blob: f7b8c45fcef87574c160665ef2019dd060f54c98 [file] [log] [blame]
Fred Drake2db76802004-12-01 05:05:47 +00001\documentclass{howto}
2\usepackage{distutils}
3% $Id$
4
Andrew M. Kuchling29b3d082006-04-14 20:35:17 +00005% Fix XXX comments
Andrew M. Kuchling5f445bf2006-04-12 18:54:00 +00006% Count up the patches and bugs
Fred Drake2db76802004-12-01 05:05:47 +00007
8\title{What's New in Python 2.5}
Andrew M. Kuchling99714cf2006-04-27 12:23:07 +00009\release{0.2}
Andrew M. Kuchling92e24952004-12-03 13:54:09 +000010\author{A.M. Kuchling}
11\authoraddress{\email{amk@amk.ca}}
Fred Drake2db76802004-12-01 05:05:47 +000012
13\begin{document}
14\maketitle
15\tableofcontents
16
17This article explains the new features in Python 2.5. No release date
Andrew M. Kuchling5eefdca2006-02-08 11:36:09 +000018for Python 2.5 has been set; it will probably be released in the
Andrew M. Kuchlingd96a6ac2006-04-04 19:17:34 +000019autumn of 2006. \pep{356} describes the planned release schedule.
Fred Drake2db76802004-12-01 05:05:47 +000020
Andrew M. Kuchling0d660c02006-04-17 14:01:36 +000021Comments, suggestions, and error reports are welcome; please e-mail them
22to the author or open a bug in the Python bug tracker.
Andrew M. Kuchling9c67ee02006-04-04 19:07:27 +000023
Andrew M. Kuchling437567c2006-03-07 20:48:55 +000024% XXX Compare with previous release in 2 - 3 sentences here.
Fred Drake2db76802004-12-01 05:05:47 +000025
26This article doesn't attempt to provide a complete specification of
27the new features, but instead provides a convenient overview. For
28full details, you should refer to the documentation for Python 2.5.
Andrew M. Kuchling437567c2006-03-07 20:48:55 +000029% XXX add hyperlink when the documentation becomes available online.
Fred Drake2db76802004-12-01 05:05:47 +000030If you want to understand the complete implementation and design
31rationale, refer to the PEP for a particular new feature.
32
33
34%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +000035\section{PEP 243: Uploading Modules to PyPI\label{pep-243}}
Andrew M. Kuchling6a67e4e2006-04-12 13:03:35 +000036
37PEP 243 describes an HTTP-based protocol for submitting software
38packages to a central archive. The Python package index at
39\url{http://cheeseshop.python.org} now supports package uploads, and
40the new \command{upload} Distutils command will upload a package to the
41repository.
42
43Before a package can be uploaded, you must be able to build a
44distribution using the \command{sdist} Distutils command. Once that
45works, you can run \code{python setup.py upload} to add your package
46to the PyPI archive. Optionally you can GPG-sign the package by
George Yoshida297bf822006-04-17 15:44:59 +000047supplying the \longprogramopt{sign} and
48\longprogramopt{identity} options.
Andrew M. Kuchling6a67e4e2006-04-12 13:03:35 +000049
50\begin{seealso}
51
52\seepep{243}{Module Repository Upload Mechanism}{PEP written by
Andrew M. Kuchling5f445bf2006-04-12 18:54:00 +000053Sean Reifschneider; implemented by Martin von~L\"owis
Andrew M. Kuchling6a67e4e2006-04-12 13:03:35 +000054and Richard Jones. Note that the PEP doesn't exactly
55describe what's implemented in PyPI.}
56
57\end{seealso}
58
59
60%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +000061\section{PEP 308: Conditional Expressions\label{pep-308}}
Andrew M. Kuchling437567c2006-03-07 20:48:55 +000062
Andrew M. Kuchlinge362d932006-03-09 13:56:25 +000063For a long time, people have been requesting a way to write
64conditional expressions, expressions that return value A or value B
65depending on whether a Boolean value is true or false. A conditional
66expression lets you write a single assignment statement that has the
67same effect as the following:
68
69\begin{verbatim}
70if condition:
71 x = true_value
72else:
73 x = false_value
74\end{verbatim}
75
76There have been endless tedious discussions of syntax on both
Andrew M. Kuchling9c67ee02006-04-04 19:07:27 +000077python-dev and comp.lang.python. A vote was even held that found the
78majority of voters wanted conditional expressions in some form,
79but there was no syntax that was preferred by a clear majority.
80Candidates included C's \code{cond ? true_v : false_v},
Andrew M. Kuchlinge362d932006-03-09 13:56:25 +000081\code{if cond then true_v else false_v}, and 16 other variations.
82
83GvR eventually chose a surprising syntax:
84
85\begin{verbatim}
86x = true_value if condition else false_value
87\end{verbatim}
88
Andrew M. Kuchling38f85072006-04-02 01:46:32 +000089Evaluation is still lazy as in existing Boolean expressions, so the
90order of evaluation jumps around a bit. The \var{condition}
91expression in the middle is evaluated first, and the \var{true_value}
92expression is evaluated only if the condition was true. Similarly,
93the \var{false_value} expression is only evaluated when the condition
94is false.
Andrew M. Kuchlinge362d932006-03-09 13:56:25 +000095
96This syntax may seem strange and backwards; why does the condition go
97in the \emph{middle} of the expression, and not in the front as in C's
98\code{c ? x : y}? The decision was checked by applying the new syntax
99to the modules in the standard library and seeing how the resulting
100code read. In many cases where a conditional expression is used, one
101value seems to be the 'common case' and one value is an 'exceptional
102case', used only on rarer occasions when the condition isn't met. The
103conditional syntax makes this pattern a bit more obvious:
104
105\begin{verbatim}
106contents = ((doc + '\n') if doc else '')
107\end{verbatim}
108
109I read the above statement as meaning ``here \var{contents} is
Andrew M. Kuchlingd0fcc022006-03-09 13:57:28 +0000110usually assigned a value of \code{doc+'\e n'}; sometimes
Andrew M. Kuchlinge362d932006-03-09 13:56:25 +0000111\var{doc} is empty, in which special case an empty string is returned.''
112I doubt I will use conditional expressions very often where there
113isn't a clear common and uncommon case.
114
115There was some discussion of whether the language should require
116surrounding conditional expressions with parentheses. The decision
117was made to \emph{not} require parentheses in the Python language's
118grammar, but as a matter of style I think you should always use them.
119Consider these two statements:
120
121\begin{verbatim}
122# First version -- no parens
123level = 1 if logging else 0
124
125# Second version -- with parens
126level = (1 if logging else 0)
127\end{verbatim}
128
129In the first version, I think a reader's eye might group the statement
130into 'level = 1', 'if logging', 'else 0', and think that the condition
131decides whether the assignment to \var{level} is performed. The
132second version reads better, in my opinion, because it makes it clear
133that the assignment is always performed and the choice is being made
134between two values.
135
136Another reason for including the brackets: a few odd combinations of
137list comprehensions and lambdas could look like incorrect conditional
138expressions. See \pep{308} for some examples. If you put parentheses
139around your conditional expressions, you won't run into this case.
140
141
142\begin{seealso}
143
144\seepep{308}{Conditional Expressions}{PEP written by
Andrew M. Kuchling67191312006-04-19 12:55:39 +0000145Guido van~Rossum and Raymond D. Hettinger; implemented by Thomas
Andrew M. Kuchlinge362d932006-03-09 13:56:25 +0000146Wouters.}
147
148\end{seealso}
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000149
150
151%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000152\section{PEP 309: Partial Function Application\label{pep-309}}
Fred Drake2db76802004-12-01 05:05:47 +0000153
Andrew M. Kuchlingb1c96fd2005-03-20 21:42:04 +0000154The \module{functional} module is intended to contain tools for
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000155functional-style programming. Currently it only contains a
156\class{partial()} function, but new functions will probably be added
157in future versions of Python.
Andrew M. Kuchlingb1c96fd2005-03-20 21:42:04 +0000158
Andrew M. Kuchling4b000cd2005-04-09 15:51:44 +0000159For programs written in a functional style, it can be useful to
160construct variants of existing functions that have some of the
161parameters filled in. Consider a Python function \code{f(a, b, c)};
162you could create a new function \code{g(b, c)} that was equivalent to
163\code{f(1, b, c)}. This is called ``partial function application'',
164and is provided by the \class{partial} class in the new
165\module{functional} module.
166
167The constructor for \class{partial} takes the arguments
168\code{(\var{function}, \var{arg1}, \var{arg2}, ...
169\var{kwarg1}=\var{value1}, \var{kwarg2}=\var{value2})}. The resulting
170object is callable, so you can just call it to invoke \var{function}
171with the filled-in arguments.
172
173Here's a small but realistic example:
174
175\begin{verbatim}
176import functional
177
178def log (message, subsystem):
179 "Write the contents of 'message' to the specified subsystem."
180 print '%s: %s' % (subsystem, message)
181 ...
182
183server_log = functional.partial(log, subsystem='server')
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000184server_log('Unable to open socket')
Andrew M. Kuchling4b000cd2005-04-09 15:51:44 +0000185\end{verbatim}
186
Andrew M. Kuchling6af7fe02005-08-02 17:20:36 +0000187Here's another example, from a program that uses PyGTk. Here a
188context-sensitive pop-up menu is being constructed dynamically. The
189callback provided for the menu option is a partially applied version
190of the \method{open_item()} method, where the first argument has been
191provided.
Andrew M. Kuchling4b000cd2005-04-09 15:51:44 +0000192
Andrew M. Kuchling6af7fe02005-08-02 17:20:36 +0000193\begin{verbatim}
194...
195class Application:
196 def open_item(self, path):
197 ...
198 def init (self):
199 open_func = functional.partial(self.open_item, item_path)
200 popup_menu.append( ("Open", open_func, 1) )
201\end{verbatim}
Andrew M. Kuchlingb1c96fd2005-03-20 21:42:04 +0000202
203
204\begin{seealso}
205
206\seepep{309}{Partial Function Application}{PEP proposed and written by
207Peter Harris; implemented by Hye-Shik Chang, with adaptations by
208Raymond Hettinger.}
209
210\end{seealso}
Fred Drake2db76802004-12-01 05:05:47 +0000211
212
213%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000214\section{PEP 314: Metadata for Python Software Packages v1.1\label{pep-314}}
Fred Drakedb7b0022005-03-20 22:19:47 +0000215
Andrew M. Kuchlingd8d732e2005-04-09 23:59:41 +0000216Some simple dependency support was added to Distutils. The
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000217\function{setup()} function now has \code{requires}, \code{provides},
218and \code{obsoletes} keyword parameters. When you build a source
219distribution using the \code{sdist} command, the dependency
220information will be recorded in the \file{PKG-INFO} file.
Andrew M. Kuchlingd8d732e2005-04-09 23:59:41 +0000221
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000222Another new keyword parameter is \code{download_url}, which should be
223set to a URL for the package's source code. This means it's now
224possible to look up an entry in the package index, determine the
225dependencies for a package, and download the required packages.
Andrew M. Kuchlingd8d732e2005-04-09 23:59:41 +0000226
Andrew M. Kuchling61434b62006-04-13 11:51:07 +0000227\begin{verbatim}
228VERSION = '1.0'
229setup(name='PyPackage',
230 version=VERSION,
231 requires=['numarray', 'zlib (>=1.1.4)'],
232 obsoletes=['OldPackage']
233 download_url=('http://www.example.com/pypackage/dist/pkg-%s.tar.gz'
234 % VERSION),
235 )
236\end{verbatim}
Andrew M. Kuchlingd8d732e2005-04-09 23:59:41 +0000237
238\begin{seealso}
239
240\seepep{314}{Metadata for Python Software Packages v1.1}{PEP proposed
241and written by A.M. Kuchling, Richard Jones, and Fred Drake;
242implemented by Richard Jones and Fred Drake.}
243
244\end{seealso}
Fred Drakedb7b0022005-03-20 22:19:47 +0000245
246
247%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000248\section{PEP 328: Absolute and Relative Imports\label{pep-328}}
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000249
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000250The simpler part of PEP 328 was implemented in Python 2.4: parentheses
251could now be used to enclose the names imported from a module using
252the \code{from ... import ...} statement, making it easier to import
253many different names.
254
255The more complicated part has been implemented in Python 2.5:
256importing a module can be specified to use absolute or
257package-relative imports. The plan is to move toward making absolute
258imports the default in future versions of Python.
