow setting
2280 their :attr:`buffer_size` attribute to change the size of the buffer
2281 used to hold character data.
2282 (Contributed by Achim Gaedke; :issue:`1137`.)
2284 * The :mod:`Queue` module now provides queue variants that retrieve entries
2285 in different orders. The :class:`PriorityQueue` class stores
2286 queued items in a heap and retrieves them in priority order,
2287 and :class:`LifoQueue` retrieves the most recently added entries first,
2288 meaning that it behaves like a stack.
2289 (Contributed by Raymond Hettinger.)
2291 * The :mod:`random` module's :class:`Random` objects can
2292 now be pickled on a 32-bit system and unpickled on a 64-bit
2293 system, and vice versa. Unfortunately, this change also means
2294 that Python 2.6's :class:`Random` objects can't be unpickled correctly
2295 on earlier versions of Python.
2296 (Contributed by Shawn Ligocki; :issue:`1727780`.)
2298 The new ``triangular(low, high, mode)`` function returns random
2299 numbers following a triangular distribution. The returned values
2300 are between *low* and *high*, not including *high* itself, and
2301 with *mode* as the most frequently occurring value
2302 in the distribution. (Contributed by Wladmir van der Laan and
2303 Raymond Hettinger; :issue:`1681432`.)
2305 * Long regular expression searches carried out by the :mod:`re`
2306 module will check for signals being delivered, so
2307 time-consuming searches can now be interrupted.
2308 (Contributed by Josh Hoyt and Ralf Schmitt; :issue:`846388`.)
2310 The regular expression module is implemented by compiling bytecodes
2311 for a tiny regex-specific virtual machine. Untrusted code
2312 could create malicious strings of bytecode directly and cause crashes,
2313 so Python 2.6 includes a verifier for the regex bytecode.
2314 (Contributed by Guido van Rossum from work for Google App Engine;
2317 * The :mod:`rgbimg` module has been removed.
2319 * The :mod:`rlcompleter` module's :meth:`Completer.complete()` method
2320 will now ignore exceptions triggered while evaluating a name.
2321 (Fixed by Lorenz Quack; :issue:`2250`.)
2323 * The :mod:`sched` module's :class:`scheduler` instances now
2324 have a read-only :attr:`queue` attribute that returns the
2325 contents of the scheduler's queue, represented as a list of
2326 named tuples with the fields ``(time, priority, action, argument)``.
2327 (Contributed by Raymond Hettinger; :issue:`1861`.)
2329 * The :mod:`select` module now has wrapper functions
2330 for the Linux :cfunc:`epoll` and BSD :cfunc:`kqueue` system calls.
2331 :meth:`modify` method was added to the existing :class:`poll`
2332 objects; ``pollobj.modify(fd, eventmask)`` takes a file descriptor
2333 or file object and an event mask, modifying the recorded event mask
2335 (Contributed by Christian Heimes; :issue:`1657`.)
2337 * The :mod:`sets` module has been deprecated; it's better to
2338 use the built-in :class:`set` and :class:`frozenset` types.
2340 * The :mod:`sha` module has been deprecated; use the :mod:`hashlib` module
2343 * The :func:`shutil.copytree` function now has an optional *ignore* argument
2344 that takes a callable object. This callable will receive each directory path
2345 and a list of the directory's contents, and returns a list of names that
2346 will be ignored, not copied.
2348 The :mod:`shutil` module also provides an :func:`ignore_patterns`
2349 function for use with this new parameter. :func:`ignore_patterns`
2350 takes an arbitrary number of glob-style patterns and returns a
2351 callable that will ignore any files and directories that match any
2352 of these patterns. The following example copies a directory tree,
2353 but skips both :file:`.svn` directories and Emacs backup files,
2354 which have names ending with '~'::
2356 shutil.copytree('Doc/library', '/tmp/library',
2357 ignore=shutil.ignore_patterns('*~', '.svn'))
2359 (Contributed by Tarek Ziadé; :issue:`2663`.)
2361 * Integrating signal handling with GUI handling event loops
2362 like those used by Tkinter or GTk+ has long been a problem; most
2363 software ends up polling, waking up every fraction of a second to check
2364 if any GUI events have occurred.
2365 The :mod:`signal` module can now make this more efficient.
2366 Calling ``signal.set_wakeup_fd(fd)`` sets a file descriptor
2367 to be used; when a signal is received, a byte is written to that
2368 file descriptor. There's also a C-level function,
2369 :cfunc:`PySignal_SetWakeupFd`, for setting the descriptor.
2371 Event loops will use this by opening a pipe to create two descriptors,
2372 one for reading and one for writing. The writable descriptor
2373 will be passed to :func:`set_wakeup_fd`, and the readable descriptor
2374 will be added to the list of descriptors monitored by the event loop via
2375 :cfunc:`select` or :cfunc:`poll`.
2376 On receiving a signal, a byte will be written and the main event loop
2377 will be woken up, avoiding the need to poll.
2379 (Contributed by Adam Olsen; :issue:`1583`.)
2381 The :func:`siginterrupt` function is now available from Python code,
2382 and allows changing whether signals can interrupt system calls or not.
2383 (Contributed by Ralf Schmitt.)
