功能¶

• All native single character format specifiers in struct module syntax(optionally prefixed with '@') are now supported.

• With some restrictions, the cast() method allows changing of format andshape of C-contiguous arrays.

• Multi-dimensional list representations are supported for any array type.

• Multi-dimensional comparisons are supported for any array type.

• One-dimensional memoryviews of hashable (read-only) types with formats B,b or c are now hashable. (Contributed by Antoine Pitrou inbpo-13411.)

• Arbitrary slicing of any 1-D arrays type is supported. For example, itis now possible to reverse a memoryview inO(1) by using a negative step.

API 變更¶

• The maximum number of dimensions is officially limited to 64.

• The representation of empty shape, strides and suboffsets is nowan empty tuple instead ofNone.

• Accessing a memoryview element with format 'B' (unsigned bytes)now returns an integer (in accordance with the struct module syntax).For returning a bytes object the view must be cast to 'c' first.

• memoryview comparisons now use the logical structure of the operandsand compare all array elements by value. All format strings in structmodule syntax are supported. Views with unrecognised format stringsare still permitted, but will always compare as unequal, regardlessof view contents.

• For further changes seeBuild and C API Changes andPorting C code.

（由 Stefan Krah 在bpo-10181 中貢獻。）

功能性¶

PEP 393 引入的更改如下：

• Python now always supports the full range of Unicode code points, includingnon-BMP ones (i.e. fromU+0000 toU+10FFFF). The distinction betweennarrow and wide builds no longer exists and Python now behaves like a widebuild, even under Windows.

• With the death of narrow builds, the problems specific to narrow builds havealso been fixed, for example:

  • len() now always returns 1 for non-BMP characters,solen('\U0010FFFF')==1;

  • surrogate pairs are not recombined in string literals,so'\uDBFF\uDFFF'!='\U0010FFFF';

  • indexing or slicing non-BMP characters returns the expected value,so'\U0010FFFF'[0] now returns'\U0010FFFF' and not'\uDBFF';

  • all other functions in the standard library now correctly handlenon-BMP code points.

• The value ofsys.maxunicode is now always1114111 (0x10FFFFin hexadecimal). ThePyUnicode_GetMax() function still returnseither0xFFFF or0x10FFFF for backward compatibility, and it shouldnot be used with the new Unicode API (seebpo-13054).

• ./configure 旗標--with-wide-unicode 已被刪除。

性能和資源使用情況¶

The storage of Unicode strings now depends on the highest code point in the string:

• pure ASCII and Latin1 strings (U+0000-U+00FF) use 1 byte per code point;

• BMP strings (U+0000-U+FFFF) use 2 bytes per code point;

• non-BMP strings (U+10000-U+10FFFF) use 4 bytes per code point.

The net effect is that for most applications, memory usage of stringstorage should decrease significantly - especially compared to formerwide unicode builds - as, in many cases, strings will be pure ASCIIeven in international contexts (because many strings store non-humanlanguage data, such as XML fragments, HTTP headers, JSON-encoded data,etc.). We also hope that it will, for the same reasons, increase CPUcache efficiency on non-trivial applications. The memory usage ofPython 3.3 is two to three times smaller than Python 3.2, and a littlebit better than Python 2.7, on a Django benchmark (see the PEP fordetails).

SimpleNamespace¶

The implementation ofsys.implementation also introduces a new type toPython:types.SimpleNamespace. In contrast to a mapping-basednamespace, likedict,SimpleNamespace is attribute-based, likeobject. However, unlikeobject,SimpleNamespace instancesare writable. This means that you can add, remove, and modify the namespacethrough normal attribute access.

新 API¶

One of the large benefits of this work is the exposure of what goes intomaking the import statement work. That means the various importers that wereonce implicit are now fully exposed as part of theimportlib package.

The abstract base classes defined inimportlib.abc have been expandedto properly delineate betweenmeta path findersandpath entry finders by introducingimportlib.abc.MetaPathFinder andimportlib.abc.PathEntryFinder, respectively. The old ABC ofimportlib.abc.Finder is now only provided for backwards-compatibilityand does not enforce any method requirements.

In terms of finders,importlib.machinery.FileFinder exposes themechanism used to search for source and bytecode files of a module. Previouslythis class was an implicit member ofsys.path_hooks.

For loaders, the new abstract base classimportlib.abc.FileLoader helpswrite a loader that uses the file system as the storage mechanism for a module'scode. The loader for source files(importlib.machinery.SourceFileLoader), sourceless bytecode files(importlib.machinery.SourcelessFileLoader), and extension modules(importlib.machinery.ExtensionFileLoader) are now available fordirect use.

ImportError now hasname andpath attributes which are set whenthere is relevant data to provide. The message for failed imports will alsoprovide the full name of the module now instead of just the tail end of themodule's name.

Theimportlib.invalidate_caches() function will now call the method withthe same name on all finders cached insys.path_importer_cache to helpclean up any stored state as necessary.

明顯的變更¶

For potential required changes to code, see thePorting Python codesection.

Beyond the expanse of whatimportlib now exposes, there are othervisible changes to import. The biggest is thatsys.meta_path andsys.path_hooks now store all of the meta path finders and path entryhooks used by import. Previously the finders were implicit and hidden withinthe C code of import instead of being directly exposed. This means that one cannow easily remove or change the order of the various finders to fit one's needs.

Another change is that all modules have a__loader__ attribute, storing theloader used to create the module.PEP 302 has been updated to make thisattribute mandatory for loaders to implement, so in the future once 3rd-partyloaders have been updated people will be able to rely on the existence of theattribute. Until such time, though, import is setting the module post-load.

Loaders are also now expected to set the__package__ attribute fromPEP 366. Once again, import itself is already setting this on all loadersfromimportlib and import itself is setting the attribute post-load.

None is now inserted intosys.path_importer_cache when no findercan be found onsys.path_hooks. Sinceimp.NullImporter is notdirectly exposed onsys.path_hooks it could no longer be relied upon toalways be available to use as a value representing no finder found.

All other changes relate to semantic changes which should be taken intoconsideration when updating code for Python 3.3, and thus should be read aboutin thePorting Python code section of this document.

（由 Brett Cannon 實作）

faulthandler¶

This new debug modulefaulthandler contains functions to dump Python tracebacks explicitly,on a fault (a crash like a segmentation fault), after a timeout, or on a usersignal. Callfaulthandler.enable() to install fault handlers for theSIGSEGV,SIGFPE,SIGABRT,SIGBUS, andSIGILL signals. You can also enable them at startup by setting thePYTHONFAULTHANDLER environment variable or by using-Xfaulthandler command line option.