259
260Let's say you have a package directory like this:
261\begin{verbatim}
262pkg/
263pkg/__init__.py
264pkg/main.py
265pkg/string.py
266\end{verbatim}
267
268This defines a package named \module{pkg} containing the
269\module{pkg.main} and \module{pkg.string} submodules.
270
271Consider the code in the \file{main.py} module. What happens if it
272executes the statement \code{import string}? In Python 2.4 and
273earlier, it will first look in the package's directory to perform a
274relative import, finds \file{pkg/string.py}, imports the contents of
275that file as the \module{pkg.string} module, and that module is bound
276to the name \samp{string} in the \module{pkg.main} module's namespace.
277
278That's fine if \module{pkg.string} was what you wanted. But what if
279you wanted Python's standard \module{string} module? There's no clean
280way to ignore \module{pkg.string} and look for the standard module;
281generally you had to look at the contents of \code{sys.modules}, which
282is slightly unclean.
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +0000283Holger Krekel's \module{py.std} package provides a tidier way to perform
284imports from the standard library, \code{import py ; py.std.string.join()},
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000285but that package isn't available on all Python installations.
286
287Reading code which relies on relative imports is also less clear,
288because a reader may be confused about which module, \module{string}
289or \module{pkg.string}, is intended to be used. Python users soon
290learned not to duplicate the names of standard library modules in the
291names of their packages' submodules, but you can't protect against
292having your submodule's name being used for a new module added in a
293future version of Python.
294
295In Python 2.5, you can switch \keyword{import}'s behaviour to
296absolute imports using a \code{from __future__ import absolute_import}
297directive. This absolute-import behaviour will become the default in
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +0000298a future version (probably Python 2.7). Once absolute imports
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000299are the default, \code{import string} will
300always find the standard library's version.
301It's suggested that users should begin using absolute imports as much
302as possible, so it's preferable to begin writing \code{from pkg import
303string} in your code.
304
305Relative imports are still possible by adding a leading period
306to the module name when using the \code{from ... import} form:
307
308\begin{verbatim}
309# Import names from pkg.string
310from .string import name1, name2
311# Import pkg.string
312from . import string
313\end{verbatim}
314
315This imports the \module{string} module relative to the current
316package, so in \module{pkg.main} this will import \var{name1} and
317\var{name2} from \module{pkg.string}. Additional leading periods
318perform the relative import starting from the parent of the current
319package. For example, code in the \module{A.B.C} module can do:
320
321\begin{verbatim}
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000322from . import D # Imports A.B.D
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000323from .. import E # Imports A.E
324from ..F import G # Imports A.F.G
325\end{verbatim}
326
327Leading periods cannot be used with the \code{import \var{modname}}
328form of the import statement, only the \code{from ... import} form.
329
330\begin{seealso}
331
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +0000332\seepep{328}{Imports: Multi-Line and Absolute/Relative}
333{PEP written by Aahz; implemented by Thomas Wouters.}
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000334
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +0000335\seeurl{http://codespeak.net/py/current/doc/index.html}
336{The py library by Holger Krekel, which contains the \module{py.std} package.}
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000337
338\end{seealso}
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000339
340
341%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000342\section{PEP 338: Executing Modules as Scripts\label{pep-338}}
Andrew M. Kuchling21d3a7c2006-03-15 11:53:09 +0000343
Andrew M. Kuchlingb182db42006-03-17 21:48:46 +0000344The \programopt{-m} switch added in Python 2.4 to execute a module as
345a script gained a few more abilities. Instead of being implemented in
346C code inside the Python interpreter, the switch now uses an
347implementation in a new module, \module{runpy}.
348
349The \module{runpy} module implements a more sophisticated import
350mechanism so that it's now possible to run modules in a package such
351as \module{pychecker.checker}. The module also supports alternative
Andrew M. Kuchling5d4cf5e2006-04-13 13:02:42 +0000352import mechanisms such as the \module{zipimport} module. This means
Andrew M. Kuchlingb182db42006-03-17 21:48:46 +0000353you can add a .zip archive's path to \code{sys.path} and then use the
354\programopt{-m} switch to execute code from the archive.
355
356
357\begin{seealso}
358
359\seepep{338}{Executing modules as scripts}{PEP written and
360implemented by Nick Coghlan.}
361
362\end{seealso}
Andrew M. Kuchling21d3a7c2006-03-15 11:53:09 +0000363
364
365%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000366\section{PEP 341: Unified try/except/finally\label{pep-341}}
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000367
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000368Until Python 2.5, the \keyword{try} statement came in two
369flavours. You could use a \keyword{finally} block to ensure that code
Andrew M. Kuchling0f1955d2006-04-13 12:09:08 +0000370is always executed, or one or more \keyword{except} blocks to catch
371specific exceptions. You couldn't combine both \keyword{except} blocks and a
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000372\keyword{finally} block, because generating the right bytecode for the
373combined version was complicated and it wasn't clear what the
374semantics of the combined should be.
375
376GvR spent some time working with Java, which does support the
377equivalent of combining \keyword{except} blocks and a
378\keyword{finally} block, and this clarified what the statement should
379mean. In Python 2.5, you can now write:
380
381\begin{verbatim}
382try:
383 block-1 ...
384except Exception1:
385 handler-1 ...
386except Exception2:
387 handler-2 ...
388else:
389 else-block
390finally:
391 final-block
392\end{verbatim}
393
394The code in \var{block-1} is executed. If the code raises an
395exception, the handlers are tried in order: \var{handler-1},
396\var{handler-2}, ... If no exception is raised, the \var{else-block}
397is executed. No matter what happened previously, the
398\var{final-block} is executed once the code block is complete and any
399raised exceptions handled. Even if there's an error in an exception
400handler or the \var{else-block} and a new exception is raised, the
401\var{final-block} is still executed.
402
403\begin{seealso}
404
405\seepep{341}{Unifying try-except and try-finally}{PEP written by Georg Brandl;
Andrew M. Kuchling9c67ee02006-04-04 19:07:27 +0000406implementation by Thomas Lee.}
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000407
408\end{seealso}
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000409
410
411%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000412\section{PEP 342: New Generator Features\label{pep-342}}
Andrew M. Kuchlinga2e21cb2005-08-02 17:13:21 +0000413
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000414Python 2.5 adds a simple way to pass values \emph{into} a generator.
Andrew M. Kuchling150e3492005-08-23 00:56:06 +0000415As introduced in Python 2.3, generators only produce output; once a
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000416generator's code is invoked to create an iterator, there's no way to
417pass any new information into the function when its execution is
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000418resumed. Sometimes the ability to pass in some information would be
419useful. Hackish solutions to this include making the generator's code
420look at a global variable and then changing the global variable's
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000421value, or passing in some mutable object that callers then modify.
Andrew M. Kuchling150e3492005-08-23 00:56:06 +0000422
423To refresh your memory of basic generators, here's a simple example:
424
425\begin{verbatim}
426def counter (maximum):
427 i = 0
428 while i < maximum:
429 yield i
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000430 i += 1
Andrew M. Kuchling150e3492005-08-23 00:56:06 +0000431\end{verbatim}
432
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000433When you call \code{counter(10)}, the result is an iterator that
434returns the values from 0 up to 9. On encountering the
435\keyword{yield} statement, the iterator returns the provided value and
436suspends the function's execution, preserving the local variables.
437Execution resumes on the following call to the iterator's
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000438\method{next()} method, picking up after the \keyword{yield} statement.
Andrew M. Kuchling150e3492005-08-23 00:56:06 +0000439
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000440In Python 2.3, \keyword{yield} was a statement; it didn't return any
441value. In 2.5, \keyword{yield} is now an expression, returning a
442value that can be assigned to a variable or otherwise operated on:
Andrew M. Kuchlinga2e21cb2005-08-02 17:13:21 +0000443
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000444\begin{verbatim}
445val = (yield i)
446\end{verbatim}
447
448I recommend that you always put parentheses around a \keyword{yield}
449expression when you're doing something with the returned value, as in
450the above example. The parentheses aren't always necessary, but it's
451easier to always add them instead of having to remember when they're
Andrew M. Kuchling3b675d22006-04-20 13:43:21 +0000452needed.
453
454(\pep{342} explains the exact rules, which are that a
455\keyword{yield}-expression must always be parenthesized except when it
456occurs at the top-level expression on the right-hand side of an
457assignment. This means you can write \code{val = yield i} but have to
458use parentheses when there's an operation, as in \code{val = (yield i)
459+ 12}.)
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000460
461Values are sent into a generator by calling its
462\method{send(\var{value})} method. The generator's code is then
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000463resumed and the \keyword{yield} expression returns the specified
464\var{value}. If the regular \method{next()} method is called, the
465\keyword{yield} returns \constant{None}.
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000466
467Here's the previous example, modified to allow changing the value of
468the internal counter.
469
470\begin{verbatim}
471def counter (maximum):
472 i = 0
473 while i < maximum:
474 val = (yield i)
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000475 # If value provided, change counter
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000476 if val is not None:
477 i = val
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000478 else:
479 i += 1
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000480\end{verbatim}
481
482And here's an example of changing the counter:
483
484\begin{verbatim}
485>>> it = counter(10)
486>>> print it.next()
4870
488>>> print it.next()
4891
490>>> print it.send(8)
4918
492>>> print it.next()
4939
494>>> print it.next()
495Traceback (most recent call last):
496 File ``t.py'', line 15, in ?
497 print it.next()
498StopIteration
Andrew M. Kuchlingc2033702005-08-29 13:30:12 +0000499\end{verbatim}
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000500
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000501Because \keyword{yield} will often be returning \constant{None}, you
502should always check for this case. Don't just use its value in
503expressions unless you're sure that the \method{send()} method
504will be the only method used resume your generator function.
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000505
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000506In addition to \method{send()}, there are two other new methods on
507generators:
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000508
509\begin{itemize}
510
511 \item \method{throw(\var{type}, \var{value}=None,
512 \var{traceback}=None)} is used to raise an exception inside the
513 generator; the exception is raised by the \keyword{yield} expression
514 where the generator's execution is paused.
515
516 \item \method{close()} raises a new \exception{GeneratorExit}
517 exception inside the generator to terminate the iteration.
518 On receiving this
519 exception, the generator's code must either raise
520 \exception{GeneratorExit} or \exception{StopIteration}; catching the
521 exception and doing anything else is illegal and will trigger
522 a \exception{RuntimeError}. \method{close()} will also be called by
523 Python's garbage collection when the generator is garbage-collected.
524
525 If you need to run cleanup code in case of a \exception{GeneratorExit},
526 I suggest using a \code{try: ... finally:} suite instead of
527 catching \exception{GeneratorExit}.
528
529\end{itemize}
530
531The cumulative effect of these changes is to turn generators from
532one-way producers of information into both producers and consumers.
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000533
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000534Generators also become \emph{coroutines}, a more generalized form of
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000535subroutines. Subroutines are entered at one point and exited at
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000536another point (the top of the function, and a \keyword{return
537statement}), but coroutines can be entered, exited, and resumed at
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000538many different points (the \keyword{yield} statements). We'll have to
539figure out patterns for using coroutines effectively in Python.
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000540
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000541The addition of the \method{close()} method has one side effect that
542isn't obvious. \method{close()} is called when a generator is
543garbage-collected, so this means the generator's code gets one last
Andrew M. Kuchling3b4fb042006-04-13 12:49:39 +0000544chance to run before the generator is destroyed. This last chance
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000545means that \code{try...finally} statements in generators can now be
546guaranteed to work; the \keyword{finally} clause will now always get a
547chance to run. The syntactic restriction that you couldn't mix
548\keyword{yield} statements with a \code{try...finally} suite has
549therefore been removed. This seems like a minor bit of language
550trivia, but using generators and \code{try...finally} is actually
551necessary in order to implement the \keyword{with} statement
Andrew M. Kuchling67191312006-04-19 12:55:39 +0000552described by PEP 343. I'll look at this new statement in the following
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000553section.
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000554
Andrew M. Kuchling3b4fb042006-04-13 12:49:39 +0000555Another even more esoteric effect of this change: previously, the
556\member{gi_frame} attribute of a generator was always a frame object.
557It's now possible for \member{gi_frame} to be \code{None}
558once the generator has been exhausted.