2385 The :func:`setitimer` and :func:`getitimer` functions have also been
2386 added (where they're available). :func:`setitimer`
2387 allows setting interval timers that will cause a signal to be
2388 delivered to the process after a specified time, measured in
2389 wall-clock time, consumed process time, or combined process+system
2390 time. (Contributed by Guilherme Polo; :issue:`2240`.)
2392 * The :mod:`smtplib` module now supports SMTP over SSL thanks to the
2393 addition of the :class:`SMTP_SSL` class. This class supports an
2394 interface identical to the existing :class:`SMTP` class.
2395 (Contributed by Monty Taylor.) Both class constructors also have an
2396 optional ``timeout`` parameter that specifies a timeout for the
2397 initial connection attempt, measured in seconds. (Contributed by
2400 An implementation of the LMTP protocol (:rfc:`2033`) was also added
2401 to the module. LMTP is used in place of SMTP when transferring
2402 e-mail between agents that don't manage a mail queue. (LMTP
2403 implemented by Leif Hedstrom; :issue:`957003`.)
2405 SMTP.starttls() now complies with :rfc:`3207` and forgets any
2406 knowledge obtained from the server not obtained from the TLS
2407 negotiation itself. (Patch contributed by Bill Fenner;
2410 * The :mod:`socket` module now supports TIPC (http://tipc.sf.net),
2411 a high-performance non-IP-based protocol designed for use in clustered
2412 environments. TIPC addresses are 4- or 5-tuples.
2413 (Contributed by Alberto Bertogli; :issue:`1646`.)
2415 A new function, :func:`create_connection`, takes an address
2416 and connects to it using an optional timeout value, returning
2417 the connected socket object.
2419 * The base classes in the :mod:`SocketServer` module now support
2420 calling a :meth:`handle_timeout` method after a span of inactivity
2421 specified by the server's :attr:`timeout` attribute. (Contributed
2422 by Michael Pomraning.) The :meth:`serve_forever` method
2423 now takes an optional poll interval measured in seconds,
2424 controlling how often the server will check for a shutdown request.
2425 (Contributed by Pedro Werneck and Jeffrey Yasskin;
2426 :issue:`742598`, :issue:`1193577`.)
2428 * The :mod:`sqlite3` module, maintained by Gerhard Haering,
2429 has been updated from version 2.3.2 in Python 2.5 to
2432 * The :mod:`struct` module now supports the C99 :ctype:`_Bool` type,
2433 using the format character ``'?'``.
2434 (Contributed by David Remahl.)
2436 * The :class:`Popen` objects provided by the :mod:`subprocess` module
2437 now have :meth:`terminate`, :meth:`kill`, and :meth:`send_signal` methods.
2438 On Windows, :meth:`send_signal` only supports the :const:`SIGTERM`
2439 signal, and all these methods are aliases for the Win32 API function
2440 :cfunc:`TerminateProcess`.
2441 (Contributed by Christian Heimes.)
2443 * A new variable in the :mod:`sys` module, :attr:`float_info`, is an
2444 object containing information derived from the :file:`float.h` file
2445 about the platform's floating-point support. Attributes of this
2446 object include :attr:`mant_dig` (number of digits in the mantissa),
2447 :attr:`epsilon` (smallest difference between 1.0 and the next
2448 largest value representable), and several others. (Contributed by
2449 Christian Heimes; :issue:`1534`.)
2451 Another new variable, :attr:`dont_write_bytecode`, controls whether Python
2452 writes any :file:`.pyc` or :file:`.pyo` files on importing a module.
2453 If this variable is true, the compiled files are not written. The
2454 variable is initially set on start-up by supplying the :option:`-B`
2455 switch to the Python interpreter, or by setting the
2456 :envvar:`PYTHONDONTWRITEBYTECODE` environment variable before
2457 running the interpreter. Python code can subsequently
2458 change the value of this variable to control whether bytecode files
2460 (Contributed by Neal Norwitz and Georg Brandl.)
2462 Information about the command-line arguments supplied to the Python
2463 interpreter is available by reading attributes of a named
2464 tuple available as ``sys.flags``. For example, the :attr:`verbose`
2465 attribute is true if Python
2466 was executed in verbose mode, :attr:`debug` is true in debugging mode, etc.
2467 These attributes are all read-only.
2468 (Contributed by Christian Heimes.)
2470 A new function, :func:`getsizeof`, takes a Python object and returns
2471 the amount of memory used by the object, measured in bytes. Built-in
2472 objects return correct results; third-party extensions may not,
2473 but can define a :meth:`__sizeof__` method to return the
2475 (Contributed by Robert Schuppenies; :issue:`2898`.)
2477 It's now possible to determine the current profiler and tracer functions
2478 by calling :func:`sys.getprofile` and :func:`sys.gettrace`.
2479 (Contributed by Georg Brandl; :issue:`1648`.)