Example of a segmentation fault on Linux:

$python-q-Xfaulthandler>>> import ctypes>>> ctypes.string_at(0)Fatal Python error: Segmentation faultCurrent thread 0x00007fb899f39700:  File "/home/python/cpython/Lib/ctypes/__init__.py", line 486 in string_at  File "<stdin>", line 1 in <module>Segmentation fault

ipaddress¶

The newipaddress module provides tools for creating and manipulatingobjects representing IPv4 and IPv6 addresses, networks and interfaces (i.e.an IP address associated with a specific IP subnet).

（由 Google 和 Peter Moody 在PEP 3144 中貢獻。）

lzma¶

The newly addedlzma module provides data compression and decompressionusing the LZMA algorithm, including support for the.xz and.lzmafile formats.

（由 Nadeem Vawda 和 Per Øyvind Karlsen 在bpo-6715 中貢獻。）

abc¶

Improved support for abstract base classes containing descriptors composed withabstract methods. The recommended approach to declaring abstract descriptors isnow to provide__isabstractmethod__ as a dynamically updatedproperty. The built-in descriptors have been updated accordingly.

• abc.abstractproperty 已被棄用，請改property 和abc.abstractmethod()。

• abc.abstractclassmethod 已被棄用，請改用classmethod 和abc.abstractmethod()。

• abc.abstractstaticmethod 已被棄用，請改用staticmethod 和abc.abstractmethod()。

（由 Darren Dale 在bpo-11610 中貢獻。）

abc.ABCMeta.register() now returns the registered subclass, which meansit can now be used as a class decorator (bpo-10868).

array¶

Thearray module supports thelonglong type usingq andQ type codes.

（由 Oren Tirosh 和 Hirokazu Yamamoto 在bpo-1172711 中貢獻。）

base64¶

ASCII-only Unicode strings are now accepted by the decoding functions of thebase64 modern interface. For example,base64.b64decode('YWJj')returnsb'abc'. (Contributed by Catalin Iacob inbpo-13641.)

binascii¶

In addition to the binary objects they normally accept, thea2b_ functionsnow all also accept ASCII-only strings as input. (Contributed by AntoinePitrou inbpo-13637.)

bz2¶

Thebz2 module has been rewritten from scratch. In the process, severalnew features have been added:

• Newbz2.open() function: open a bzip2-compressed file in binary ortext mode.

• bz2.BZ2File can now read from and write to arbitrary file-likeobjects, by means of its constructor'sfileobj argument.

  （由 Nadeem Vawda 在bpo-5863 中貢獻。）

• bz2.BZ2File andbz2.decompress() can now decompressmulti-stream inputs (such as those produced by thepbzip2 tool).bz2.BZ2File can now also be used to create this type of file, usingthe'a' (append) mode.

  （由 Nir Aides 在bpo-1625 中貢獻。）

• bz2.BZ2File now implements all of theio.BufferedIOBase API,except for thedetach() andtruncate() methods.

codecs¶

Thembcs codec has been rewritten to handle correctlyreplace andignore error handlers on all Windows versions. Thembcs codec now supports all error handlers, instead of onlyreplace to encode andignore to decode.

A new Windows-only codec has been added:cp65001 (bpo-13216). It is theWindows code page 65001 (Windows UTF-8,CP_UTF8). For example, it is usedbysys.stdout if the console output code page is set to cp65001 (e.g., usingchcp65001 command).

Multibyte CJK decoders now resynchronize faster. They only ignore the firstbyte of an invalid byte sequence. For example,b'\xff\n'.decode('gb2312','replace') now returns a\n after the replacement character.

(bpo-12016)

Incremental CJK codec encoders are no longer reset at each call to theirencode() methods. For example:

>>>importcodecs>>>encoder=codecs.getincrementalencoder('hz')('strict')>>>b''.join(encoder.encode(x)forxin'\u52ff\u65bd\u65bc\u4eba\u3002 Bye.')b'~{NpJ)l6HK!#~} Bye.'

This example givesb'~{Np~}~{J)~}~{l6~}~{HK~}~{!#~}Bye.' with older Pythonversions.

(bpo-12100)

unicode_internal 編解碼器已被棄用。

collections¶

Addition of a newChainMap class to allow treating anumber of mappings as a single unit. (Written by Raymond Hettinger forbpo-11089, made public inbpo-11297.)

The abstract base classes have been moved in a newcollections.abcmodule, to better differentiate between the abstract and the concretecollections classes. Aliases for ABCs are still present in thecollections module to preserve existing imports. (bpo-11085)

TheCounter class now supports the unary+ and-operators, as well as the in-place operators+=,-=,|=, and&=. (Contributed by Raymond Hettinger inbpo-13121.)

contextlib¶

ExitStack now provides a solid foundation forprogrammatic manipulation of context managers and similar cleanupfunctionality. Unlike the previouscontextlib.nested API (which wasdeprecated and removed), the new API is designed to work correctlyregardless of whether context managers acquire their resources intheir__init__ method (for example, file objects) or in their__enter__ method (for example, synchronisation objects from thethreading module).

(bpo-13585)

crypt¶

Addition of salt and modular crypt format (hashing method) and themksalt()function to thecrypt module.

(bpo-10924)

curses¶

• If thecurses module is linked to the ncursesw library, use Unicodefunctions when Unicode strings or characters are passed (e.g.waddwstr()), and bytes functions otherwise (e.g.waddstr()).

• Use the locale encoding instead ofutf-8 to encode Unicode strings.

• curses.window 有一個新的curses.window.encoding 屬性。

• curses.window 類別有一個新的get_wch() 方法來獲取寬字元

• Thecurses module has a newunget_wch() function topush a wide character so the nextget_wch() will returnit

（由 Iñigo Serna 在bpo-6755 中貢獻。）

datetime¶

• Equality comparisons between naive and awaredatetimeinstances now returnFalse instead of raisingTypeError(bpo-15006).

• Newdatetime.datetime.timestamp() method: Return POSIX timestampcorresponding to thedatetime instance.

• Thedatetime.datetime.strftime() method supports formatting yearsolder than 1000.

• Thedatetime.datetime.astimezone() method can now becalled without arguments to convert datetime instance to the systemtimezone.

decimal¶

bpo-7652 - integrate fast native decimal arithmetic.