559
Andrew M. Kuchlinga2e21cb2005-08-02 17:13:21 +0000560\begin{seealso}
561
562\seepep{342}{Coroutines via Enhanced Generators}{PEP written by
Andrew M. Kuchling67191312006-04-19 12:55:39 +0000563Guido van~Rossum and Phillip J. Eby;
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000564implemented by Phillip J. Eby. Includes examples of
565some fancier uses of generators as coroutines.}
566
567\seeurl{http://en.wikipedia.org/wiki/Coroutine}{The Wikipedia entry for
568coroutines.}
569
Neal Norwitz09179882006-03-04 23:31:45 +0000570\seeurl{http://www.sidhe.org/\~{}dan/blog/archives/000178.html}{An
Andrew M. Kuchling07382062005-08-27 18:45:47 +0000571explanation of coroutines from a Perl point of view, written by Dan
572Sugalski.}
Andrew M. Kuchlinga2e21cb2005-08-02 17:13:21 +0000573
574\end{seealso}
575
576
577%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000578\section{PEP 343: The 'with' statement\label{pep-343}}
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000579
Andrew M. Kuchling0a7ed8c2006-04-24 14:30:47 +0000580The '\keyword{with}' statement clarifies code that previously would
581use \code{try...finally} blocks to ensure that clean-up code is
582executed. In this section, I'll discuss the statement as it will
583commonly be used. In the next section, I'll examine the
584implementation details and show how to write objects for use with this
585statement.
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000586
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +0000587The '\keyword{with}' statement is a new control-flow structure whose
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000588basic structure is:
589
590\begin{verbatim}
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000591with expression [as variable]:
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000592 with-block
593\end{verbatim}
594
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000595The expression is evaluated, and it should result in an object that
596supports the context management protocol. This object may return a
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000597value that can optionally be bound to the name \var{variable}. (Note
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000598carefully that \var{variable} is \emph{not} assigned the result of
599\var{expression}.) The object can then run set-up code
600before \var{with-block} is executed and some clean-up code
601is executed after the block is done, even if the block raised an exception.
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000602
603To enable the statement in Python 2.5, you need
604to add the following directive to your module:
605
606\begin{verbatim}
607from __future__ import with_statement
608\end{verbatim}
609
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000610The statement will always be enabled in Python 2.6.
611
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000612Some standard Python objects now support the context management
613protocol and can be used with the '\keyword{with}' statement. File
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000614objects are one example:
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000615
616\begin{verbatim}
617with open('/etc/passwd', 'r') as f:
618 for line in f:
619 print line
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000620 ... more processing code ...
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000621\end{verbatim}
622
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000623After this statement has executed, the file object in \var{f} will
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +0000624have been automatically closed, even if the 'for' loop
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000625raised an exception part-way through the block.
626
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000627The \module{threading} module's locks and condition variables
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +0000628also support the '\keyword{with}' statement:
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000629
630\begin{verbatim}
631lock = threading.Lock()
632with lock:
633 # Critical section of code
634 ...
635\end{verbatim}
636
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000637The lock is acquired before the block is executed and always released once
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000638the block is complete.
639
640The \module{decimal} module's contexts, which encapsulate the desired
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000641precision and rounding characteristics for computations, also work.
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000642
643\begin{verbatim}
644import decimal
645
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000646# Displays with default precision of 28 digits
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000647v1 = decimal.Decimal('578')
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000648print v1.sqrt()
649
650with decimal.Context(prec=16):
651 # All code in this block uses a precision of 16 digits.
652 # The original context is restored on exiting the block.
653 print v1.sqrt()
654\end{verbatim}
655
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000656\subsection{Writing Context Managers\label{context-managers}}
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000657
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +0000658Under the hood, the '\keyword{with}' statement is fairly complicated.
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000659Most people will only use '\keyword{with}' in company with existing
Andrew M. Kuchling0a7ed8c2006-04-24 14:30:47 +0000660objects and don't need to know these details, so you can skip the rest
661of this section if you like. Authors of new objects will need to
662understand the details of the underlying implementation and should
663keep reading.
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000664
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000665A high-level explanation of the context management protocol is:
666
667\begin{itemize}
668\item The expression is evaluated and should result in an object
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000669with a \method{__context__()} method (called a ``context manager'').
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000670
Andrew M. Kuchling0a7ed8c2006-04-24 14:30:47 +0000671\item The context specifier's \method{__context__()} method is called,
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000672and must return another object (called a ``with-statement context object'') that has
Andrew M. Kuchling0a7ed8c2006-04-24 14:30:47 +0000673\method{__enter__()} and \method{__exit__()} methods.
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000674
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000675\item The context object's \method{__enter__()} method is called. The value
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000676returned is assigned to \var{VAR}. If no \code{'as \var{VAR}'} clause
677is present, the value is simply discarded.
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000678
679\item The code in \var{BLOCK} is executed.
680
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000681\item If \var{BLOCK} raises an exception, the
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000682\method{__exit__(\var{type}, \var{value}, \var{traceback})} is called
683with the exception's information, the same values returned by
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000684\function{sys.exc_info()}. The method's return value controls whether
685the exception is re-raised: any false value re-raises the exception,
686and \code{True} will result in suppressing it. You'll only rarely
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000687want to suppress the exception, because if you do
688the author of the code containing the
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000689'\keyword{with}' statement will never realize anything went wrong.
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000690
691\item If \var{BLOCK} didn't raise an exception,
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000692the \method{__exit__()} method is still called,
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000693but \var{type}, \var{value}, and \var{traceback} are all \code{None}.
694
695\end{itemize}
696
697Let's think through an example. I won't present detailed code but
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000698will only sketch the methods necessary for a database that supports
699transactions.
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000700
701(For people unfamiliar with database terminology: a set of changes to
702the database are grouped into a transaction. Transactions can be
703either committed, meaning that all the changes are written into the
704database, or rolled back, meaning that the changes are all discarded
705and the database is unchanged. See any database textbook for more
706information.)
707% XXX find a shorter reference?
708
709Let's assume there's an object representing a database connection.
710Our goal will be to let the user write code like this:
711
712\begin{verbatim}
713db_connection = DatabaseConnection()
714with db_connection as cursor:
715 cursor.execute('insert into ...')
716 cursor.execute('delete from ...')
717 # ... more operations ...
718\end{verbatim}
719
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000720The transaction should be committed if the code in the block
721runs flawlessly or rolled back if there's an exception.
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000722
723First, the \class{DatabaseConnection} needs a \method{__context__()}
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000724method. Sometimes an object can simply return \code{self}; the
725\module{threading} module's lock objects do this, for example. For
726our database example, though, we need to create a new object; I'll
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000727call this class \class{DatabaseContext}. Our \method{__context__()}
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000728method must therefore look like this:
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000729
730\begin{verbatim}
731class DatabaseConnection:
732 ...
733 def __context__ (self):
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000734 return DatabaseContext(self)
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000735
736 # Database interface
737 def cursor (self):
738 "Returns a cursor object and starts a new transaction"
739 def commit (self):
740 "Commits current transaction"
741 def rollback (self):
742 "Rolls back current transaction"
743\end{verbatim}
744
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000745Instances of \class{DatabaseContext} need the connection object so that
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000746the connection object's \method{commit()} or \method{rollback()}
747methods can be called:
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000748
749\begin{verbatim}
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000750class DatabaseContext:
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000751 def __init__ (self, connection):
752 self.connection = connection
753\end{verbatim}
754
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000755The \method {__enter__()} method is pretty easy, having only to start
756a new transaction. For this application the resulting cursor object
757would be a useful result, so the method will return it. The user can
758then add \code{as cursor} to their '\keyword{with}' statement to bind
759the cursor to a variable name.
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000760
761\begin{verbatim}
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000762class DatabaseContext:
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000763 ...
764 def __enter__ (self):
765 # Code to start a new transaction
766 cursor = self.connection.cursor()
767 return cursor
768\end{verbatim}
769
770The \method{__exit__()} method is the most complicated because it's
771where most of the work has to be done. The method has to check if an
772exception occurred. If there was no exception, the transaction is
773committed. The transaction is rolled back if there was an exception.
Andrew M. Kuchling0a7ed8c2006-04-24 14:30:47 +0000774
775In the code below, execution will just fall off the end of the
776function, returning the default value of \code{None}. \code{None} is
777false, so the exception will be re-raised automatically. If you
778wished, you could be more explicit and add a \keyword{return}
779statement at the marked location.
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000780
781\begin{verbatim}
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000782class DatabaseContext:
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000783 ...
784 def __exit__ (self, type, value, tb):
785 if tb is None:
786 # No exception, so commit
787 self.connection.commit()
788 else:
789 # Exception occurred, so rollback.
790 self.connection.rollback()
791 # return False
792\end{verbatim}
793
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +0000794
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +0000795\subsection{The contextlib module\label{module-contextlib}}
Andrew M. Kuchling9c67ee02006-04-04 19:07:27 +0000796
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000797The new \module{contextlib} module provides some functions and a
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000798decorator that are useful for writing objects for use with the
799'\keyword{with}' statement.
Andrew M. Kuchling9c67ee02006-04-04 19:07:27 +0000800
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000801The decorator is called \function{contextfactory}, and lets you write
802a single generator function instead of defining a new class. The generator
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000803should yield exactly one value. The code up to the \keyword{yield}
804will be executed as the \method{__enter__()} method, and the value
805yielded will be the method's return value that will get bound to the
806variable in the '\keyword{with}' statement's \keyword{as} clause, if
807any. The code after the \keyword{yield} will be executed in the
808\method{__exit__()} method. Any exception raised in the block will be
809raised by the \keyword{yield} statement.
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +0000810
811Our database example from the previous section could be written
812using this decorator as:
813
814\begin{verbatim}
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000815from contextlib import contextfactory
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +0000816
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000817@contextfactory
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +0000818def db_transaction (connection):
819 cursor = connection.cursor()
820 try:
821 yield cursor
822 except:
823 connection.rollback()
824 raise
825 else:
826 connection.commit()
827
828db = DatabaseConnection()
829with db_transaction(db) as cursor:
830 ...
831\end{verbatim}
832
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000833You can also use this decorator to write the \method{__context__()}
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000834method for a class:
Andrew M. Kuchling67191312006-04-19 12:55:39 +0000835
836\begin{verbatim}
837class DatabaseConnection:
838
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000839 @contextfactory
Andrew M. Kuchling67191312006-04-19 12:55:39 +0000840 def __context__ (self):
841 cursor = self.cursor()
842 try:
843 yield cursor
844 except:
845 self.rollback()
846 raise
847 else:
848 self.commit()
849\end{verbatim}
850
851
Andrew M. Kuchlingd798a182006-04-25 12:47:25 +0000852The \module{contextlib} module also has a \function{nested(\var{mgr1},
853\var{mgr2}, ...)} function that combines a number of contexts so you
854don't need to write nested '\keyword{with}' statements. In this
855example, the single '\keyword{with}' statement both starts a database
856transaction and acquires a thread lock:
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +0000857
858\begin{verbatim}
859lock = threading.Lock()
860with nested (db_transaction(db), lock) as (cursor, locked):
861 ...
862\end{verbatim}
863
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000864Finally, the \function{closing(\var{object})} function
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +0000865returns \var{object} so that it can be bound to a variable,
866and calls \code{\var{object}.close()} at the end of the block.
867
868\begin{verbatim}
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +0000869import urllib, sys
870from contextlib import closing
871
872with closing(urllib.urlopen('http://www.yahoo.com')) as f:
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +0000873 for line in f:
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +0000874 sys.stdout.write(line)
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +0000875\end{verbatim}
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000876
877\begin{seealso}
878
Andrew M. Kuchling67191312006-04-19 12:55:39 +0000879\seepep{343}{The ``with'' statement}{PEP written by Guido van~Rossum
880and Nick Coghlan; implemented by Mike Bland, Guido van~Rossum, and
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +0000881Neal Norwitz. The PEP shows the code generated for a '\keyword{with}'
Andrew M. Kuchlingedb575e2006-04-23 21:01:04 +0000882statement, which can be helpful in learning how the statement works.}
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +0000883
884\seeurl{../lib/module-contextlib.html}{The documentation
885for the \module{contextlib} module.}
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000886
887\end{seealso}
888
Andrew M. Kuchling437567c2006-03-07 20:48:55 +0000889
890%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000891\section{PEP 352: Exceptions as New-Style Classes\label{pep-352}}
Andrew M. Kuchling8f4d2552006-03-08 01:50:20 +0000892
Andrew M. Kuchling38f85072006-04-02 01:46:32 +0000893Exception classes can now be new-style classes, not just classic
894classes, and the built-in \exception{Exception} class and all the
895standard built-in exceptions (\exception{NameError},
896\exception{ValueError}, etc.) are now new-style classes.