2481 * The :mod:`tarfile` module now supports POSIX.1-2001 (pax) tarfiles in
2482 addition to the POSIX.1-1988 (ustar) and GNU tar formats that were
2483 already supported. The default format is GNU tar; specify the
2484 ``format`` parameter to open a file using a different format::
2486 tar = tarfile.open("output.tar", "w",
2487 format=tarfile.PAX_FORMAT)
2489 The new ``encoding`` and ``errors`` parameters specify an encoding and
2490 an error handling scheme for character conversions. ``'strict'``,
2491 ``'ignore'``, and ``'replace'`` are the three standard ways Python can
2493 ``'utf-8'`` is a special value that replaces bad characters with
2494 their UTF-8 representation. (Character conversions occur because the
2495 PAX format supports Unicode filenames, defaulting to UTF-8 encoding.)
2497 The :meth:`TarFile.add` method now accepts an ``exclude`` argument that's
2498 a function that can be used to exclude certain filenames from
2500 The function must take a filename and return true if the file
2501 should be excluded or false if it should be archived.
2502 The function is applied to both the name initially passed to :meth:`add`
2503 and to the names of files in recursively-added directories.
2505 (All changes contributed by Lars Gustäbel).
2507 * An optional ``timeout`` parameter was added to the
2508 :class:`telnetlib.Telnet` class constructor, specifying a timeout
2509 measured in seconds. (Added by Facundo Batista.)
2511 * The :class:`tempfile.NamedTemporaryFile` class usually deletes
2512 the temporary file it created when the file is closed. This
2513 behaviour can now be changed by passing ``delete=False`` to the
2514 constructor. (Contributed by Damien Miller; :issue:`1537850`.)
2516 A new class, :class:`SpooledTemporaryFile`, behaves like
2517 a temporary file but stores its data in memory until a maximum size is
2518 exceeded. On reaching that limit, the contents will be written to
2519 an on-disk temporary file. (Contributed by Dustin J. Mitchell.)
2521 The :class:`NamedTemporaryFile` and :class:`SpooledTemporaryFile` classes
2522 both work as context managers, so you can write
2523 ``with tempfile.NamedTemporaryFile() as tmp: ...``.
2524 (Contributed by Alexander Belopolsky; :issue:`2021`.)
2526 * The :mod:`test.test_support` module gained a number
2527 of context managers useful for writing tests.
2528 :func:`EnvironmentVarGuard` is a
2529 context manager that temporarily changes environment variables and
2530 automatically restores them to their old values.
2532 Another context manager, :class:`TransientResource`, can surround calls
2533 to resources that may or may not be available; it will catch and
2534 ignore a specified list of exceptions. For example,
2535 a network test may ignore certain failures when connecting to an
2538 with test_support.TransientResource(IOError,
2539 errno=errno.ETIMEDOUT):
2540 f = urllib.urlopen('https://sf.net')
2543 Finally, :func:`check_warnings` resets the :mod:`warning` module's
2544 warning filters and returns an object that will record all warning
2545 messages triggered (:issue:`3781`)::
2547 with test_support.check_warnings() as wrec:
2548 warnings.simplefilter("always")
2549 # ... code that triggers a warning ...
2550 assert str(wrec.message) == "function is outdated"
2551 assert len(wrec.warnings) == 1, "Multiple warnings raised"
2553 (Contributed by Brett Cannon.)
2555 * The :mod:`textwrap` module can now preserve existing whitespace
2556 at the beginnings and ends of the newly-created lines
2557 by specifying ``drop_whitespace=False``
2560 >>> S = """This sentence has a bunch of
2561 ... extra whitespace."""
2562 >>> print textwrap.fill(S, width=15)
2567 >>> print textwrap.fill(S, drop_whitespace=False, width=15)
2574 (Contributed by Dwayne Bailey; :issue:`1581073`.)
2576 * The :mod:`threading` module API is being changed to use properties
2577 such as :attr:`daemon` instead of :meth:`setDaemon` and
2578 :meth:`isDaemon` methods, and some methods have been renamed to use
2579 underscores instead of camel-case; for example, the
2580 :meth:`activeCount` method is renamed to :meth:`active_count`. Both
2581 the 2.6 and 3.0 versions of the module support the same properties
2582 and renamed methods, but don't remove the old methods. No date has been set
2583 for the deprecation of the old APIs in Python 3.x; the old APIs won't
2584 be removed in any 2.x version.
2585 (Carried out by several people, most notably Benjamin Peterson.)
2587 The :mod:`threading` module's :class:`Thread` objects
2588 gained an :attr:`ident` property that returns the thread's
2589 identifier, a nonzero integer. (Contributed by Gregory P. Smith;
2592 * The :mod:`timeit` module now accepts callables as well as strings
2593 for the statement being timed and for the setup code.
2594 Two convenience functions were added for creating
2595 :class:`Timer` instances:
2596 ``repeat(stmt, setup, time, repeat, number)`` and
2597 ``timeit(stmt, setup, time, number)`` create an instance and call
2598 the corresponding method. (Contributed by Erik Demaine;
2601 * The :mod:`Tkinter` module now accepts lists and tuples for options,
2602 separating the elements by spaces before passing the resulting value to
2604 (Contributed by Guilherme Polo; :issue:`2906`.)
2606 * The :mod:`turtle` module for turtle graphics was greatly enhanced by
2607 Gregor Lingl. New features in the module include:
2609 * Better animation of turtle movement and rotation.
2610 * Control over turtle movement using the new :meth:`delay`,
2611 :meth:`tracer`, and :meth:`speed` methods.
2612 * The ability to set new shapes for the turtle, and to
2613 define a new coordinate system.