由 Stefan Krah 編寫的 C 模組和 libmpdec。

The new C version of the decimal module integrates the high speed libmpdeclibrary for arbitrary precision correctly rounded decimal floating-pointarithmetic. libmpdec conforms to IBM's General Decimal Arithmetic Specification.

Performance gains range from 10x for database applications to 100x fornumerically intensive applications. These numbers are expected gainsfor standard precisions used in decimal floating-point arithmetic. Sincethe precision is user configurable, the exact figures may vary. For example,in integer bignum arithmetic the differences can be significantly higher.

The following table is meant as an illustration. Benchmarks are availableathttps://www.bytereef.org/mpdecimal/quickstart.html.

	decimal.py	_decimal	speedup
pi	42.02s	0.345s	120x
telco	172.19s	5.68s	30x
psycopg	3.57s	0.29s	12x

email¶

mypolicy=compat32.clone(linesep='\r\n')

>>>m=Message(policy=SMTP)>>>m['To']='Éric <foo@example.com>'>>>m['to']'Éric <foo@example.com>'>>>m['to'].addresses(Address(display_name='Éric', username='foo', domain='example.com'),)>>>m['to'].addresses[0].username'foo'>>>m['to'].addresses[0].display_name'Éric'>>>m['Date']=email.utils.localtime()>>>m['Date'].datetimedatetime.datetime(2012, 5, 25, 21, 39, 24, 465484, tzinfo=datetime.timezone(datetime.timedelta(-1, 72000), 'EDT'))>>>m['Date']'Fri, 25 May 2012 21:44:27 -0400'>>>print(m)To: =?utf-8?q?=C3=89ric?= <foo@example.com>Date: Fri, 25 May 2012 21:44:27 -0400

>>>m['cc']=[Group('pals',[Address('Bob','bob','example.com'),...Address('Sally','sally','example.com')]),...Address('Bonzo',addr_spec='bonz@laugh.com')]>>>print(m)To: =?utf-8?q?=C3=89ric?= <foo@example.com>Date: Fri, 25 May 2012 21:44:27 -0400cc: pals: Bob <bob@example.com>, Sally <sally@example.com>;, Bonzo <bonz@laugh.com>

>>>m2=message_from_string(str(m))>>>m2['to']'Éric <foo@example.com>'

>>>m2['cc'].addresses(Address(display_name='Bob', username='bob', domain='example.com'), Address(display_name='Sally', username='sally', domain='example.com'), Address(display_name='Bonzo', username='bonz', domain='laugh.com'))>>>m2['cc'].groups(Group(display_name='pals', addresses=(Address(display_name='Bob', username='bob', domain='example.com'), Address(display_name='Sally', username='sally', domain='example.com')), Group(display_name=None, addresses=(Address(display_name='Bonzo', username='bonz', domain='laugh.com'),))

ftplib¶

• ftplib.FTP now accepts asource_address keyword argument tospecify the(host,port) to use as the source address in the bind callwhen creating the outgoing socket. (Contributed by Giampaolo Rodolàinbpo-8594.)

• TheFTP_TLS class now provides a newccc() function to revert control channel back toplaintext. This can be useful to take advantage of firewalls that know howto handle NAT with non-secure FTP without opening fixed ports. (Contributedby Giampaolo Rodolà inbpo-12139.)

• Addedftplib.FTP.mlsd() method which provides a parsable directorylisting format and deprecatesftplib.FTP.nlst() andftplib.FTP.dir(). (Contributed by Giampaolo Rodolà inbpo-11072.)

functools¶

Thefunctools.lru_cache() decorator now accepts atyped keywordargument (that defaults toFalse to ensure that it caches values ofdifferent types that compare equal in separate cache slots. (Contributedby Raymond Hettinger inbpo-13227.)

gc¶

It is now possible to register callbacks invoked by the garbage collectorbefore and after collection using the newcallbacks list.

hmac¶

A newcompare_digest() function has been added to prevent sidechannel attacks on digests through timing analysis. (Contributed by NickCoghlan and Christian Heimes inbpo-15061.)

http¶

http.server.BaseHTTPRequestHandler now buffers the headers and writesthem all at once whenend_headers() iscalled. A new methodflush_headers()can be used to directly manage when the accumulated headers are sent.(Contributed by Andrew Schaaf inbpo-3709.)

http.server now produces validHTML4.01strict output.(Contributed by Ezio Melotti inbpo-13295.)

http.client.HTTPResponse now has areadinto() method, which means it can be usedas anio.RawIOBase class. (Contributed by John Kuhn inbpo-13464.)

html¶

html.parser.HTMLParser is now able to parse broken markup withoutraising errors, therefore thestrict argument of the constructor and theHTMLParseError exception are now deprecated.The ability to parse broken markup is the result of a number of bug fixes thatare also available on the latest bug fix releases of Python 2.7/3.2.(Contributed by Ezio Melotti inbpo-15114, andbpo-14538,bpo-13993,bpo-13960,bpo-13358,bpo-1745761,bpo-755670,bpo-13357,bpo-12629,bpo-1200313,bpo-670664,bpo-13273,bpo-12888,bpo-7311.)

A newhtml5 dictionary that maps HTML5 named characterreferences to the equivalent Unicode character(s) (e.g.html5['gt;']=='>') has been added to thehtml.entities module. The dictionary isnow also used byHTMLParser. (Contributed by EzioMelotti inbpo-11113 andbpo-15156.)

imaplib¶

TheIMAP4_SSL constructor now accepts an SSLContextparameter to control parameters of the secure channel.

（由 Sijin Joseph 在bpo-8808 中貢獻。）

inspect¶

A newgetclosurevars() function has been added. This functionreports the current binding of all names referenced from the function body andwhere those names were resolved, making it easier to verify correct internalstate when testing code that relies on stateful closures.

（由 Meador Inge 與 Nick Coghlan 在bpo-13062 中貢獻。）

A newgetgeneratorlocals() function has been added. Thisfunction reports the current binding of local variables in the generator'sstack frame, making it easier to verify correct internal state when testinggenerators.

（由 Meador Inge 於bpo-15153 中貢獻。）

io¶

Theopen() function has a new'x' mode that can be used toexclusively create a new file, and raise aFileExistsError if the filealready exists. It is based on the C11 'x' mode to fopen().