Andrew M. Kuchlingaeadf952006-03-09 19:06:05 +0000897
898The inheritance hierarchy for exceptions has been rearranged a bit.
899In 2.5, the inheritance relationships are:
900
901\begin{verbatim}
902BaseException # New in Python 2.5
903|- KeyboardInterrupt
904|- SystemExit
905|- Exception
906 |- (all other current built-in exceptions)
907\end{verbatim}
908
909This rearrangement was done because people often want to catch all
910exceptions that indicate program errors. \exception{KeyboardInterrupt} and
911\exception{SystemExit} aren't errors, though, and usually represent an explicit
912action such as the user hitting Control-C or code calling
913\function{sys.exit()}. A bare \code{except:} will catch all exceptions,
914so you commonly need to list \exception{KeyboardInterrupt} and
915\exception{SystemExit} in order to re-raise them. The usual pattern is:
916
917\begin{verbatim}
918try:
919 ...
920except (KeyboardInterrupt, SystemExit):
921 raise
922except:
923 # Log error...
924 # Continue running program...
925\end{verbatim}
926
927In Python 2.5, you can now write \code{except Exception} to achieve
928the same result, catching all the exceptions that usually indicate errors
929but leaving \exception{KeyboardInterrupt} and
930\exception{SystemExit} alone. As in previous versions,
931a bare \code{except:} still catches all exceptions.
932
933The goal for Python 3.0 is to require any class raised as an exception
934to derive from \exception{BaseException} or some descendant of
935\exception{BaseException}, and future releases in the
936Python 2.x series may begin to enforce this constraint. Therefore, I
937suggest you begin making all your exception classes derive from
938\exception{Exception} now. It's been suggested that the bare
939\code{except:} form should be removed in Python 3.0, but Guido van~Rossum
940hasn't decided whether to do this or not.
941
942Raising of strings as exceptions, as in the statement \code{raise
943"Error occurred"}, is deprecated in Python 2.5 and will trigger a
944warning. The aim is to be able to remove the string-exception feature
945in a few releases.
946
947
948\begin{seealso}
949
Andrew M. Kuchlingc3749a92006-04-04 19:14:41 +0000950\seepep{352}{Required Superclass for Exceptions}{PEP written by
Andrew M. Kuchling67191312006-04-19 12:55:39 +0000951Brett Cannon and Guido van~Rossum; implemented by Brett Cannon.}
Andrew M. Kuchlingaeadf952006-03-09 19:06:05 +0000952
953\end{seealso}
Andrew M. Kuchling8f4d2552006-03-08 01:50:20 +0000954
955
956%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +0000957\section{PEP 353: Using ssize_t as the index type\label{pep-353}}
Andrew M. Kuchlingc3749a92006-04-04 19:14:41 +0000958
959A wide-ranging change to Python's C API, using a new
960\ctype{Py_ssize_t} type definition instead of \ctype{int},
961will permit the interpreter to handle more data on 64-bit platforms.
962This change doesn't affect Python's capacity on 32-bit platforms.
963
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +0000964Various pieces of the Python interpreter used C's \ctype{int} type to
965store sizes or counts; for example, the number of items in a list or
966tuple were stored in an \ctype{int}. The C compilers for most 64-bit
967platforms still define \ctype{int} as a 32-bit type, so that meant
968that lists could only hold up to \code{2**31 - 1} = 2147483647 items.
969(There are actually a few different programming models that 64-bit C
970compilers can use -- see
971\url{http://www.unix.org/version2/whatsnew/lp64_wp.html} for a
972discussion -- but the most commonly available model leaves \ctype{int}
973as 32 bits.)
974
975A limit of 2147483647 items doesn't really matter on a 32-bit platform
976because you'll run out of memory before hitting the length limit.
977Each list item requires space for a pointer, which is 4 bytes, plus
978space for a \ctype{PyObject} representing the item. 2147483647*4 is
979already more bytes than a 32-bit address space can contain.
980
981It's possible to address that much memory on a 64-bit platform,
982however. The pointers for a list that size would only require 16GiB
983of space, so it's not unreasonable that Python programmers might
984construct lists that large. Therefore, the Python interpreter had to
985be changed to use some type other than \ctype{int}, and this will be a
98664-bit type on 64-bit platforms. The change will cause
987incompatibilities on 64-bit machines, so it was deemed worth making
988the transition now, while the number of 64-bit users is still
989relatively small. (In 5 or 10 years, we may \emph{all} be on 64-bit
990machines, and the transition would be more painful then.)
991
992This change most strongly affects authors of C extension modules.
993Python strings and container types such as lists and tuples
994now use \ctype{Py_ssize_t} to store their size.
995Functions such as \cfunction{PyList_Size()}
996now return \ctype{Py_ssize_t}. Code in extension modules
997may therefore need to have some variables changed to
998\ctype{Py_ssize_t}.
999
1000The \cfunction{PyArg_ParseTuple()} and \cfunction{Py_BuildValue()} functions
1001have a new conversion code, \samp{n}, for \ctype{Py_ssize_t}.
Andrew M. Kuchlinga4d651f2006-04-06 13:24:58 +00001002\cfunction{PyArg_ParseTuple()}'s \samp{s\#} and \samp{t\#} still output
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +00001003\ctype{int} by default, but you can define the macro
1004\csimplemacro{PY_SSIZE_T_CLEAN} before including \file{Python.h}
1005to make them return \ctype{Py_ssize_t}.
1006
1007\pep{353} has a section on conversion guidelines that
1008extension authors should read to learn about supporting 64-bit
1009platforms.
Andrew M. Kuchlingc3749a92006-04-04 19:14:41 +00001010
1011\begin{seealso}
1012
Andrew M. Kuchling5f445bf2006-04-12 18:54:00 +00001013\seepep{353}{Using ssize_t as the index type}{PEP written and implemented by Martin von~L\"owis.}
Andrew M. Kuchlingc3749a92006-04-04 19:14:41 +00001014
1015\end{seealso}
1016
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +00001017
Andrew M. Kuchlingc3749a92006-04-04 19:14:41 +00001018%======================================================================
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +00001019\section{PEP 357: The '__index__' method\label{pep-357}}
Andrew M. Kuchling437567c2006-03-07 20:48:55 +00001020
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001021The NumPy developers had a problem that could only be solved by adding
1022a new special method, \method{__index__}. When using slice notation,
Fred Drake1c0e3282006-04-02 03:30:06 +00001023as in \code{[\var{start}:\var{stop}:\var{step}]}, the values of the
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001024\var{start}, \var{stop}, and \var{step} indexes must all be either
1025integers or long integers. NumPy defines a variety of specialized
1026integer types corresponding to unsigned and signed integers of 8, 16,
102732, and 64 bits, but there was no way to signal that these types could
1028be used as slice indexes.
1029
1030Slicing can't just use the existing \method{__int__} method because
1031that method is also used to implement coercion to integers. If
1032slicing used \method{__int__}, floating-point numbers would also
1033become legal slice indexes and that's clearly an undesirable
1034behaviour.
1035
1036Instead, a new special method called \method{__index__} was added. It
1037takes no arguments and returns an integer giving the slice index to
1038use. For example:
1039
1040\begin{verbatim}
1041class C:
1042 def __index__ (self):
1043 return self.value
1044\end{verbatim}
1045
1046The return value must be either a Python integer or long integer.
1047The interpreter will check that the type returned is correct, and
1048raises a \exception{TypeError} if this requirement isn't met.
1049
1050A corresponding \member{nb_index} slot was added to the C-level
1051\ctype{PyNumberMethods} structure to let C extensions implement this
1052protocol. \cfunction{PyNumber_Index(\var{obj})} can be used in
1053extension code to call the \method{__index__} function and retrieve
1054its result.
1055
1056\begin{seealso}
1057
1058\seepep{357}{Allowing Any Object to be Used for Slicing}{PEP written
Andrew M. Kuchling9c67ee02006-04-04 19:07:27 +00001059and implemented by Travis Oliphant.}
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001060
1061\end{seealso}
Andrew M. Kuchling437567c2006-03-07 20:48:55 +00001062
1063
1064%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001065\section{Other Language Changes\label{other-lang}}
Fred Drake2db76802004-12-01 05:05:47 +00001066
1067Here are all of the changes that Python 2.5 makes to the core Python
1068language.
1069
1070\begin{itemize}
Andrew M. Kuchling1cae3f52004-12-03 14:57:21 +00001071
Andrew M. Kuchlingc7095842006-04-14 12:41:19 +00001072\item The \class{dict} type has a new hook for letting subclasses
1073provide a default value when a key isn't contained in the dictionary.
1074When a key isn't found, the dictionary's
1075\method{__missing__(\var{key})}
1076method will be called. This hook is used to implement
1077the new \class{defaultdict} class in the \module{collections}
1078module. The following example defines a dictionary
1079that returns zero for any missing key:
1080
1081\begin{verbatim}
1082class zerodict (dict):
1083 def __missing__ (self, key):
1084 return 0
1085
1086d = zerodict({1:1, 2:2})
1087print d[1], d[2] # Prints 1, 2
1088print d[3], d[4] # Prints 0, 0
1089\end{verbatim}
1090
Andrew M. Kuchling1cae3f52004-12-03 14:57:21 +00001091\item The \function{min()} and \function{max()} built-in functions
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +00001092gained a \code{key} keyword parameter analogous to the \code{key}
1093argument for \method{sort()}. This parameter supplies a function that
Andrew M. Kuchlingc7095842006-04-14 12:41:19 +00001094takes a single argument and is called for every value in the list;
Andrew M. Kuchling1cae3f52004-12-03 14:57:21 +00001095\function{min()}/\function{max()} will return the element with the
1096smallest/largest return value from this function.
1097For example, to find the longest string in a list, you can do:
1098
1099\begin{verbatim}
1100L = ['medium', 'longest', 'short']
1101# Prints 'longest'
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +00001102print max(L, key=len)
Andrew M. Kuchling1cae3f52004-12-03 14:57:21 +00001103# Prints 'short', because lexicographically 'short' has the largest value
1104print max(L)
1105\end{verbatim}
1106
1107(Contributed by Steven Bethard and Raymond Hettinger.)
Fred Drake2db76802004-12-01 05:05:47 +00001108
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001109\item Two new built-in functions, \function{any()} and
1110\function{all()}, evaluate whether an iterator contains any true or
1111false values. \function{any()} returns \constant{True} if any value
1112returned by the iterator is true; otherwise it will return
1113\constant{False}. \function{all()} returns \constant{True} only if
1114all of the values returned by the iterator evaluate as being true.
Andrew M. Kuchling6e3a66d2006-04-07 12:46:06 +00001115(Suggested by GvR, and implemented by Raymond Hettinger.)
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001116
Andrew M. Kuchling5f445bf2006-04-12 18:54:00 +00001117\item ASCII is now the default encoding for modules. It's now
1118a syntax error if a module contains string literals with 8-bit
1119characters but doesn't have an encoding declaration. In Python 2.4
1120this triggered a warning, not a syntax error. See \pep{263}
1121for how to declare a module's encoding; for example, you might add
1122a line like this near the top of the source file:
1123
1124\begin{verbatim}
1125# -*- coding: latin1 -*-
1126\end{verbatim}
1127
Andrew M. Kuchlingc9236112006-04-30 01:07:09 +00001128\item One error that Python programmers sometimes make is forgetting
1129to include an \file{__init__.py} module in a package directory.
1130Debugging this mistake can be confusing, and usually requires running
1131Python with the \programopt{-v} switch to log all the paths searched.
1132In Python 2.5, a new \exception{ImportWarning} warning is raised when
1133an import would have picked up a directory as a package but no
1134\file{__init__.py} was found. (Implemented by Thomas Wouters.)
1135
Andrew M. Kuchlinge9b1bf42005-03-20 19:26:30 +00001136\item The list of base classes in a class definition can now be empty.
1137As an example, this is now legal:
1138
1139\begin{verbatim}
1140class C():
1141 pass
1142\end{verbatim}
1143(Implemented by Brett Cannon.)
1144
Fred Drake2db76802004-12-01 05:05:47 +00001145\end{itemize}
1146
1147
1148%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001149\subsection{Interactive Interpreter Changes\label{interactive}}
Andrew M. Kuchlingda376042006-03-17 15:56:41 +00001150
1151In the interactive interpreter, \code{quit} and \code{exit}
1152have long been strings so that new users get a somewhat helpful message
1153when they try to quit:
1154
1155\begin{verbatim}
1156>>> quit
1157'Use Ctrl-D (i.e. EOF) to exit.'