2614 * Turtles now have an :meth:`undo()` method that can roll back actions.
2615 * Simple support for reacting to input events such as mouse and keyboard
2616 activity, making it possible to write simple games.
2617 * A :file:`turtle.cfg` file can be used to customize the starting appearance
2618 of the turtle's screen.
2619 * The module's docstrings can be replaced by new docstrings that have been
2620 translated into another language.
2624 * An optional ``timeout`` parameter was added to the
2625 :func:`urllib.urlopen` function and the
2626 :class:`urllib.ftpwrapper` class constructor, as well as the
2627 :func:`urllib2.urlopen` function. The parameter specifies a timeout
2628 measured in seconds. For example::
2630 >>> u = urllib2.urlopen("http://slow.example.com",
2632 Traceback (most recent call last):
2634 urllib2.URLError: <urlopen error timed out>
2637 (Added by Facundo Batista.)
2639 * The Unicode database provided by the :mod:`unicodedata` module
2640 has been updated to version 5.1.0. (Updated by
2641 Martin von Loewis; :issue:`3811`.)
2643 * The :mod:`warnings` module's :func:`formatwarning` and :func:`showwarning`
2644 gained an optional *line* argument that can be used to supply the
2645 line of source code. (Added as part of :issue:`1631171`, which re-implemented
2646 part of the :mod:`warnings` module in C code.)
2648 A new function, :func:`catch_warnings`, is a context manager
2649 intended for testing purposes that lets you temporarily modify the
2650 warning filters and then restore their original values (:issue:`3781`).
2652 * The XML-RPC :class:`SimpleXMLRPCServer` and :class:`DocXMLRPCServer`
2653 classes can now be prevented from immediately opening and binding to
2654 their socket by passing True as the ``bind_and_activate``
2655 constructor parameter. This can be used to modify the instance's
2656 :attr:`allow_reuse_address` attribute before calling the
2657 :meth:`server_bind` and :meth:`server_activate` methods to
2658 open the socket and begin listening for connections.
2659 (Contributed by Peter Parente; :issue:`1599845`.)
2661 :class:`SimpleXMLRPCServer` also has a :attr:`_send_traceback_header`
2662 attribute; if true, the exception and formatted traceback are returned
2663 as HTTP headers "X-Exception" and "X-Traceback". This feature is
2664 for debugging purposes only and should not be used on production servers
2665 because the tracebacks might reveal passwords or other sensitive
2666 information. (Contributed by Alan McIntyre as part of his
2667 project for Google's Summer of Code 2007.)
2669 * The :mod:`xmlrpclib` module no longer automatically converts
2670 :class:`datetime.date` and :class:`datetime.time` to the
2671 :class:`xmlrpclib.DateTime` type; the conversion semantics were
2672 not necessarily correct for all applications. Code using
2673 :mod:`xmlrpclib` should convert :class:`date` and :class:`time`
2674 instances. (:issue:`1330538`) The code can also handle
2675 dates before 1900 (contributed by Ralf Schmitt; :issue:`2014`)
2676 and 64-bit integers represented by using ``<i8>`` in XML-RPC responses
2677 (contributed by Riku Lindblad; :issue:`2985`).
2679 * The :mod:`zipfile` module's :class:`ZipFile` class now has
2680 :meth:`extract` and :meth:`extractall` methods that will unpack
2681 a single file or all the files in the archive to the current directory, or
2682 to a specified directory::
2684 z = zipfile.ZipFile('python-251.zip')
2686 # Unpack a single file, writing it relative
2687 # to the /tmp directory.
2688 z.extract('Python/sysmodule.c', '/tmp')
2690 # Unpack all the files in the archive.
2693 (Contributed by Alan McIntyre; :issue:`467924`.)
2695 The :meth:`open`, :meth:`read` and :meth:`extract` methods can now
2696 take either a filename or a :class:`ZipInfo` object. This is useful when an
2697 archive accidentally contains a duplicated filename.
2698 (Contributed by Graham Horler; :issue:`1775025`.)
2700 Finally, :mod:`zipfile` now supports using Unicode filenames
2701 for archived files. (Contributed by Alexey Borzenkov; :issue:`1734346`.)