（由 David Townshend 於bpo-12760 中貢獻。）

The constructor of theTextIOWrapper class has a newwrite_through optional argument. Ifwrite_through isTrue, calls towrite() are guaranteed not to be buffered: any datawritten on theTextIOWrapper object is immediately handled to itsunderlying binary buffer.

itertools¶

accumulate() now takes an optionalfunc argument forproviding a user-supplied binary function.

logging¶

ThebasicConfig() function now supports an optionalhandlersargument taking an iterable of handlers to be added to the root logger.

A class level attributeappend_nul hasbeen added toSysLogHandler to allow control of theappending of theNUL (\000) byte to syslog records, since for somedaemons it is required while for others it is passed through to the log.

math¶

Themath module has a new function,log2(), which returnsthe base-2 logarithm ofx.

（由 Mark Dickinson 於bpo-11888 中撰寫）

mmap¶

Theread() method is now more compatible with other file-likeobjects: if the argument is omitted or specified asNone, it returns thebytes from the current file position to the end of the mapping. (Contributedby Petri Lehtinen inbpo-12021.)

multiprocessing¶

The newmultiprocessing.connection.wait() function allows pollingmultiple objects (such as connections, sockets and pipes) with a timeout.(Contributed by Richard Oudkerk inbpo-12328.)

multiprocessing.Connection objects can now be transferred overmultiprocessing connections.(Contributed by Richard Oudkerk inbpo-4892.)

multiprocessing.Process now accepts adaemon keyword argumentto override the default behavior of inheriting thedaemon flag fromthe parent process (bpo-6064).

New attributemultiprocessing.Process.sentinel allows aprogram to wait on multipleProcess objects at onetime using the appropriate OS primitives (for example,select onposix systems).

New methodsmultiprocessing.pool.Pool.starmap() andstarmap_async() provideitertools.starmap() equivalents to the existingmultiprocessing.pool.Pool.map() andmap_async() functions. (Contributed by HynekSchlawack inbpo-12708.)

nntplib¶

Thenntplib.NNTP class now supports the context management protocol tounconditionally consumesocket.error exceptions and to close the NNTPconnection when done:

>>>fromnntplibimportNNTP>>>withNNTP('news.gmane.org')asn:...n.group('gmane.comp.python.committers')...('211 1755 1 1755 gmane.comp.python.committers', 1755, 1, 1755, 'gmane.comp.python.committers')>>>

（由 Giampaolo Rodolà 於bpo-9795 中貢獻。）

os¶

• Theos module has a newpipe2() function that makes itpossible to create a pipe withO_CLOEXEC orO_NONBLOCK flags set atomically. This is especially useful toavoid race conditions in multi-threaded programs.

• Theos module has a newsendfile() function which providesan efficient "zero-copy" way for copying data from one file (or socket)descriptor to another. The phrase "zero-copy" refers to the fact that all ofthe copying of data between the two descriptors is done entirely by thekernel, with no copying of data into userspace buffers.sendfile()can be used to efficiently copy data from a file on disk to a network socket,e.g. for downloading a file.

  （由 Ross Lagerwall 和 Giampaolo Rodolà 於bpo-10882 提交補丁。）

• To avoid race conditions like symlink attacks and issues with temporaryfiles and directories, it is more reliable (and also faster) to manipulatefile descriptors instead of file names. Python 3.3 enhances existing functionsand introduces new functions to work on file descriptors (bpo-4761,bpo-10755 andbpo-14626).

  • Theos module has a newfwalk() function similar towalk() except that it also yields file descriptors referring to thedirectories visited. This is especially useful to avoid symlink races.

  • The following functions get new optionaldir_fd (paths relative todirectory descriptors) and/orfollow_symlinks (notfollowing symlinks):access(),chflags(),chmod(),chown(),link(),lstat(),mkdir(),mkfifo(),mknod(),open(),readlink(),remove(),rename(),replace(),rmdir(),stat(),symlink(),unlink(),utime(). Platformsupport for using these parameters can be checked via the setsos.supports_dir_fd andos.supports_follows_symlinks.

  • The following functions now support a file descriptor for their path argument:chdir(),chmod(),chown(),execve(),listdir(),pathconf(),exists(),stat(),statvfs(),utime(). Platform supportfor this can be checked via theos.supports_fd set.

• access() accepts aneffective_ids keyword argument to turn onusing the effective uid/gid rather than the real uid/gid in the access check.Platform support for this can be checked via thesupports_effective_ids set.

• Theos module has two new functions:getpriority() andsetpriority(). They can be used to get or set processniceness/priority in a fashion similar toos.nice() but extended to allprocesses instead of just the current one.

  （由 Giampaolo Rodolà 於bpo-10784 提交補丁。）

• The newos.replace() function allows cross-platform renaming of afile with overwriting the destination. Withos.rename(), an existingdestination file is overwritten under POSIX, but raises an error underWindows.(Contributed by Antoine Pitrou inbpo-8828.)

• The stat family of functions (stat(),fstat(),andlstat()) now support reading a file's timestampswith nanosecond precision. Symmetrically,utime()can now write file timestamps with nanosecond precision. (Contributed byLarry Hastings inbpo-14127.)

• The newos.get_terminal_size() function queries the size of theterminal attached to a file descriptor. See alsoshutil.get_terminal_size().(Contributed by Zbigniew Jędrzejewski-Szmek inbpo-13609.)

• New functions to support Linux extended attributes (bpo-12720):getxattr(),listxattr(),removexattr(),setxattr().

• New interface to the scheduler. These functionscontrol how a process is allocated CPU time by the operating system. Newfunctions:sched_get_priority_max(),sched_get_priority_min(),sched_getaffinity(),sched_getparam(),sched_getscheduler(),sched_rr_get_interval(),sched_setaffinity(),sched_setparam(),sched_setscheduler(),sched_yield(),

• New functions to control the file system:

  • posix_fadvise(): Announces an intention to access data in aspecific pattern thus allowing the kernel to make optimizations.

  • posix_fallocate(): Ensures that enough disk space is allocatedfor a file.

  • sync(): Force write of everything to disk.

• 其他新的 posix 函式：

  • lockf(): Apply, test or remove a POSIX lock on an open file descriptor.

  • pread(): Read from a file descriptor at an offset, the fileoffset remains unchanged.

  • pwrite(): Write to a file descriptor from an offset, leavingthe file offset unchanged.

  • readv(): Read from a file descriptor into a number of writable buffers.

  • truncate(): Truncate the file corresponding topath, so thatit is at mostlength bytes in size.