1158\end{verbatim}
1159
1160In Python 2.5, \code{quit} and \code{exit} are now objects that still
1161produce string representations of themselves, but are also callable.
1162Newbies who try \code{quit()} or \code{exit()} will now exit the
1163interpreter as they expect. (Implemented by Georg Brandl.)
1164
1165
1166%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001167\subsection{Optimizations\label{opts}}
Fred Drake2db76802004-12-01 05:05:47 +00001168
1169\begin{itemize}
1170
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001171\item When they were introduced
1172in Python 2.4, the built-in \class{set} and \class{frozenset} types
1173were built on top of Python's dictionary type.
1174In 2.5 the internal data structure has been customized for implementing sets,
1175and as a result sets will use a third less memory and are somewhat faster.
1176(Implemented by Raymond Hettinger.)
Fred Drake2db76802004-12-01 05:05:47 +00001177
Andrew M. Kuchling45bb98e2006-04-16 19:53:27 +00001178\item The performance of some Unicode operations, such as
1179character map decoding, has been improved.
1180% Patch 1313939
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001181
1182\item The code generator's peephole optimizer now performs
1183simple constant folding in expressions. If you write something like
1184\code{a = 2+3}, the code generator will do the arithmetic and produce
1185code corresponding to \code{a = 5}.
1186
Fred Drake2db76802004-12-01 05:05:47 +00001187\end{itemize}
1188
1189The net result of the 2.5 optimizations is that Python 2.5 runs the
Andrew M. Kuchling9c67ee02006-04-04 19:07:27 +00001190pystone benchmark around XXX\% faster than Python 2.4.
Fred Drake2db76802004-12-01 05:05:47 +00001191
1192
1193%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001194\section{New, Improved, and Removed Modules\label{modules}}
Fred Drake2db76802004-12-01 05:05:47 +00001195
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +00001196The standard library received many enhancements and bug fixes in
1197Python 2.5. Here's a partial list of the most notable changes, sorted
1198alphabetically by module name. Consult the \file{Misc/NEWS} file in
1199the source tree for a more complete list of changes, or look through
1200the SVN logs for all the details.
Fred Drake2db76802004-12-01 05:05:47 +00001201
1202\begin{itemize}
1203
Andrew M. Kuchling6fc69762006-04-13 12:37:21 +00001204\item The \module{audioop} module now supports the a-LAW encoding,
1205and the code for u-LAW encoding has been improved. (Contributed by
1206Lars Immisch.)
1207
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +00001208\item The \module{codecs} module gained support for incremental
1209codecs. The \function{codec.lookup()} function now
1210returns a \class{CodecInfo} instance instead of a tuple.
1211\class{CodecInfo} instances behave like a 4-tuple to preserve backward
1212compatibility but also have the attributes \member{encode},
1213\member{decode}, \member{incrementalencoder}, \member{incrementaldecoder},
1214\member{streamwriter}, and \member{streamreader}. Incremental codecs
1215can receive input and produce output in multiple chunks; the output is
1216the same as if the entire input was fed to the non-incremental codec.
1217See the \module{codecs} module documentation for details.
1218(Designed and implemented by Walter D\"orwald.)
1219% Patch 1436130
1220
Andrew M. Kuchlingc7095842006-04-14 12:41:19 +00001221\item The \module{collections} module gained a new type,
1222\class{defaultdict}, that subclasses the standard \class{dict}
1223type. The new type mostly behaves like a dictionary but constructs a
1224default value when a key isn't present, automatically adding it to the
1225dictionary for the requested key value.
1226
1227The first argument to \class{defaultdict}'s constructor is a factory
1228function that gets called whenever a key is requested but not found.
1229This factory function receives no arguments, so you can use built-in
1230type constructors such as \function{list()} or \function{int()}. For
1231example,
1232you can make an index of words based on their initial letter like this:
1233
1234\begin{verbatim}
1235words = """Nel mezzo del cammin di nostra vita
1236mi ritrovai per una selva oscura
1237che la diritta via era smarrita""".lower().split()
1238
1239index = defaultdict(list)
1240
1241for w in words:
1242 init_letter = w[0]
1243 index[init_letter].append(w)
1244\end{verbatim}
1245
1246Printing \code{index} results in the following output:
1247
1248\begin{verbatim}
1249defaultdict(<type 'list'>, {'c': ['cammin', 'che'], 'e': ['era'],
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +00001250 'd': ['del', 'di', 'diritta'], 'm': ['mezzo', 'mi'],
1251 'l': ['la'], 'o': ['oscura'], 'n': ['nel', 'nostra'],
1252 'p': ['per'], 's': ['selva', 'smarrita'],
1253 'r': ['ritrovai'], 'u': ['una'], 'v': ['vita', 'via']}
Andrew M. Kuchlingc7095842006-04-14 12:41:19 +00001254\end{verbatim}
1255
1256The \class{deque} double-ended queue type supplied by the
1257\module{collections} module now has a \method{remove(\var{value})}
1258method that removes the first occurrence of \var{value} in the queue,
1259raising \exception{ValueError} if the value isn't found.
1260
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +00001261\item New module: The \module{contextlib} module contains helper functions for use
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +00001262with the new '\keyword{with}' statement. See
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +00001263section~\ref{module-contextlib} for more about this module.
Andrew M. Kuchlingde0a23f2006-04-16 18:45:11 +00001264
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +00001265\item New module: The \module{cProfile} module is a C implementation of
Andrew M. Kuchlingc7095842006-04-14 12:41:19 +00001266the existing \module{profile} module that has much lower overhead.
1267The module's interface is the same as \module{profile}: you run
1268\code{cProfile.run('main()')} to profile a function, can save profile
1269data to a file, etc. It's not yet known if the Hotshot profiler,
1270which is also written in C but doesn't match the \module{profile}
1271module's interface, will continue to be maintained in future versions
1272of Python. (Contributed by Armin Rigo.)
1273
Andrew M. Kuchling0a7ed8c2006-04-24 14:30:47 +00001274Also, the \module{pstats} module for analyzing the data measured by
1275the profiler now supports directing the output to any file object
Andrew M. Kuchlinge78eeb12006-04-21 13:26:42 +00001276by supplying a \var{stream} argument to the \class{Stats} constructor.
1277(Contributed by Skip Montanaro.)
1278
Andrew M. Kuchling952f1962006-04-18 12:38:19 +00001279\item The \module{csv} module, which parses files in
1280comma-separated value format, received several enhancements and a
1281number of bugfixes. You can now set the maximum size in bytes of a
1282field by calling the \method{csv.field_size_limit(\var{new_limit})}
1283function; omitting the \var{new_limit} argument will return the
1284currently-set limit. The \class{reader} class now has a
1285\member{line_num} attribute that counts the number of physical lines
1286read from the source; records can span multiple physical lines, so
1287\member{line_num} is not the same as the number of records read.
1288(Contributed by Skip Montanaro and Andrew McNamara.)
1289
Andrew M. Kuchling67191312006-04-19 12:55:39 +00001290\item The \class{datetime} class in the \module{datetime}
1291module now has a \method{strptime(\var{string}, \var{format})}
1292method for parsing date strings, contributed by Josh Spoerri.
1293It uses the same format characters as \function{time.strptime()} and
1294\function{time.strftime()}:
1295
1296\begin{verbatim}
1297from datetime import datetime
1298
1299ts = datetime.strptime('10:13:15 2006-03-07',
1300 '%H:%M:%S %Y-%m-%d')
1301\end{verbatim}
1302
Andrew M. Kuchlingb33842a2006-04-25 12:31:38 +00001303\item The \module{doctest} module gained a \code{SKIP} option that
1304keeps an example from being executed at all. This is intended for
1305code snippets that are usage examples intended for the reader and
1306aren't actually test cases.
1307
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +00001308\item The \module{fileinput} module was made more flexible.
1309Unicode filenames are now supported, and a \var{mode} parameter that
1310defaults to \code{"r"} was added to the
1311\function{input()} function to allow opening files in binary or
1312universal-newline mode. Another new parameter, \var{openhook},
1313lets you use a function other than \function{open()}
1314to open the input files. Once you're iterating over
1315the set of files, the \class{FileInput} object's new
1316\method{fileno()} returns the file descriptor for the currently opened file.
1317(Contributed by Georg Brandl.)
1318
Andrew M. Kuchlingda376042006-03-17 15:56:41 +00001319\item In the \module{gc} module, the new \function{get_count()} function
1320returns a 3-tuple containing the current collection counts for the
1321three GC generations. This is accounting information for the garbage
1322collector; when these counts reach a specified threshold, a garbage
1323collection sweep will be made. The existing \function{gc.collect()}
1324function now takes an optional \var{generation} argument of 0, 1, or 2
1325to specify which generation to collect.
1326
Andrew M. Kuchlinge9b1bf42005-03-20 19:26:30 +00001327\item The \function{nsmallest()} and
1328\function{nlargest()} functions in the \module{heapq} module
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +00001329now support a \code{key} keyword parameter similar to the one
Andrew M. Kuchlinge9b1bf42005-03-20 19:26:30 +00001330provided by the \function{min()}/\function{max()} functions
1331and the \method{sort()} methods. For example:
1332Example:
1333
1334\begin{verbatim}
1335>>> import heapq
1336>>> L = ["short", 'medium', 'longest', 'longer still']
1337>>> heapq.nsmallest(2, L) # Return two lowest elements, lexicographically
1338['longer still', 'longest']
1339>>> heapq.nsmallest(2, L, key=len) # Return two shortest elements
1340['short', 'medium']
1341\end{verbatim}
1342
1343(Contributed by Raymond Hettinger.)
1344
Andrew M. Kuchling511a3a82005-03-20 19:52:18 +00001345\item The \function{itertools.islice()} function now accepts
1346\code{None} for the start and step arguments. This makes it more
1347compatible with the attributes of slice objects, so that you can now write
1348the following:
1349
1350\begin{verbatim}
1351s = slice(5) # Create slice object
1352itertools.islice(iterable, s.start, s.stop, s.step)
1353\end{verbatim}
1354
1355(Contributed by Raymond Hettinger.)
Andrew M. Kuchling3e41b052005-03-01 00:53:46 +00001356
Andrew M. Kuchlingd4c21772006-04-23 21:51:10 +00001357\item The \module{mailbox} module underwent a massive rewrite to add
1358the capability to modify mailboxes in addition to reading them. A new
1359set of classes that include \class{mbox}, \class{MH}, and
1360\class{Maildir} are used to read mailboxes, and have an
1361\method{add(\var{message})} method to add messages,
1362\method{remove(\var{key})} to remove messages, and
1363\method{lock()}/\method{unlock()} to lock/unlock the mailbox. The
1364following example converts a maildir-format mailbox into an mbox-format one:
1365
1366\begin{verbatim}
1367import mailbox
1368
1369# 'factory=None' uses email.Message.Message as the class representing
1370# individual messages.
1371src = mailbox.Maildir('maildir', factory=None)
1372dest = mailbox.mbox('/tmp/mbox')
1373
1374for msg in src:
1375 dest.add(msg)
1376\end{verbatim}
1377
1378(Contributed by Gregory K. Johnson. Funding was provided by Google's
13792005 Summer of Code.)
1380
Andrew M. Kuchling75ba2442006-04-14 10:29:55 +00001381\item The \module{nis} module now supports accessing domains other
1382than the system default domain by supplying a \var{domain} argument to
1383the \function{nis.match()} and \function{nis.maps()} functions.
1384(Contributed by Ben Bell.)
1385
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001386\item The \module{operator} module's \function{itemgetter()}
1387and \function{attrgetter()} functions now support multiple fields.
1388A call such as \code{operator.attrgetter('a', 'b')}
1389will return a function
1390that retrieves the \member{a} and \member{b} attributes. Combining
1391this new feature with the \method{sort()} method's \code{key} parameter
1392lets you easily sort lists using multiple fields.
Andrew M. Kuchling6e3a66d2006-04-07 12:46:06 +00001393(Contributed by Raymond Hettinger.)
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001394
Andrew M. Kuchlingd4c21772006-04-23 21:51:10 +00001395\item The \module{optparse} module was updated to version 1.5.1 of the
1396Optik library. The \class{OptionParser} class gained an
1397\member{epilog} attribute, a string that will be printed after the
1398help message, and a \method{destroy()} method to break reference
1399cycles created by the object. (Contributed by Greg Ward.)