2703 .. ======================================================================
2704 .. whole new modules get described in subsections here
2706 The :mod:`ast` module
2707 ----------------------
2709 The :mod:`ast` module provides an Abstract Syntax Tree
2710 representation of Python code, and Armin Ronacher
2711 contributed a set of helper functions that perform a variety of
2712 common tasks. These will be useful for HTML templating
2713 packages, code analyzers, and similar tools that process
2716 The :func:`parse` function takes an expression and returns an AST.
2717 The :func:`dump` function outputs a representation of a tree, suitable
2724 for i in 'abcdefghijklm':
2725 d[i + i] = ord(i) - ord('a') + 1
2730 This outputs a deeply nested tree::
2734 Name(id='d', ctx=Store())
2735 ], value=Dict(keys=[], values=[]))
2736 For(target=Name(id='i', ctx=Store()),
2737 iter=Str(s='abcdefghijklm'), body=[
2740 Name(id='d', ctx=Load()),
2743 BinOp(left=Name(id='i', ctx=Load()), op=Add(),
2744 right=Name(id='i', ctx=Load()))), ctx=Store())
2749 Name(id='ord', ctx=Load()), args=[
2750 Name(id='i', ctx=Load())
2751 ], keywords=[], starargs=None, kwargs=None),
2752 op=Sub(), right=Call(func=
2753 Name(id='ord', ctx=Load()), args=[
2755 ], keywords=[], starargs=None, kwargs=None)),
2756 op=Add(), right=Num(n=1)))
2758 Print(dest=None, values=[
2759 Name(id='d', ctx=Load())
2763 The :func:`literal_eval` method takes a string or an AST
2764 representing a literal expression, parses and evaluates it, and
2765 returns the resulting value. A literal expression is a Python
2766 expression containing only strings, numbers, dictionaries,
2767 etc. but no statements or function calls. If you need to
2768 evaluate an expression but cannot accept the security risk of using an
2769 :func:`eval` call, :func:`literal_eval` will handle it safely::
2771 >>> literal = '("a", "b", {2:4, 3:8, 1:2})'
2772 >>> print ast.literal_eval(literal)
2773 ('a', 'b', {1: 2, 2: 4, 3: 8})
2774 >>> print ast.literal_eval('"a" + "b"')
2775 Traceback (most recent call last):
2777 ValueError: malformed string
2779 The module also includes :class:`NodeVisitor` and
2780 :class:`NodeTransformer` classes for traversing and modifying an AST,
2781 and functions for common transformations such as changing line
2784 .. ======================================================================
2786 The :mod:`future_builtins` module
2787 --------------------------------------
2789 Python 3.0 makes many changes to the repertoire of built-in
2790 functions, and most of the changes can't be introduced in the Python
2791 2.x series because they would break compatibility.
2792 The :mod:`future_builtins` module provides versions
2793 of these built-in functions that can be imported when writing
2794 3.0-compatible code.
2796 The functions in this module currently include:
2798 * ``ascii(obj)``: equivalent to :func:`repr`. In Python 3.0,
2799 :func:`repr` will return a Unicode string, while :func:`ascii` will
2800 return a pure ASCII bytestring.
2802 * ``filter(predicate, iterable)``,
2803 ``map(func, iterable1, ...)``: the 3.0 versions
2804 return iterators, unlike the 2.x built-ins which return lists.
2806 * ``hex(value)``, ``oct(value)``: instead of calling the
2807 :meth:`__hex__` or :meth:`__oct__` methods, these versions will
2808 call the :meth:`__index__` method and convert the result to hexadecimal
2809 or octal. :func:`oct` will use the new ``0o`` notation for its
2812 .. ======================================================================
2814 The :mod:`json` module: JavaScript Object Notation
2815 --------------------------------------------------------------------
2817 The new :mod:`json` module supports the encoding and decoding of Python types in
2818 JSON (Javascript Object Notation). JSON is a lightweight interchange format
2819 often used in web applications. For more information about JSON, see
2820 http://www.json.org.
2822 :mod:`json` comes with support for decoding and encoding most builtin Python
2823 types. The following example encodes and decodes a dictionary::
2826 >>> data = {"spam" : "foo", "parrot" : 42}
2827 >>> in_json = json.dumps(data) # Encode the data
2829 '{"parrot": 42, "spam": "foo"}'
2830 >>> json.loads(in_json) # Decode into a Python object
2831 {"spam" : "foo", "parrot" : 42}
2833 It's also possible to write your own decoders and encoders to support
2834 more types. Pretty-printing of the JSON strings is also supported.
2836 :mod:`json` (originally called simplejson) was written by Bob
2840 .. ======================================================================
2842 The :mod:`plistlib` module: A Property-List Parser
2843 --------------------------------------------------
2845 The ``.plist`` format is commonly used on Mac OS X to
2846 store basic data types (numbers, strings, lists,
2847 and dictionaries) by serializing them into an XML-based format.
2848 It resembles the XML-RPC serialization of data types.
2850 Despite being primarily used on Mac OS X, the format
2851 has nothing Mac-specific about it and the Python implementation works
2852 on any platform that Python supports, so the :mod:`plistlib` module
2853 has been promoted to the standard library.
2855 Using the module is simple::
2861 # Create data structure
2862 data_struct = dict(lastAccessed=datetime.datetime.now(),
2864 categories=('Personal','Shared','Private'))
2866 # Create string containing XML.
2867 plist_str = plistlib.writePlistToString(data_struct)
2868 new_struct = plistlib.readPlistFromString(plist_str)
2872 # Write data structure to a file and read it back.
2873 plistlib.writePlist(data_struct, '/tmp/customizations.plist')
2874 new_struct = plistlib.readPlist('/tmp/customizations.plist')
2876 # read/writePlist accepts file-like objects as well as paths.
2877 plistlib.writePlist(data_struct, sys.stdout)
2879 .. ======================================================================
2882 --------------------------------------------------
2884 Thomas Heller continued to maintain and enhance the
2885 :mod:`ctypes` module.
2887 :mod:`ctypes` now supports a :class:`c_bool` datatype
2888 that represents the C99 ``bool`` type. (Contributed by David Remahl;
2891 The :mod:`ctypes` string, buffer and array types have improved
2892 support for extended slicing syntax,
2893 where various combinations of ``(start, stop, step)`` are supplied.