  • waitid(): Wait for the completion of one or more child processes.

  • writev(): Write the contents ofbuffers to a file descriptor,wherebuffers is an arbitrary sequence of buffers.

  • getgrouplist() (bpo-9344): Return list of group ids thatspecified user belongs to.

• times() anduname(): Return type changed from a tuple toa tuple-like object with named attributes.

• Some platforms now support additional constants for thelseek()function, such asos.SEEK_HOLE andos.SEEK_DATA.

• New constantsRTLD_LAZY,RTLD_NOW,RTLD_GLOBAL,RTLD_LOCAL,RTLD_NODELETE,RTLD_NOLOAD, andRTLD_DEEPBIND are available onplatforms that support them. These are for use with thesys.setdlopenflags() function, and supersede the similar constantsdefined inctypes andDLFCN. (Contributed by Victor Stinnerinbpo-13226.)

• os.symlink() now accepts (and ignores) thetarget_is_directorykeyword argument on non-Windows platforms, to ease cross-platform support.

pdb¶

Tab-completion is now available not only for command names, but also theirarguments. For example, for thebreak command, function and file namesare completed.

（由 Georg Brandl 在bpo-14210 中貢獻）

pickle¶

pickle.Pickler objects now have an optionaldispatch_table attribute allowing per-picklerreduction functions to be set.

（由 Richard Oudkerk 在bpo-14166 中貢獻。）

pydoc¶

The Tk GUI and theserve() function have been removed from thepydoc module:pydoc-g andserve() have been deprecatedin Python 3.2.

re¶

str regular expressions now support\u and\U escapes.

（由 Serhiy Storchaka 在bpo-3665 中貢獻。）

sched¶

• run() now accepts ablocking parameter which whenset to false makes the method execute the scheduled events due to expiresoonest (if any) and then return immediately.This is useful in case you want to use thescheduler innon-blocking applications. (Contributed by Giampaolo Rodolà inbpo-13449.)

• scheduler class can now be safely used in multi-threadedenvironments. (Contributed by Josiah Carlson and Giampaolo Rodolà inbpo-8684.)

• timefunc anddelayfunct parameters ofscheduler classconstructor are now optional and defaults totime.time() andtime.sleep() respectively. (Contributed by Chris Clark inbpo-13245.)

• enter() 和enterabs()argument 參數現在是可選的。（由 Chris Clark 在bpo-13245 中貢獻。）

• enter() 和enterabs() 現在接受kwargs 參數。（由 Chris Clark 在bpo-13245 中貢獻。）

select¶

Solaris and derivative platforms have a new classselect.devpollfor high performance asynchronous sockets via/dev/poll.(Contributed by Jesús Cea Avión inbpo-6397.)

shlex¶

The previously undocumented helper functionquote from thepipes modules has been moved to theshlex module anddocumented.quote() properly escapes all characters in a stringthat might be otherwise given special meaning by the shell.

shutil¶

• 新功能：

  • disk_usage(): provides total, used and free disk spacestatistics. (Contributed by Giampaolo Rodolà inbpo-12442.)

  • chown(): allows one to change user and/or group of the givenpath also specifying the user/group names and not only their numericids. (Contributed by Sandro Tosi inbpo-12191.)

  • shutil.get_terminal_size(): returns the size of the terminal windowto which the interpreter is attached. (Contributed by ZbigniewJędrzejewski-Szmek inbpo-13609.)

• copy2() andcopystat() now preserve filetimestamps with nanosecond precision on platforms that support it.They also preserve file "extended attributes" on Linux. (Contributedby Larry Hastings inbpo-14127 andbpo-15238.)

• Several functions now take an optionalsymlinks argument: when thatparameter is true, symlinks aren't dereferenced and the operation insteadacts on the symlink itself (or creates one, if relevant).(Contributed by Hynek Schlawack inbpo-12715.)

• When copying files to a different file system,move() nowhandles symlinks the way the posixmv command does, recreating thesymlink rather than copying the target file contents. (Contributed byJonathan Niehof inbpo-9993.)move() now also returnsthedst argument as its result.

• rmtree() is now resistant to symlink attacks on platformswhich support the newdir_fd parameter inos.open() andos.unlink(). (Contributed by Martin von Löwis and Hynek Schlawackinbpo-4489.)

signal¶

• signal 模組有了新的函式：

  • pthread_sigmask(): fetch and/or change the signal mask of thecalling thread (Contributed by Jean-Paul Calderone inbpo-8407);

  • pthread_kill()：向執行緒發送訊號；

  • sigpending()：檢查未定的函式；

  • sigwait()：等待訊號；

  • sigwaitinfo()：等待訊號，回傳有關該訊號的詳細資訊；

  • sigtimedwait()：類似於sigwaitinfo() 但有超時。

• The signal handler writes the signal number as a single byte instead ofa nul byte into the wakeup file descriptor. So it is possible to wait morethan one signal and know which signals were raised.

• signal.signal() 和signal.siginterrupt() 會引發 OSError，而不是 RuntimeError：OSError 有一個 errno 屬性。

smtpd¶

Thesmtpd module now supportsRFC 5321 (extended SMTP) andRFC 1870(size extension). Per the standard, these extensions are enabled if and onlyif the client initiates the session with anEHLO command.

(InitialELHO support by Alberto Trevino. Size extension by JuhanaJauhiainen. Substantial additional work on the patch contributed by MicheleOrrù and Dan Boswell.bpo-8739)

smtplib¶

TheSMTP,SMTP_SSL, andLMTP classes now accept asource_address keyword argumentto specify the(host,port) to use as the source address in the bind callwhen creating the outgoing socket. (Contributed by Paulo Scardine inbpo-11281.)