Andrew M. Kuchling3e41b052005-03-01 00:53:46 +00001400
Andrew M. Kuchling0f1955d2006-04-13 12:09:08 +00001401\item The \module{os} module underwent several changes. The
Andrew M. Kuchlinge9b1bf42005-03-20 19:26:30 +00001402\member{stat_float_times} variable now defaults to true, meaning that
1403\function{os.stat()} will now return time values as floats. (This
1404doesn't necessarily mean that \function{os.stat()} will return times
1405that are precise to fractions of a second; not all systems support
1406such precision.)
Andrew M. Kuchling3e41b052005-03-01 00:53:46 +00001407
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001408Constants named \member{os.SEEK_SET}, \member{os.SEEK_CUR}, and
Andrew M. Kuchlinge9b1bf42005-03-20 19:26:30 +00001409\member{os.SEEK_END} have been added; these are the parameters to the
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001410\function{os.lseek()} function. Two new constants for locking are
1411\member{os.O_SHLOCK} and \member{os.O_EXLOCK}.
1412
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001413Two new functions, \function{wait3()} and \function{wait4()}, were
1414added. They're similar the \function{waitpid()} function which waits
1415for a child process to exit and returns a tuple of the process ID and
1416its exit status, but \function{wait3()} and \function{wait4()} return
1417additional information. \function{wait3()} doesn't take a process ID
1418as input, so it waits for any child process to exit and returns a
14193-tuple of \var{process-id}, \var{exit-status}, \var{resource-usage}
1420as returned from the \function{resource.getrusage()} function.
1421\function{wait4(\var{pid})} does take a process ID.
Andrew M. Kuchling6e3a66d2006-04-07 12:46:06 +00001422(Contributed by Chad J. Schroeder.)
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001423
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001424On FreeBSD, the \function{os.stat()} function now returns
1425times with nanosecond resolution, and the returned object
1426now has \member{st_gen} and \member{st_birthtime}.
1427The \member{st_flags} member is also available, if the platform supports it.
Andrew M. Kuchling6e3a66d2006-04-07 12:46:06 +00001428(Contributed by Antti Louko and Diego Petten\`o.)
1429% (Patch 1180695, 1212117)
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001430
Andrew M. Kuchlingb33842a2006-04-25 12:31:38 +00001431\item The Python debugger provided by the \module{pdb} module
1432can now store lists of commands to execute when a breakpoint is
George Yoshida3bbbc492006-04-25 14:09:58 +00001433reached and execution stops. Once breakpoint \#1 has been created,
Andrew M. Kuchlingb33842a2006-04-25 12:31:38 +00001434enter \samp{commands 1} and enter a series of commands to be executed,
1435finishing the list with \samp{end}. The command list can include
1436commands that resume execution, such as \samp{continue} or
1437\samp{next}. (Contributed by Gr\'egoire Dooms.)
1438% Patch 790710
1439
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +00001440\item The \module{pickle} and \module{cPickle} modules no
1441longer accept a return value of \code{None} from the
1442\method{__reduce__()} method; the method must return a tuple of
1443arguments instead. The ability to return \code{None} was deprecated
1444in Python 2.4, so this completes the removal of the feature.
1445
Andrew M. Kuchlingaa013da2006-04-29 12:10:43 +00001446\item The \module{pkgutil} module, containing various utility
1447functions for finding packages, was enhanced to support PEP 302's
1448import hooks and now also works for packages stored in ZIP-format archives.
1449(Contributed by Phillip J. Eby.)
1450
Andrew M. Kuchlingc9236112006-04-30 01:07:09 +00001451\item The pybench benchmark suite by Marc-Andr\'e~Lemburg is now
1452included in the \file{Tools/pybench} directory. The pybench suite is
1453an improvement on the commonly used \file{pystone.py} program because
1454pybench provides a more detailed measurement of the interpreter's
1455performance. It times particular operations such as function calls,
1456tuple slicing, method lookups, and numeric operations, instead of
1457performing many different operations and reducing the result to a
1458single number as \file{pystone.py} does.
1459
Andrew M. Kuchling01e3d262006-03-17 15:38:39 +00001460\item The old \module{regex} and \module{regsub} modules, which have been
1461deprecated ever since Python 2.0, have finally been deleted.
Andrew M. Kuchlingf4b06602006-03-17 15:39:52 +00001462Other deleted modules: \module{statcache}, \module{tzparse},
1463\module{whrandom}.
Andrew M. Kuchling01e3d262006-03-17 15:38:39 +00001464
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +00001465\item Also deleted: the \file{lib-old} directory,
Andrew M. Kuchling01e3d262006-03-17 15:38:39 +00001466which includes ancient modules such as \module{dircmp} and
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +00001467\module{ni}, was removed. \file{lib-old} wasn't on the default
Andrew M. Kuchling01e3d262006-03-17 15:38:39 +00001468\code{sys.path}, so unless your programs explicitly added the directory to
1469\code{sys.path}, this removal shouldn't affect your code.
1470
Andrew M. Kuchling4678dc82006-01-15 16:11:28 +00001471\item The \module{socket} module now supports \constant{AF_NETLINK}
1472sockets on Linux, thanks to a patch from Philippe Biondi.
1473Netlink sockets are a Linux-specific mechanism for communications
1474between a user-space process and kernel code; an introductory
1475article about them is at \url{http://www.linuxjournal.com/article/7356}.
1476In Python code, netlink addresses are represented as a tuple of 2 integers,
1477\code{(\var{pid}, \var{group_mask})}.
1478
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001479Socket objects also gained accessor methods \method{getfamily()},
1480\method{gettype()}, and \method{getproto()} methods to retrieve the
1481family, type, and protocol values for the socket.
1482
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +00001483\item New module: the \module{spwd} module provides functions for
1484accessing the shadow password database on systems that support
1485shadow passwords.
Fred Drake2db76802004-12-01 05:05:47 +00001486
Andrew M. Kuchling61434b62006-04-13 11:51:07 +00001487\item The Python developers switched from CVS to Subversion during the 2.5
1488development process. Information about the exact build version is
1489available as the \code{sys.subversion} variable, a 3-tuple
1490of \code{(\var{interpreter-name}, \var{branch-name}, \var{revision-range})}.
1491For example, at the time of writing
1492my copy of 2.5 was reporting \code{('CPython', 'trunk', '45313:45315')}.
1493
1494This information is also available to C extensions via the
1495\cfunction{Py_GetBuildInfo()} function that returns a
1496string of build information like this:
1497\code{"trunk:45355:45356M, Apr 13 2006, 07:42:19"}.
1498(Contributed by Barry Warsaw.)
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001499
Andrew M. Kuchlinge9b1bf42005-03-20 19:26:30 +00001500\item The \class{TarFile} class in the \module{tarfile} module now has
Georg Brandl08c02db2005-07-22 18:39:19 +00001501an \method{extractall()} method that extracts all members from the
Andrew M. Kuchlinge9b1bf42005-03-20 19:26:30 +00001502archive into the current working directory. It's also possible to set
1503a different directory as the extraction target, and to unpack only a
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001504subset of the archive's members.
Andrew M. Kuchlinge9b1bf42005-03-20 19:26:30 +00001505
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001506A tarfile's compression can be autodetected by
1507using the mode \code{'r|*'}.
1508% patch 918101
1509(Contributed by Lars Gust\"abel.)
Gregory P. Smithf21a5f72005-08-21 18:45:59 +00001510
Andrew M. Kuchlingf688cc52006-03-10 18:50:08 +00001511\item The \module{unicodedata} module has been updated to use version 4.1.0
1512of the Unicode character database. Version 3.2.0 is required
1513by some specifications, so it's still available as
1514\member{unicodedata.db_3_2_0}.
1515
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +00001516\item The \module{webbrowser} module received a number of
1517enhancements.
1518It's now usable as a script with \code{python -m webbrowser}, taking a
1519URL as the argument; there are a number of switches
1520to control the behaviour (\programopt{-n} for a new browser window,
1521\programopt{-t} for a new tab). New module-level functions,
1522\function{open_new()} and \function{open_new_tab()}, were added
1523to support this. The module's \function{open()} function supports an
1524additional feature, an \var{autoraise} parameter that signals whether
1525to raise the open window when possible. A number of additional
1526browsers were added to the supported list such as Firefox, Opera,
1527Konqueror, and elinks. (Contributed by Oleg Broytmann and George
1528Brandl.)
1529% Patch #754022
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001530
Fredrik Lundh7e0aef02005-12-12 18:54:55 +00001531
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001532\item The \module{xmlrpclib} module now supports returning
1533 \class{datetime} objects for the XML-RPC date type. Supply
1534 \code{use_datetime=True} to the \function{loads()} function
1535 or the \class{Unmarshaller} class to enable this feature.
Andrew M. Kuchling6e3a66d2006-04-07 12:46:06 +00001536 (Contributed by Skip Montanaro.)
1537% Patch 1120353
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001538
Gregory P. Smithf21a5f72005-08-21 18:45:59 +00001539
Fred Drake114b8ca2005-03-21 05:47:11 +00001540\end{itemize}
Andrew M. Kuchlinge9b1bf42005-03-20 19:26:30 +00001541
Fred Drake2db76802004-12-01 05:05:47 +00001542
1543
1544%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001545\subsection{The ctypes package\label{module-ctypes}}
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001546
1547The \module{ctypes} package, written by Thomas Heller, has been added
1548to the standard library. \module{ctypes} lets you call arbitrary functions
Andrew M. Kuchling28c5f1f2006-04-13 02:04:42 +00001549in shared libraries or DLLs. Long-time users may remember the \module{dl} module, which
1550provides functions for loading shared libraries and calling functions in them. The \module{ctypes} package is much fancier.
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001551
Andrew M. Kuchling28c5f1f2006-04-13 02:04:42 +00001552To load a shared library or DLL, you must create an instance of the
1553\class{CDLL} class and provide the name or path of the shared library
1554or DLL. Once that's done, you can call arbitrary functions
1555by accessing them as attributes of the \class{CDLL} object.
1556
1557\begin{verbatim}
1558import ctypes
1559
1560libc = ctypes.CDLL('libc.so.6')
1561result = libc.printf("Line of output\n")
1562\end{verbatim}
1563
1564Type constructors for the various C types are provided: \function{c_int},
1565\function{c_float}, \function{c_double}, \function{c_char_p} (equivalent to \ctype{char *}), and so forth. Unlike Python's types, the C versions are all mutable; you can assign to their \member{value} attribute
1566to change the wrapped value. Python integers and strings will be automatically
1567converted to the corresponding C types, but for other types you
1568must call the correct type constructor. (And I mean \emph{must};
1569getting it wrong will often result in the interpreter crashing
1570with a segmentation fault.)
1571
1572You shouldn't use \function{c_char_p} with a Python string when the C function will be modifying the memory area, because Python strings are
1573supposed to be immutable; breaking this rule will cause puzzling bugs. When you need a modifiable memory area,
Neal Norwitz5f5a69b2006-04-13 03:41:04 +00001574use \function{create_string_buffer()}:
Andrew M. Kuchling28c5f1f2006-04-13 02:04:42 +00001575
1576\begin{verbatim}
1577s = "this is a string"
1578buf = ctypes.create_string_buffer(s)
1579libc.strfry(buf)
1580\end{verbatim}
1581
1582C functions are assumed to return integers, but you can set
1583the \member{restype} attribute of the function object to
1584change this:
1585
1586\begin{verbatim}
1587>>> libc.atof('2.71828')
1588-1783957616
1589>>> libc.atof.restype = ctypes.c_double
1590>>> libc.atof('2.71828')
15912.71828
1592\end{verbatim}
1593
1594\module{ctypes} also provides a wrapper for Python's C API
1595as the \code{ctypes.pythonapi} object. This object does \emph{not}
1596release the global interpreter lock before calling a function, because the lock must be held when calling into the interpreter's code.
1597There's a \class{py_object()} type constructor that will create a
1598\ctype{PyObject *} pointer. A simple usage:
1599
1600\begin{verbatim}
1601import ctypes
1602
1603d = {}
1604ctypes.pythonapi.PyObject_SetItem(ctypes.py_object(d),
1605 ctypes.py_object("abc"), ctypes.py_object(1))
1606# d is now {'abc', 1}.
1607\end{verbatim}
1608
1609Don't forget to use \class{py_object()}; if it's omitted you end
1610up with a segmentation fault.