2894 (Implemented by Thomas Wouters.)
2898 All :mod:`ctypes` data types now support
2899 :meth:`from_buffer` and :meth:`from_buffer_copy`
2900 methods that create a ctypes instance based on a
2901 provided buffer object. :meth:`from_buffer_copy` copies
2902 the contents of the object,
2903 while :meth:`from_buffer` will share the same memory area.
2905 A new calling convention tells :mod:`ctypes` to clear the ``errno`` or
2906 Win32 LastError variables at the outset of each wrapped call.
2907 (Implemented by Thomas Heller; :issue:`1798`.)
2909 You can now retrieve the Unix ``errno`` variable after a function
2910 call. When creating a wrapped function, you can supply
2911 ``use_errno=True`` as a keyword parameter to the :func:`DLL` function
2912 and then call the module-level methods :meth:`set_errno` and
2913 :meth:`get_errno` to set and retrieve the error value.
2915 The Win32 LastError variable is similarly supported by
2916 the :func:`DLL`, :func:`OleDLL`, and :func:`WinDLL` functions.
2917 You supply ``use_last_error=True`` as a keyword parameter
2918 and then call the module-level methods :meth:`set_last_error`
2919 and :meth:`get_last_error`.
2921 The :func:`byref` function, used to retrieve a pointer to a ctypes
2922 instance, now has an optional *offset* parameter that is a byte
2923 count that will be added to the returned pointer.
2925 .. ======================================================================
2927 Improved SSL Support
2928 --------------------------------------------------
2930 Bill Janssen made extensive improvements to Python 2.6's support for
2931 the Secure Sockets Layer by adding a new module, :mod:`ssl`, that's
2932 built atop the `OpenSSL <http://www.openssl.org/>`__ library.
2933 This new module provides more control over the protocol negotiated,
2934 the X.509 certificates used, and has better support for writing SSL
2935 servers (as opposed to clients) in Python. The existing SSL support
2936 in the :mod:`socket` module hasn't been removed and continues to work,
2937 though it will be removed in Python 3.0.
2939 To use the new module, you must first create a TCP connection in the
2940 usual way and then pass it to the :func:`ssl.wrap_socket` function.
2941 It's possible to specify whether a certificate is required, and to
2942 obtain certificate info by calling the :meth:`getpeercert` method.
2946 The documentation for the :mod:`ssl` module.
2948 .. ======================================================================
2951 Build and C API Changes
2952 =======================
2954 Changes to Python's build process and to the C API include:
2956 * Python now must be compiled with C89 compilers (after 19
2957 years!). This means that the Python source tree has dropped its
2958 own implementations of :cfunc:`memmove` and :cfunc:`strerror`, which
2959 are in the C89 standard library.
2961 * Python 2.6 can be built with Microsoft Visual Studio 2008 (version
2962 9.0), and this is the new default compiler. See the
2963 :file:`PCbuild` directory for the build files. (Implemented by
2966 * On Mac OS X, Python 2.6 can be compiled as a 4-way universal build.
2967 The :program:`configure` script
2968 can take a :option:`--with-universal-archs=[32-bit|64-bit|all]`
2969 switch, controlling whether the binaries are built for 32-bit
2970 architectures (x86, PowerPC), 64-bit (x86-64 and PPC-64), or both.
2971 (Contributed by Ronald Oussoren.)
2973 * The BerkeleyDB module now has a C API object, available as
2974 ``bsddb.db.api``. This object can be used by other C extensions
2975 that wish to use the :mod:`bsddb` module for their own purposes.
2976 (Contributed by Duncan Grisby; :issue:`1551895`.)
2978 * The new buffer interface, previously described in
2979 `the PEP 3118 section <#pep-3118-revised-buffer-protocol>`__,
2980 adds :cfunc:`PyObject_GetBuffer` and :cfunc:`PyBuffer_Release`,
2981 as well as a few other functions.
2983 * Python's use of the C stdio library is now thread-safe, or at least
2984 as thread-safe as the underlying library is. A long-standing potential
2985 bug occurred if one thread closed a file object while another thread
2986 was reading from or writing to the object. In 2.6 file objects
2987 have a reference count, manipulated by the
2988 :cfunc:`PyFile_IncUseCount` and :cfunc:`PyFile_DecUseCount`
2989 functions. File objects can't be closed unless the reference count
2990 is zero. :cfunc:`PyFile_IncUseCount` should be called while the GIL
2991 is still held, before carrying out an I/O operation using the
2992 ``FILE *`` pointer, and :cfunc:`PyFile_DecUseCount` should be called
2993 immediately after the GIL is re-acquired.
2994 (Contributed by Antoine Pitrou and Gregory P. Smith.)
2996 * Importing modules simultaneously in two different threads no longer
2997 deadlocks; it will now raise an :exc:`ImportError`. A new API
2998 function, :cfunc:`PyImport_ImportModuleNoBlock`, will look for a
2999 module in ``sys.modules`` first, then try to import it after
3000 acquiring an import lock. If the import lock is held by another
3001 thread, an :exc:`ImportError` is raised.
3002 (Contributed by Christian Heimes.)