SMTP 現在支援情境管理協議，允許在with 陳述式中使用SMTP 實例。（由 Giampaolo Rodolà 在bpo-11289 中貢獻。）

SMTP_SSL 構造函式和starttls() 方法現在接受 SSLContext 參數來控制安全通道的參數。（由 Kasun Herath 在bpo-8809 中貢獻。）

socket¶

• 當底層平台支援時，socket 類現在公開了額外的方法來處理輔助資料：

  • sendmsg()

  • recvmsg()

  • recvmsg_into()

  （由 David Watson 在bpo-6560 中貢獻，其基於 Heiko Wundram 的早期補丁）

• socket 類別現在支援 Linux (https://lwn.net/Articles/253425) 上的 PF_CAN 協議系列 (https://en.wikipedia.org/wiki/Socketcan)。

  （在bpo-10141 中由 Matthias Fuchs 貢獻、並由 Tiago Gonçalves 更新。）

• socket 類別現在支援 PF_RDS 協議系列（https://en.wikipedia.org/wiki/Reliable_Datagram_Sockets 和https://oss.oracle.com/projects/rds）。

• socket 類別現在支援 OS X 上的PF_SYSTEM 協議系列。（由 Michael Goderbauer 在bpo-13777 中貢獻。）

• New functionsethostname() allows the hostname to be seton Unix systems if the calling process has sufficient privileges.(Contributed by Ross Lagerwall inbpo-10866.)

socketserver¶

BaseServer now has an overridable methodservice_actions() that is called by theserve_forever() method in the service loop.ForkingMixIn now uses this to clean up zombiechild processes. (Contributed by Justin Warkentin inbpo-11109.)

sqlite3¶

Newsqlite3.Connection methodset_trace_callback() can be used to capture a trace ofall sql commands processed by sqlite. (Contributed by Torsten Landschoffinbpo-11688.)

ssl¶

• ssl 模組有兩個新的隨機生成函式：

  • RAND_bytes()：生成加密的強偽隨機位元組。

  • RAND_pseudo_bytes()：生成偽隨機位元組。

  （由 Victor Stinner 在bpo-12049 中貢獻。）

• Thessl module now exposes a finer-grained exception hierarchyin order to make it easier to inspect the various kinds of errors.(Contributed by Antoine Pitrou inbpo-11183.)

• load_cert_chain() now accepts apassword argumentto be used if the private key is encrypted.(Contributed by Adam Simpkins inbpo-12803.)

• Diffie-Hellman key exchange, both regular and Elliptic Curve-based, isnow supported through theload_dh_params() andset_ecdh_curve() methods.(Contributed by Antoine Pitrou inbpo-13626 andbpo-13627.)

• SSL sockets have a newget_channel_binding() methodallowing the implementation of certain authentication mechanisms such asSCRAM-SHA-1-PLUS. (Contributed by Jacek Konieczny inbpo-12551.)

• You can query the SSL compression algorithm used by an SSL socket, thanksto its newcompression() method. The new attributeOP_NO_COMPRESSION can be used to disable compression.(Contributed by Antoine Pitrou inbpo-13634.)

• Support has been added for the Next Protocol Negotiation extension usingthessl.SSLContext.set_npn_protocols() method.(Contributed by Colin Marc inbpo-14204.)

• SSL errors can now be introspected more easily thanks tolibrary andreason attributes.(Contributed by Antoine Pitrou inbpo-14837.)

• Theget_server_certificate() function now supports IPv6.(Contributed by Charles-François Natali inbpo-11811.)

• New attributeOP_CIPHER_SERVER_PREFERENCE allows settingSSLv3 server sockets to use the server's cipher ordering preference ratherthan the client's (bpo-13635).

stat¶

The undocumented tarfile.filemode function has been moved tostat.filemode(). It can be used to convert a file's mode to a string ofthe form '-rwxrwxrwx'.

（由 Giampaolo Rodolà 在bpo-14807 中貢獻。）

struct¶

Thestruct module now supportsssize_t andsize_t via thenew codesn andN, respectively. (Contributed by Antoine Pitrouinbpo-3163.)

subprocess¶

Command strings can now be bytes objects on posix platforms. (Contributed byVictor Stinner inbpo-8513.)

A new constantDEVNULL allows suppressing output in aplatform-independent fashion. (Contributed by Ross Lagerwall inbpo-5870.)

sys¶

Thesys module has a newthread_infonamedtuple holding information about the thread implementation(bpo-11223).

tarfile¶

tarfile 現在透過lzma 模組支援lzma 編碼。（由 Lars Gustäbel 在bpo-5689 中貢獻。）

tempfile¶

tempfile.SpooledTemporaryFile'struncate() method now acceptsasize parameter. (Contributed by Ryan Kelly inbpo-9957.)

textwrap¶

Thetextwrap module has a newindent() that makesit straightforward to add a common prefix to selected lines in a blockof text (bpo-13857).

threading¶

threading.Condition,threading.Semaphore,threading.BoundedSemaphore,threading.Event, andthreading.Timer, all of which used to be factory functions returning aclass instance, are now classes and may be subclassed. (Contributed by ÉricAraujo inbpo-10968.)

Thethreading.Thread constructor now accepts adaemon keywordargument to override the default behavior of inheriting thedaemon flagvalue from the parent thread (bpo-6064).

The formerly private function_thread.get_ident is now available as thepublic functionthreading.get_ident(). This eliminates several cases ofdirect access to the_thread module in the stdlib. Third party code thatused_thread.get_ident should likewise be changed to use the new publicinterface.

time¶

PEP 418 向time 模組新增了新功能：

• get_clock_info()：取得時鐘資訊。

• monotonic()：單調時鐘（不能倒退），不受系統時鐘更新的影響。

• perf_counter(): Performance counter with the highest availableresolution to measure a short duration.

• process_time()：目前行程的系統和使用者 CPU 時間總和。

其他新功能：

• clock_getres()、clock_gettime() 和clock_settime() 函式帶有CLOCK_xxx 常數。（由 Victor Stinner 在bpo-10278 中貢獻。）

To improve cross platform consistency,sleep() now raises aValueError when passed a negative sleep value. Previously this was anerror on posix, but produced an infinite sleep on Windows.

types¶

Add a newtypes.MappingProxyType class: Read-only proxy of a mapping.(bpo-14386)

The new functionstypes.new_class() andtypes.prepare_class() provide supportforPEP 3115 compliant dynamic type creation. (bpo-14588)

unittest¶

assertRaises(),assertRaisesRegex(),assertWarns(), andassertWarnsRegex() now accept a keyword argumentmsg when used ascontext managers. (Contributed by Ezio Melotti and Winston Ewert inbpo-10775.)

unittest.TestCase.run() 現在回傳TestResult 物件。

urllib¶

TheRequest class, now accepts amethod argumentused byget_method() to determine what HTTP methodshould be used. For example, this will send a'HEAD' request:

>>>urlopen(Request('https://www.python.org',method='HEAD'))

(bpo-1673007)

webbrowser¶

Thewebbrowser module supports more "browsers": Google Chrome (namedchrome,chromium,chrome-browser orchromium-browser depending on the version and operating system),and the generic launchersxdg-open, from the FreeDesktop.orgproject, andgvfs-open, which is the default URI handler for GNOME3. (The former contributed by Arnaud Calmettes inbpo-13620, the latterby Matthias Klose inbpo-14493.)

xml.etree.ElementTree¶

Thexml.etree.ElementTree module now imports its C accelerator bydefault; there is no longer a need to explicitly importxml.etree.cElementTree (this module stays for backwards compatibility,but is now deprecated). In addition, theiter family of methods ofElement has been optimized (rewritten in C).The module's documentation has also been greatly improved with added examplesand a more detailed reference.

zlib¶

New attributezlib.Decompress.eof makes it possible to distinguishbetween a properly formed compressed stream and an incomplete or truncated one.(Contributed by Nadeem Vawda inbpo-12646.)