1611
1612\module{ctypes} has been around for a while, but people still write
1613and distribution hand-coded extension modules because you can't rely on \module{ctypes} being present.
1614Perhaps developers will begin to write
1615Python wrappers atop a library accessed through \module{ctypes} instead
1616of extension modules, now that \module{ctypes} is included with core Python.
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001617
Andrew M. Kuchling28c5f1f2006-04-13 02:04:42 +00001618\begin{seealso}
1619
1620\seeurl{http://starship.python.net/crew/theller/ctypes/}
1621{The ctypes web page, with a tutorial, reference, and FAQ.}
1622
1623\end{seealso}
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001624
Andrew M. Kuchling61434b62006-04-13 11:51:07 +00001625
1626%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001627\subsection{The ElementTree package\label{module-etree}}
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001628
1629A subset of Fredrik Lundh's ElementTree library for processing XML has
Andrew M. Kuchling16ed5212006-04-10 22:28:11 +00001630been added to the standard library as \module{xmlcore.etree}. The
Georg Brandlce27a062006-04-11 06:27:12 +00001631available modules are
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001632\module{ElementTree}, \module{ElementPath}, and
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +00001633\module{ElementInclude} from ElementTree 1.2.6.
1634The \module{cElementTree} accelerator module is also included.
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001635
Andrew M. Kuchling16ed5212006-04-10 22:28:11 +00001636The rest of this section will provide a brief overview of using
1637ElementTree. Full documentation for ElementTree is available at
1638\url{http://effbot.org/zone/element-index.htm}.
1639
1640ElementTree represents an XML document as a tree of element nodes.
1641The text content of the document is stored as the \member{.text}
1642and \member{.tail} attributes of
1643(This is one of the major differences between ElementTree and
1644the Document Object Model; in the DOM there are many different
1645types of node, including \class{TextNode}.)
1646
1647The most commonly used parsing function is \function{parse()}, that
1648takes either a string (assumed to contain a filename) or a file-like
1649object and returns an \class{ElementTree} instance:
1650
1651\begin{verbatim}
1652from xmlcore.etree import ElementTree as ET
1653
1654tree = ET.parse('ex-1.xml')
1655
1656feed = urllib.urlopen(
1657 'http://planet.python.org/rss10.xml')
1658tree = ET.parse(feed)
1659\end{verbatim}
1660
1661Once you have an \class{ElementTree} instance, you
1662can call its \method{getroot()} method to get the root \class{Element} node.
1663
1664There's also an \function{XML()} function that takes a string literal
1665and returns an \class{Element} node (not an \class{ElementTree}).
1666This function provides a tidy way to incorporate XML fragments,
1667approaching the convenience of an XML literal:
1668
1669\begin{verbatim}
1670svg = et.XML("""<svg width="10px" version="1.0">
1671 </svg>""")
1672svg.set('height', '320px')
1673svg.append(elem1)
1674\end{verbatim}
1675
1676Each XML element supports some dictionary-like and some list-like
Andrew M. Kuchling075e0232006-04-11 13:14:56 +00001677access methods. Dictionary-like operations are used to access attribute
1678values, and list-like operations are used to access child nodes.
Andrew M. Kuchling16ed5212006-04-10 22:28:11 +00001679
Andrew M. Kuchling075e0232006-04-11 13:14:56 +00001680\begin{tableii}{c|l}{code}{Operation}{Result}
1681 \lineii{elem[n]}{Returns n'th child element.}
1682 \lineii{elem[m:n]}{Returns list of m'th through n'th child elements.}
1683 \lineii{len(elem)}{Returns number of child elements.}
1684 \lineii{elem.getchildren()}{Returns list of child elements.}
1685 \lineii{elem.append(elem2)}{Adds \var{elem2} as a child.}
1686 \lineii{elem.insert(index, elem2)}{Inserts \var{elem2} at the specified location.}
1687 \lineii{del elem[n]}{Deletes n'th child element.}
1688 \lineii{elem.keys()}{Returns list of attribute names.}
1689 \lineii{elem.get(name)}{Returns value of attribute \var{name}.}
1690 \lineii{elem.set(name, value)}{Sets new value for attribute \var{name}.}
1691 \lineii{elem.attrib}{Retrieves the dictionary containing attributes.}
1692 \lineii{del elem.attrib[name]}{Deletes attribute \var{name}.}
1693\end{tableii}
1694
1695Comments and processing instructions are also represented as
1696\class{Element} nodes. To check if a node is a comment or processing
1697instructions:
1698
1699\begin{verbatim}
1700if elem.tag is ET.Comment:
1701 ...
1702elif elem.tag is ET.ProcessingInstruction:
1703 ...
1704\end{verbatim}
Andrew M. Kuchling16ed5212006-04-10 22:28:11 +00001705
1706To generate XML output, you should call the
1707\method{ElementTree.write()} method. Like \function{parse()},
1708it can take either a string or a file-like object:
1709
1710\begin{verbatim}
1711# Encoding is US-ASCII
1712tree.write('output.xml')
1713
1714# Encoding is UTF-8
1715f = open('output.xml', 'w')
1716tree.write(f, 'utf-8')
1717\end{verbatim}
1718
1719(Caution: the default encoding used for output is ASCII, which isn't
1720very useful for general XML work, raising an exception if there are
1721any characters with values greater than 127. You should always
1722specify a different encoding such as UTF-8 that can handle any Unicode
1723character.)
1724
Andrew M. Kuchling075e0232006-04-11 13:14:56 +00001725This section is only a partial description of the ElementTree interfaces.
1726Please read the package's official documentation for more details.
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001727
Andrew M. Kuchling16ed5212006-04-10 22:28:11 +00001728\begin{seealso}
1729
1730\seeurl{http://effbot.org/zone/element-index.htm}
1731{Official documentation for ElementTree.}
1732
1733
1734\end{seealso}
1735
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001736
Andrew M. Kuchling61434b62006-04-13 11:51:07 +00001737%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001738\subsection{The hashlib package\label{module-hashlib}}
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001739
Andrew M. Kuchling29b3d082006-04-14 20:35:17 +00001740A new \module{hashlib} module, written by Gregory P. Smith,
1741has been added to replace the
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001742\module{md5} and \module{sha} modules. \module{hashlib} adds support
1743for additional secure hashes (SHA-224, SHA-256, SHA-384, and SHA-512).
1744When available, the module uses OpenSSL for fast platform optimized
1745implementations of algorithms.
1746
1747The old \module{md5} and \module{sha} modules still exist as wrappers
1748around hashlib to preserve backwards compatibility. The new module's
1749interface is very close to that of the old modules, but not identical.
1750The most significant difference is that the constructor functions
1751for creating new hashing objects are named differently.
1752
1753\begin{verbatim}
1754# Old versions
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +00001755h = md5.md5()
1756h = md5.new()
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001757
1758# New version
1759h = hashlib.md5()
1760
1761# Old versions
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +00001762h = sha.sha()
1763h = sha.new()
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001764
1765# New version
1766h = hashlib.sha1()
1767
1768# Hash that weren't previously available
1769h = hashlib.sha224()
1770h = hashlib.sha256()
1771h = hashlib.sha384()
1772h = hashlib.sha512()
1773
1774# Alternative form
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +00001775h = hashlib.new('md5') # Provide algorithm as a string
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001776\end{verbatim}
1777
1778Once a hash object has been created, its methods are the same as before:
1779\method{update(\var{string})} hashes the specified string into the
1780current digest state, \method{digest()} and \method{hexdigest()}
1781return the digest value as a binary string or a string of hex digits,
1782and \method{copy()} returns a new hashing object with the same digest state.
1783
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001784
Andrew M. Kuchling61434b62006-04-13 11:51:07 +00001785%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001786\subsection{The sqlite3 package\label{module-sqlite}}
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001787
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001788The pysqlite module (\url{http://www.pysqlite.org}), a wrapper for the
1789SQLite embedded database, has been added to the standard library under
Andrew M. Kuchling29b3d082006-04-14 20:35:17 +00001790the package name \module{sqlite3}.
1791
1792SQLite is a C library that provides a SQL-language database that
1793stores data in disk files without requiring a separate server process.
1794pysqlite was written by Gerhard H\"aring and provides a SQL interface
1795compliant with the DB-API 2.0 specification described by
1796\pep{249}. This means that it should be possible to write the first
1797version of your applications using SQLite for data storage. If
1798switching to a larger database such as PostgreSQL or Oracle is
1799later necessary, the switch should be relatively easy.
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001800
1801If you're compiling the Python source yourself, note that the source
Andrew M. Kuchling29b3d082006-04-14 20:35:17 +00001802tree doesn't include the SQLite code, only the wrapper module.
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001803You'll need to have the SQLite libraries and headers installed before
1804compiling Python, and the build process will compile the module when
1805the necessary headers are available.
1806
Andrew M. Kuchlingd58baf82006-04-10 21:40:16 +00001807To use the module, you must first create a \class{Connection} object
1808that represents the database. Here the data will be stored in the
1809\file{/tmp/example} file:
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001810
Andrew M. Kuchlingd58baf82006-04-10 21:40:16 +00001811\begin{verbatim}
1812conn = sqlite3.connect('/tmp/example')
1813\end{verbatim}
1814
1815You can also supply the special name \samp{:memory:} to create
1816a database in RAM.
1817
1818Once you have a \class{Connection}, you can create a \class{Cursor}
1819object and call its \method{execute()} method to perform SQL commands:
1820
1821\begin{verbatim}
1822c = conn.cursor()
1823
1824# Create table
1825c.execute('''create table stocks
1826(date timestamp, trans varchar, symbol varchar,
1827 qty decimal, price decimal)''')
1828
1829# Insert a row of data
1830c.execute("""insert into stocks
Andrew M. Kuchling29b3d082006-04-14 20:35:17 +00001831 values ('2006-01-05','BUY','RHAT',100,35.14)""")
Andrew M. Kuchlingd58baf82006-04-10 21:40:16 +00001832\end{verbatim}
1833
Andrew M. Kuchling29b3d082006-04-14 20:35:17 +00001834Usually your SQL operations will need to use values from Python
Andrew M. Kuchlingd58baf82006-04-10 21:40:16 +00001835variables. You shouldn't assemble your query using Python's string
1836operations because doing so is insecure; it makes your program
Andrew M. Kuchling29b3d082006-04-14 20:35:17 +00001837vulnerable to an SQL injection attack.
1838
1839Instead, use SQLite's parameter substitution. Put \samp{?} as a
1840placeholder wherever you want to use a value, and then provide a tuple
1841of values as the second argument to the cursor's \method{execute()}
1842method. For example:
Andrew M. Kuchlingd58baf82006-04-10 21:40:16 +00001843
1844\begin{verbatim}
1845# Never do this -- insecure!
1846symbol = 'IBM'
1847c.execute("... where symbol = '%s'" % symbol)
1848
1849# Do this instead
1850t = (symbol,)
Andrew M. Kuchling7e5abb92006-04-26 12:21:06 +00001851c.execute('select * from stocks where symbol=?', t)
Andrew M. Kuchlingd58baf82006-04-10 21:40:16 +00001852
1853# Larger example
1854for t in (('2006-03-28', 'BUY', 'IBM', 1000, 45.00),
Andrew M. Kuchlingd058d002006-04-16 18:20:05 +00001855 ('2006-04-05', 'BUY', 'MSOFT', 1000, 72.00),
1856 ('2006-04-06', 'SELL', 'IBM', 500, 53.00),
1857 ):
Andrew M. Kuchlingd58baf82006-04-10 21:40:16 +00001858 c.execute('insert into stocks values (?,?,?,?,?)', t)
1859\end{verbatim}
1860
1861To retrieve data after executing a SELECT statement, you can either
1862treat the cursor as an iterator, call the cursor's \method{fetchone()}
1863method to retrieve a single matching row,
1864or call \method{fetchall()} to get a list of the matching rows.
1865
1866This example uses the iterator form:
1867
1868\begin{verbatim}
1869>>> c = conn.cursor()
1870>>> c.execute('select * from stocks order by price')
1871>>> for row in c:
1872... print row
1873...
1874(u'2006-01-05', u'BUY', u'RHAT', 100, 35.140000000000001)
1875(u'2006-03-28', u'BUY', u'IBM', 1000, 45.0)
1876(u'2006-04-06', u'SELL', u'IBM', 500, 53.0)
1877(u'2006-04-05', u'BUY', u'MSOFT', 1000, 72.0)
1878>>>
1879\end{verbatim}
1880
Andrew M. Kuchlingd58baf82006-04-10 21:40:16 +00001881For more information about the SQL dialect supported by SQLite, see
1882\url{http://www.sqlite.org}.