3004 * Several functions return information about the platform's
3005 floating-point support. :cfunc:`PyFloat_GetMax` returns
3006 the maximum representable floating point value,
3007 and :cfunc:`PyFloat_GetMin` returns the minimum
3008 positive value. :cfunc:`PyFloat_GetInfo` returns an object
3009 containing more information from the :file:`float.h` file, such as
3010 ``"mant_dig"`` (number of digits in the mantissa), ``"epsilon"``
3011 (smallest difference between 1.0 and the next largest value
3012 representable), and several others.
3013 (Contributed by Christian Heimes; :issue:`1534`.)
3015 * C functions and methods that use
3016 :cfunc:`PyComplex_AsCComplex` will now accept arguments that
3017 have a :meth:`__complex__` method. In particular, the functions in the
3018 :mod:`cmath` module will now accept objects with this method.
3019 This is a backport of a Python 3.0 change.
3020 (Contributed by Mark Dickinson; :issue:`1675423`.)
3022 * Python's C API now includes two functions for case-insensitive string
3023 comparisons, ``PyOS_stricmp(char*, char*)``
3024 and ``PyOS_strnicmp(char*, char*, Py_ssize_t)``.
3025 (Contributed by Christian Heimes; :issue:`1635`.)
3027 * Many C extensions define their own little macro for adding
3028 integers and strings to the module's dictionary in the
3029 ``init*`` function. Python 2.6 finally defines standard macros
3030 for adding values to a module, :cmacro:`PyModule_AddStringMacro`
3031 and :cmacro:`PyModule_AddIntMacro()`. (Contributed by
3034 * Some macros were renamed in both 3.0 and 2.6 to make it clearer that
3036 not functions. :cmacro:`Py_Size()` became :cmacro:`Py_SIZE()`,
3037 :cmacro:`Py_Type()` became :cmacro:`Py_TYPE()`, and
3038 :cmacro:`Py_Refcnt()` became :cmacro:`Py_REFCNT()`.
3039 The mixed-case macros are still available
3040 in Python 2.6 for backward compatibility.
3043 * Distutils now places C extensions it builds in a
3044 different directory when running on a debug version of Python.
3045 (Contributed by Collin Winter; :issue:`1530959`.)
3047 * Several basic data types, such as integers and strings, maintain
3048 internal free lists of objects that can be re-used. The data
3049 structures for these free lists now follow a naming convention: the
3050 variable is always named ``free_list``, the counter is always named
3051 ``numfree``, and a macro ``Py<typename>_MAXFREELIST`` is
3054 * A new Makefile target, "make patchcheck", prepares the Python source tree
3055 for making a patch: it fixes trailing whitespace in all modified
3056 ``.py`` files, checks whether the documentation has been changed,
3057 and reports whether the :file:`Misc/ACKS` and :file:`Misc/NEWS` files
3059 (Contributed by Brett Cannon.)
3061 Another new target, "make profile-opt", compiles a Python binary
3062 using GCC's profile-guided optimization. It compiles Python with
3063 profiling enabled, runs the test suite to obtain a set of profiling
3064 results, and then compiles using these results for optimization.
3065 (Contributed by Gregory P. Smith.)
3067 .. ======================================================================
3069 Port-Specific Changes: Windows
3070 -----------------------------------
3072 * The support for Windows 95, 98, ME and NT4 has been dropped.
3073 Python 2.6 requires at least Windows 2000 SP4.
3075 * The new default compiler on Windows is Visual Studio 2008 (version
3076 9.0). The build directories for Visual Studio 2003 (version 7.1) and
3077 2005 (version 8.0) were moved into the PC/ directory. The new
3078 :file:`PCbuild` directory supports cross compilation for X64, debug
3079 builds and Profile Guided Optimization (PGO). PGO builds are roughly
3080 10% faster than normal builds. (Contributed by Christian Heimes
3081 with help from Amaury Forgeot d'Arc and Martin von Loewis.)
3083 * The :mod:`msvcrt` module now supports
3084 both the normal and wide char variants of the console I/O
3085 API. The :func:`getwch` function reads a keypress and returns a Unicode
3086 value, as does the :func:`getwche` function. The :func:`putwch` function
3087 takes a Unicode character and writes it to the console.
3088 (Contributed by Christian Heimes.)
3090 * :func:`os.path.expandvars` will now expand environment variables in
3091 the form "%var%", and "~user" will be expanded into the user's home
3092 directory path. (Contributed by Josiah Carlson; :issue:`957650`.)
3094 * The :mod:`socket` module's socket objects now have an
3095 :meth:`ioctl` method that provides a limited interface to the
3096 :cfunc:`WSAIoctl` system interface.
3098 * The :mod:`_winreg` module now has a function,
3099 :func:`ExpandEnvironmentStrings`,
3100 that expands environment variable references such as ``%NAME%``
3101 in an input string. The handle objects provided by this
3102 module now support the context protocol, so they can be used
3103 in :keyword:`with` statements. (Contributed by Christian Heimes.)
3105 :mod:`_winreg` also has better support for x64 systems,
3106 exposing the :func:`DisableReflectionKey`, :func:`EnableReflectionKey`,
3107 and :func:`QueryReflectionKey` functions, which enable and disable
3108 registry reflection for 32-bit processes running on 64-bit systems.