New attributezlib.ZLIB_RUNTIME_VERSION reports the version string ofthe underlyingzlib library that is loaded at runtime. (Contributed byTorsten Landschoff inbpo-12306.)

不支援的作業系統¶

由於缺乏維護者，OS/2 和 VMS 不再受支援。

Windows 2000 and Windows platforms which setCOMSPEC tocommand.comare no longer supported due to maintenance burden.

OSF 支援在 3.2 中已棄用，現已完全刪除。

已棄用的 Python 模組、函式和方法¶

• Passing a non-empty string toobject.__format__() is deprecated, andwill produce aTypeError in Python 3.4 (bpo-9856).

• Theunicode_internal codec has been deprecated because of thePEP 393, use UTF-8, UTF-16 (utf-16-le orutf-16-be), or UTF-32(utf-32-le orutf-32-be)

• ftplib.FTP.nlst() 和ftplib.FTP.dir()：使用ftplib.FTP.mlsd()

• platform.popen()：使用subprocess 模組。請特別檢查Replacing Older Functions with the subprocess Module 部分 (bpo-11377)。

• bpo-13374: The Windows bytes API has been deprecated in theosmodule. Use Unicode filenames, instead of bytes filenames, to not depend onthe ANSI code page anymore and to support any filename.

• bpo-13988: Thexml.etree.cElementTree module is deprecated. Theaccelerator is used automatically whenever available.

• The behaviour oftime.clock() depends on the platform: use the newtime.perf_counter() ortime.process_time() function instead,depending on your requirements, to have a well defined behaviour.

• os.stat_float_times() 函式已棄用。

• abc 模組：

  • abc.abstractproperty 已被棄用，請改property 和abc.abstractmethod()。

  • abc.abstractclassmethod 已被棄用，請改用classmethod 和abc.abstractmethod()。

  • abc.abstractstaticmethod 已被棄用，請改用staticmethod 和abc.abstractmethod()。

• importlib 套件：

  • importlib.abc.SourceLoader.path_mtime() is now deprecated in favour ofimportlib.abc.SourceLoader.path_stats() as bytecode files now storeboth the modification time and size of the source file the bytecode file wascompiled from.

C API 中已棄用的函式和型別¶

Py_UNICODE 已被PEP 393 棄用，並將在 Python 4 中刪除。所有使用此型別的函式均已棄用：

Unicode functions and methods usingPy_UNICODE andPy_UNICODE* types:

• PyUnicode_FromUnicode：使用PyUnicode_FromWideChar() 或PyUnicode_FromKindAndData()

• PyUnicode_AS_UNICODE、PyUnicode_AsUnicode()、PyUnicode_AsUnicodeAndSize()：使用PyUnicode_AsWideCharString()

• PyUnicode_AS_DATA：將PyUnicode_DATA 與PyUnicode_READ 和PyUnicode_WRITE 一起使用

• PyUnicode_GET_SIZE、PyUnicode_GetSize()：使用PyUnicode_GET_LENGTH 或PyUnicode_GetLength()

• PyUnicode_GET_DATA_SIZE：使用PyUnicode_GET_LENGTH(str)*PyUnicode_KIND(str)（僅適用於準備好的字串）

• PyUnicode_AsUnicodeCopy()：使用PyUnicode_AsUCS4Copy() 或PyUnicode_AsWideCharString()

• PyUnicode_GetMax()

操作 Py_UNICODE* 字串的函式和巨集：

• Py_UNICODE_strlen()：使用PyUnicode_GetLength() 或PyUnicode_GET_LENGTH

• Py_UNICODE_strcat()：使用PyUnicode_CopyCharacters() 或PyUnicode_FromFormat()

• Py_UNICODE_strcpy()、Py_UNICODE_strncpy()、Py_UNICODE_COPY()：使用PyUnicode_CopyCharacters() 或PyUnicode_Substring()

• Py_UNICODE_strcmp()：使用PyUnicode_Compare()

• Py_UNICODE_strncmp(): 使用PyUnicode_Tailmatch()

• Py_UNICODE_strchr(),Py_UNICODE_strrchr(): 使用PyUnicode_FindChar()

• Py_UNICODE_FILL(): 使用PyUnicode_Fill()

• Py_UNICODE_MATCH

編碼器：

• PyUnicode_Encode()：使用PyUnicode_AsEncodedObject()

• PyUnicode_EncodeUTF7()

• PyUnicode_EncodeUTF8()：使用PyUnicode_AsUTF8() 或PyUnicode_AsUTF8String()

• PyUnicode_EncodeUTF32()

• PyUnicode_EncodeUTF16()

• PyUnicode_EncodeUnicodeEscape() 使用PyUnicode_AsUnicodeEscapeString()

• PyUnicode_EncodeRawUnicodeEscape() 使用PyUnicode_AsRawUnicodeEscapeString()

• PyUnicode_EncodeLatin1(): 使用PyUnicode_AsLatin1String()

• PyUnicode_EncodeASCII()：使用PyUnicode_AsASCIIString()

• PyUnicode_EncodeCharmap()

• PyUnicode_TranslateCharmap()

• PyUnicode_EncodeMBCS()：使用PyUnicode_AsMBCSString() 或PyUnicode_EncodeCodePage() （帶有CP_ACP code_page）

• PyUnicode_EncodeDecimal()、PyUnicode_TransformDecimalToASCII()

已棄用的功能¶

Thearray module's'u' format code is now deprecated and will beremoved in Python 4 together with the rest of the (Py_UNICODE) API.

移植 Python 程式碼¶

• Hash randomization is enabled by default. Set thePYTHONHASHSEEDenvironment variable to0 to disable hash randomization. See also theobject.__hash__() method.