1883
1884\begin{seealso}
1885
1886\seeurl{http://www.pysqlite.org}
1887{The pysqlite web page.}
1888
1889\seeurl{http://www.sqlite.org}
1890{The SQLite web page; the documentation describes the syntax and the
1891available data types for the supported SQL dialect.}
1892
1893\seepep{249}{Database API Specification 2.0}{PEP written by
1894Marc-Andr\'e Lemburg.}
1895
1896\end{seealso}
Andrew M. Kuchlingaf7ee992006-04-03 12:41:37 +00001897
Fred Drake2db76802004-12-01 05:05:47 +00001898
1899% ======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001900\section{Build and C API Changes\label{build-api}}
Fred Drake2db76802004-12-01 05:05:47 +00001901
1902Changes to Python's build process and to the C API include:
1903
1904\begin{itemize}
1905
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +00001906\item The largest change to the C API came from \pep{353},
1907which modifies the interpreter to use a \ctype{Py_ssize_t} type
1908definition instead of \ctype{int}. See the earlier
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +00001909section~\ref{pep-353} for a discussion of this change.
Andrew M. Kuchling4d8cd892006-04-06 13:03:04 +00001910
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00001911\item The design of the bytecode compiler has changed a great deal, to
1912no longer generate bytecode by traversing the parse tree. Instead
Andrew M. Kuchlingdb85ed52005-10-23 21:52:59 +00001913the parse tree is converted to an abstract syntax tree (or AST), and it is
1914the abstract syntax tree that's traversed to produce the bytecode.
1915
Andrew M. Kuchling4e861952006-04-12 12:16:31 +00001916It's possible for Python code to obtain AST objects by using the
Andrew M. Kuchling5f445bf2006-04-12 18:54:00 +00001917\function{compile()} built-in and specifying \code{_ast.PyCF_ONLY_AST}
1918as the value of the
Andrew M. Kuchling4e861952006-04-12 12:16:31 +00001919\var{flags} parameter:
1920
1921\begin{verbatim}
Andrew M. Kuchling5f445bf2006-04-12 18:54:00 +00001922from _ast import PyCF_ONLY_AST
Andrew M. Kuchling4e861952006-04-12 12:16:31 +00001923ast = compile("""a=0
1924for i in range(10):
1925 a += i
Andrew M. Kuchling5f445bf2006-04-12 18:54:00 +00001926""", "<string>", 'exec', PyCF_ONLY_AST)
Andrew M. Kuchling4e861952006-04-12 12:16:31 +00001927
1928assignment = ast.body[0]
1929for_loop = ast.body[1]
1930\end{verbatim}
1931
Andrew M. Kuchlingdb85ed52005-10-23 21:52:59 +00001932No documentation has been written for the AST code yet. To start
1933learning about it, read the definition of the various AST nodes in
1934\file{Parser/Python.asdl}. A Python script reads this file and
1935generates a set of C structure definitions in
1936\file{Include/Python-ast.h}. The \cfunction{PyParser_ASTFromString()}
1937and \cfunction{PyParser_ASTFromFile()}, defined in
1938\file{Include/pythonrun.h}, take Python source as input and return the
1939root of an AST representing the contents. This AST can then be turned
1940into a code object by \cfunction{PyAST_Compile()}. For more
1941information, read the source code, and then ask questions on
1942python-dev.
1943
1944% List of names taken from Jeremy's python-dev post at
1945% http://mail.python.org/pipermail/python-dev/2005-October/057500.html
1946The AST code was developed under Jeremy Hylton's management, and
1947implemented by (in alphabetical order) Brett Cannon, Nick Coghlan,
1948Grant Edwards, John Ehresman, Kurt Kaiser, Neal Norwitz, Tim Peters,
1949Armin Rigo, and Neil Schemenauer, plus the participants in a number of
1950AST sprints at conferences such as PyCon.
1951
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001952\item The built-in set types now have an official C API. Call
1953\cfunction{PySet_New()} and \cfunction{PyFrozenSet_New()} to create a
1954new set, \cfunction{PySet_Add()} and \cfunction{PySet_Discard()} to
1955add and remove elements, and \cfunction{PySet_Contains} and
1956\cfunction{PySet_Size} to examine the set's state.
Andrew M. Kuchling29b3d082006-04-14 20:35:17 +00001957(Contributed by Raymond Hettinger.)
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001958
Andrew M. Kuchling61434b62006-04-13 11:51:07 +00001959\item C code can now obtain information about the exact revision
1960of the Python interpreter by calling the
1961\cfunction{Py_GetBuildInfo()} function that returns a
1962string of build information like this:
1963\code{"trunk:45355:45356M, Apr 13 2006, 07:42:19"}.
1964(Contributed by Barry Warsaw.)
1965
Andrew M. Kuchling29b3d082006-04-14 20:35:17 +00001966\item The CPython interpreter is still written in C, but
1967the code can now be compiled with a {\Cpp} compiler without errors.
1968(Implemented by Anthony Baxter, Martin von~L\"owis, Skip Montanaro.)
1969
Andrew M. Kuchling150e3492005-08-23 00:56:06 +00001970\item The \cfunction{PyRange_New()} function was removed. It was
1971never documented, never used in the core code, and had dangerously lax
1972error checking.
Fred Drake2db76802004-12-01 05:05:47 +00001973
1974\end{itemize}
1975
1976
1977%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00001978\subsection{Port-Specific Changes\label{ports}}
Fred Drake2db76802004-12-01 05:05:47 +00001979
Andrew M. Kuchling6fc69762006-04-13 12:37:21 +00001980\begin{itemize}
1981
1982\item MacOS X (10.3 and higher): dynamic loading of modules
1983now uses the \cfunction{dlopen()} function instead of MacOS-specific
1984functions.
1985
Andrew M. Kuchlingb37bcb52006-04-29 11:53:15 +00001986\item MacOS X: a \longprogramopt{enable-universalsdk} switch was added
1987to the \program{configure} script that compiles the interpreter as a
1988universal binary able to run on both PowerPC and Intel processors.
1989(Contributed by Ronald Oussoren.)
1990
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +00001991\item Windows: \file{.dll} is no longer supported as a filename extension for
1992extension modules. \file{.pyd} is now the only filename extension that will
1993be searched for.
1994
Andrew M. Kuchling6fc69762006-04-13 12:37:21 +00001995\end{itemize}
Fred Drake2db76802004-12-01 05:05:47 +00001996
1997
1998%======================================================================
1999\section{Other Changes and Fixes \label{section-other}}
2000
2001As usual, there were a bunch of other improvements and bugfixes
Andrew M. Kuchlingf688cc52006-03-10 18:50:08 +00002002scattered throughout the source tree. A search through the SVN change
Fred Drake2db76802004-12-01 05:05:47 +00002003logs finds there were XXX patches applied and YYY bugs fixed between
Andrew M. Kuchling92e24952004-12-03 13:54:09 +00002004Python 2.4 and 2.5. Both figures are likely to be underestimates.
Fred Drake2db76802004-12-01 05:05:47 +00002005
2006Some of the more notable changes are:
2007
2008\begin{itemize}
2009
Andrew M. Kuchling01e3d262006-03-17 15:38:39 +00002010\item Evan Jones's patch to obmalloc, first described in a talk
2011at PyCon DC 2005, was applied. Python 2.4 allocated small objects in
2012256K-sized arenas, but never freed arenas. With this patch, Python
2013will free arenas when they're empty. The net effect is that on some
2014platforms, when you allocate many objects, Python's memory usage may
2015actually drop when you delete them, and the memory may be returned to
2016the operating system. (Implemented by Evan Jones, and reworked by Tim
2017Peters.)
Fred Drake2db76802004-12-01 05:05:47 +00002018
Andrew M. Kuchlingf7c62902006-04-12 12:27:50 +00002019Note that this change means extension modules need to be more careful
Andrew M. Kuchling0f1955d2006-04-13 12:09:08 +00002020with how they allocate memory. Python's API has many different
Andrew M. Kuchlingf7c62902006-04-12 12:27:50 +00002021functions for allocating memory that are grouped into families. For
2022example, \cfunction{PyMem_Malloc()}, \cfunction{PyMem_Realloc()}, and
2023\cfunction{PyMem_Free()} are one family that allocates raw memory,
2024while \cfunction{PyObject_Malloc()}, \cfunction{PyObject_Realloc()},
2025and \cfunction{PyObject_Free()} are another family that's supposed to
2026be used for creating Python objects.
2027
2028Previously these different families all reduced to the platform's
2029\cfunction{malloc()} and \cfunction{free()} functions. This meant
2030it didn't matter if you got things wrong and allocated memory with the
2031\cfunction{PyMem} function but freed it with the \cfunction{PyObject}
2032function. With the obmalloc change, these families now do different
2033things, and mismatches will probably result in a segfault. You should
2034carefully test your C extension modules with Python 2.5.
2035
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00002036\item Coverity, a company that markets a source code analysis tool
2037 called Prevent, provided the results of their examination of the Python
Andrew M. Kuchling0f1955d2006-04-13 12:09:08 +00002038 source code. The analysis found about 60 bugs that
2039 were quickly fixed. Many of the bugs were refcounting problems, often
2040 occurring in error-handling code. See
2041 \url{http://scan.coverity.com} for the statistics.
Andrew M. Kuchling38f85072006-04-02 01:46:32 +00002042
Fred Drake2db76802004-12-01 05:05:47 +00002043\end{itemize}
2044
2045
2046%======================================================================
Andrew M. Kuchling98189242006-04-26 12:23:39 +00002047\section{Porting to Python 2.5\label{porting}}
Fred Drake2db76802004-12-01 05:05:47 +00002048
2049This section lists previously described changes that may require
2050changes to your code:
2051
2052\begin{itemize}
2053
Andrew M. Kuchling5f445bf2006-04-12 18:54:00 +00002054\item ASCII is now the default encoding for modules. It's now
2055a syntax error if a module contains string literals with 8-bit
2056characters but doesn't have an encoding declaration. In Python 2.4
2057this triggered a warning, not a syntax error.
2058
Andrew M. Kuchling3b4fb042006-04-13 12:49:39 +00002059\item Previously, the \member{gi_frame} attribute of a generator
2060was always a frame object. Because of the \pep{342} changes
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +00002061described in section~\ref{pep-342}, it's now possible
Andrew M. Kuchling3b4fb042006-04-13 12:49:39 +00002062for \member{gi_frame} to be \code{None}.
2063
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +00002064
2065\item Library: The \module{pickle} and \module{cPickle} modules no
2066longer accept a return value of \code{None} from the
2067\method{__reduce__()} method; the method must return a tuple of
2068arguments instead. The modules also no longer accept the deprecated
2069\var{bin} keyword parameter.
2070
Andrew M. Kuchlingf7c62902006-04-12 12:27:50 +00002071\item C API: Many functions now use \ctype{Py_ssize_t}
Andrew M. Kuchling42c6e2f2006-04-21 13:01:45 +00002072instead of \ctype{int} to allow processing more data on 64-bit
2073machines. Extension code may need to make the same change to avoid
2074warnings and to support 64-bit machines. See the earlier
Andrew M. Kuchlingfb08e732006-04-21 13:08:02 +00002075section~\ref{pep-353} for a discussion of this change.
Andrew M. Kuchlingf7c62902006-04-12 12:27:50 +00002076
2077\item C API:
2078The obmalloc changes mean that
2079you must be careful to not mix usage
2080of the \cfunction{PyMem_*()} and \cfunction{PyObject_*()}
2081families of functions. Memory allocated with
2082one family's \cfunction{*_Malloc()} must be
2083freed with the corresponding family's \cfunction{*_Free()} function.
2084
Fred Drake2db76802004-12-01 05:05:47 +00002085\end{itemize}
2086
2087
2088%======================================================================
2089\section{Acknowledgements \label{acks}}
2090
2091The author would like to thank the following people for offering
2092suggestions, corrections and assistance with various drafts of this
Andrew M. Kuchling63fe9b52006-04-20 13:36:06 +00002093article: Phillip J. Eby, Kent Johnson, Martin von~L\"owis, Gustavo
Andrew M. Kuchling7e5abb92006-04-26 12:21:06 +00002094Niemeyer, James Pryor, Mike Rovner, Thomas Wouters.
Fred Drake2db76802004-12-01 05:05:47 +00002095
2096\end{document}