3111 * The :mod:`msilib` module's :class:`Record` object
3112 gained :meth:`GetInteger` and :meth:`GetString` methods that
3113 return field values as an integer or a string.
3114 (Contributed by Floris Bruynooghe; :issue:`2125`.)
3116 .. ======================================================================
3118 Port-Specific Changes: Mac OS X
3119 -----------------------------------
3121 * When compiling a framework build of Python, you can now specify the
3122 framework name to be used by providing the
3123 :option:`--with-framework-name=` option to the
3124 :program:`configure` script.
3126 * The :mod:`macfs` module has been removed. This in turn required the
3127 :func:`macostools.touched` function to be removed because it depended on the
3128 :mod:`macfs` module. (:issue:`1490190`)
3130 * Many other Mac OS modules have been deprecated and will removed in
3132 :mod:`_builtinSuites`,
3137 :mod:`appletrawmain`,
3138 :mod:`appletrunner`,
3139 :mod:`argvemulator`,
3162 :mod:`OSATerminology`,
3164 :mod:`PixMapWrapper`,
3166 :mod:`SystemEvents`,
3167 :mod:`Terminal`, and
3168 :mod:`terminalcommand`.
3170 .. ======================================================================
3172 Port-Specific Changes: IRIX
3173 -----------------------------------
3175 A number of old IRIX-specific modules were deprecated and will
3176 be removed in Python 3.0:
3177 :mod:`al` and :mod:`AL`,
3181 :mod:`CL` and :mod:`cl`,
3185 :mod:`FL` and :mod:`fl`,
3190 :mod:`GL` and :mod:`gl`,
3196 :mod:`SV` and :mod:`sv`,
3198 :mod:`videoreader`, and
3201 .. ======================================================================
3204 Porting to Python 2.6
3205 =====================
3207 This section lists previously described changes and other bugfixes
3208 that may require changes to your code:
3210 * Classes that aren't supposed to be hashable should
3211 set ``__hash__ = None`` in their definitions to indicate
3214 * String exceptions have been removed. Attempting to use them raises a
3217 * The :meth:`__init__` method of :class:`collections.deque`
3218 now clears any existing contents of the deque
3219 before adding elements from the iterable. This change makes the
3220 behavior match ``list.__init__()``.
3222 * :meth:`object.__init__` previously accepted arbitrary arguments and
3223 keyword arguments, ignoring them. In Python 2.6, this is no longer
3224 allowed and will result in a :exc:`TypeError`. This will affect
3225 :meth:`__init__` methods that end up calling the corresponding
3226 method on :class:`object` (perhaps through using :func:`super`).
3227 See :issue:`1683368` for discussion.
3229 * The :class:`Decimal` constructor now accepts leading and trailing
3230 whitespace when passed a string. Previously it would raise an
3231 :exc:`InvalidOperation` exception. On the other hand, the
3232 :meth:`create_decimal` method of :class:`Context` objects now
3233 explicitly disallows extra whitespace, raising a
3234 :exc:`ConversionSyntax` exception.
3236 * Due to an implementation accident, if you passed a file path to
3237 the built-in :func:`__import__` function, it would actually import
3238 the specified file. This was never intended to work, however, and
3239 the implementation now explicitly checks for this case and raises
3240 an :exc:`ImportError`.
3242 * C API: the :cfunc:`PyImport_Import` and :cfunc:`PyImport_ImportModule`
3243 functions now default to absolute imports, not relative imports.
3244 This will affect C extensions that import other modules.
3246 * C API: extension data types that shouldn't be hashable
3247 should define their ``tp_hash`` slot to
3248 :cfunc:`PyObject_HashNotImplemented`.
3250 * The :mod:`socket` module exception :exc:`socket.error` now inherits
3251 from :exc:`IOError`. Previously it wasn't a subclass of
3252 :exc:`StandardError` but now it is, through :exc:`IOError`.
3253 (Implemented by Gregory P. Smith; :issue:`1706815`.)
3255 * The :mod:`xmlrpclib` module no longer automatically converts
3256 :class:`datetime.date` and :class:`datetime.time` to the
3257 :class:`xmlrpclib.DateTime` type; the conversion semantics were
3258 not necessarily correct for all applications. Code using
3259 :mod:`xmlrpclib` should convert :class:`date` and :class:`time`
3260 instances. (:issue:`1330538`)
3262 * (3.0-warning mode) The :class:`Exception` class now warns
3263 when accessed using slicing or index access; having
3264 :class:`Exception` behave like a tuple is being phased out.
3266 * (3.0-warning mode) inequality comparisons between two dictionaries
3267 or two objects that don't implement comparison methods are reported
3268 as warnings. ``dict1 == dict2`` still works, but ``dict1 < dict2``
3269 is being phased out.
3271 Comparisons between cells, which are an implementation detail of Python's
3272 scoping rules, also cause warnings because such comparisons are forbidden
3275 .. ======================================================================
3283 The author would like to thank the following people for offering
3284 suggestions, corrections and assistance with various drafts of this
3285 article: Georg Brandl, Steve Brown, Nick Coghlan, Ralph Corderoy,
3286 Jim Jewett, Kent Johnson, Chris Lambacher, Martin Michlmayr,
3287 Antoine Pitrou, Brian Warner.