• bpo-12326: On Linux, sys.platform doesn't contain the major versionanymore. It is now always 'linux', instead of 'linux2' or 'linux3' dependingon the Linux version used to build Python. Replace sys.platform == 'linux2'with sys.platform.startswith('linux'), or directly sys.platform == 'linux' ifyou don't need to support older Python versions.

• bpo-13847,bpo-14180:time anddatetime:OverflowError is now raised instead ofValueError if atimestamp is out of range.OSError is now raised if C functionsgmtime() orlocaltime() failed.

• The default finders used by import now utilize a cache of what is containedwithin a specific directory. If you create a Python source file or sourcelessbytecode file, make sure to callimportlib.invalidate_caches() to clearout the cache for the finders to notice the new file.

• ImportError now uses the full name of the module that was attempted tobe imported. Doctests that check ImportErrors' message will need to beupdated to use the full name of the module instead of just the tail of thename.

• Theindex argument to__import__() now defaults to 0 instead of -1and no longer support negative values. It was an oversight whenPEP 328 wasimplemented that the default value remained -1. If you need to continue toperform a relative import followed by an absolute import, then perform therelative import using an index of 1, followed by another import using anindex of 0. It is preferred, though, that you useimportlib.import_module() rather than call__import__() directly.

• __import__() no longer allows one to use an index value other than 0for top-level modules. E.g.__import__('sys',level=1) is now an error.

• Becausesys.meta_path andsys.path_hooks now have finders onthem by default, you will most likely want to uselist.insert() insteadoflist.append() to add to those lists.

• BecauseNone is now inserted intosys.path_importer_cache, if youare clearing out entries in the dictionary of paths that do not have afinder, you will need to remove keys paired with values ofNoneandimp.NullImporter to be backwards-compatible. This will lead to extraoverhead on older versions of Python that re-insertNone intosys.path_importer_cache where it represents the use of implicitfinders, but semantically it should not change anything.

• importlib.abc.Finder no longer specifies afind_module() abstractmethod that must be implemented. If you were relying on subclasses toimplement that method, make sure to check for the method's existence first.You will probably want to check forfind_loader() first, though, in thecase of working withpath entry finders.

• pkgutil has been converted to useimportlib internally. Thiseliminates many edge cases where the old behaviour of thePEP 302 importemulation failed to match the behaviour of the real import system. Theimport emulation itself is still present, but is now deprecated. Thepkgutil.iter_importers() andpkgutil.walk_packages() functionsspecial case the standard import hooks so they are still supported eventhough they do not provide the non-standarditer_modules() method.

• A longstanding RFC-compliance bug (bpo-1079) in the parsing done byemail.header.decode_header() has been fixed. Code that uses thestandard idiom to convert encoded headers into unicode(str(make_header(decode_header(h))) will see no change, but code thatlooks at the individual tuples returned by decode_header will see thatwhitespace that precedes or followsASCII sections is now included in theASCII section. Code that builds headers usingmake_header shouldalso continue to work without change, sincemake_header continues to addwhitespace betweenASCII and non-ASCII sections if it is not alreadypresent in the input strings.

• email.utils.formataddr() now does the correct content transferencoding when passed non-ASCII display names. Any code that depended onthe previous buggy behavior that preserved the non-ASCII unicode in theformatted output string will need to be changed (bpo-1690608).

• poplib.POP3.quit() may now raise protocol errors like all otherpoplib methods. Code that assumesquit does not raisepoplib.error_proto errors may need to be changed if errors onquitare encountered by a particular application (bpo-11291).

• 自 Python 2.4 以來已棄用的email.parser.Parser 的strict 引數終於被刪除了。

• 已棄用的方法unittest.TestCase.assertSameElements 已被刪除。

• 已棄用的變數time.accept2dyear 已被刪除。

• The deprecatedContext._clamp attribute has been removed from thedecimal module. It was previously replaced by the public attributeclamp. (Seebpo-8540.)

• The undocumented internal helper classSSLFakeFile has been removedfromsmtplib, since its functionality has long been provided directlybysocket.socket.makefile().

• Passing a negative value totime.sleep() on Windows now raises anerror instead of sleeping forever. It has always raised an error on posix.

• Theast.__version__ constant has been removed. If you need tomake decisions affected by the AST version, usesys.version_infoto make the decision.

• Code that used to work around the fact that thethreading module usedfactory functions by subclassing the private classes will need to change tosubclass the now-public classes.

• The undocumented debugging machinery in the threading module has beenremoved, simplifying the code. This should have no effect on productioncode, but is mentioned here in case any application debug frameworks wereinteracting with it (bpo-13550).

Porting C code¶

• In the course of changes to the buffer API the undocumentedsmalltable member of thePy_buffer structure has been removed and thelayout of thePyMemoryViewObject has changed.

  All extensions relying on the relevant parts inmemoryobject.horobject.h must be rebuilt.

• Due toPEP 393, thePy_UNICODE type and allfunctions using this type are deprecated (but will stay available forat least five years). If you were using low-level Unicode APIs toconstruct and access unicode objects and you want to benefit of thememory footprint reduction provided byPEP 393, you have to convertyour code to the newUnicode API.

  However, if you only have been using high-level functions such asPyUnicode_Concat(),PyUnicode_Join() orPyUnicode_FromFormat(), your code will automatically takeadvantage of the new unicode representations.

• PyImport_GetMagicNumber() 現在在失敗時回傳-1。

• As a negative value for thelevel argument to__import__() is nolonger valid, the same now holds forPyImport_ImportModuleLevel().This also means that the value oflevel used byPyImport_ImportModuleEx() is now0 instead of-1.

Building C extensions¶

• The range of possible file names for C extensions has been narrowed.Very rarely used spellings have been suppressed: under POSIX, filesnamedxxxmodule.so,xxxmodule.abi3.so andxxxmodule.cpython-*.so are no longer recognized as implementingthexxx module. If you had been generating such files, you haveto switch to the other spellings (i.e., remove themodule stringfrom the file names).

  （於bpo-14040 中實作。）

命令列開關更改¶

• The -Q command-line flag and related artifacts have been removed. Codechecking sys.flags.division_warning will need updating.

  （由 Éric Araujo 於bpo-10998 中實作。）

• Whenpython is started with-S,importsitewill no longer add site-specific paths to the module search paths. Inprevious versions, it did.

  （bpo-11591，由 Carl Meyer 貢獻並由 Éric Araujo 修訂。）