PEP 308: Conditional Expressions

For a long time, people have been requesting a way to write conditional expressions, which are expressions that return value A or value B depending on whether a Boolean value is true or false. A conditional expression lets you write a single assignment statement that has the same effect as the following:

if condition:
    x = true_value
else:
    x = false_value

There have been endless tedious discussions of syntax on both python-dev and comp.lang.python. A vote was even held that found the majority of voters wanted conditional expressions in some form, but there was no syntax that was preferred by a clear majority. Candidates included C’s cond ? true_v : false_v, if cond then true_v else false_v, and 16 other variations.

Guido van Rossum eventually chose a surprising syntax:

x = true_value if condition else false_value

Evaluation is still lazy as in existing Boolean expressions, so the order of evaluation jumps around a bit. The condition expression in the middle is evaluated first, and the true_value expression is evaluated only if the condition was true. Similarly, the false_value expression is only evaluated when the condition is false.

This syntax may seem strange and backwards; why does the condition go in the middle of the expression, and not in the front as in C’s c ? x : y? The decision was checked by applying the new syntax to the modules in the standard library and seeing how the resulting code read. In many cases where a conditional expression is used, one value seems to be the ‘common case’ and one value is an ‘exceptional case’, used only on rarer occasions when the condition isn’t met. The conditional syntax makes this pattern a bit more obvious:

contents = ((doc + '\n') if doc else '')

I read the above statement as meaning “here contents is usually assigned a value of doc+'\n'; sometimes doc is empty, in which special case an empty string is returned.” I doubt I will use conditional expressions very often where there isn’t a clear common and uncommon case.

There was some discussion of whether the language should require surrounding conditional expressions with parentheses. The decision was made to not require parentheses in the Python language’s grammar, but as a matter of style I think you should always use them. Consider these two statements:

# First version -- no parens
level = 1 if logging else 0

# Second version -- with parens
level = (1 if logging else 0)

In the first version, I think a reader’s eye might group the statement into ‘level = 1’, ‘if logging’, ‘else 0’, and think that the condition decides whether the assignment to level is performed. The second version reads better, in my opinion, because it makes it clear that the assignment is always performed and the choice is being made between two values.

Another reason for including the brackets: a few odd combinations of list comprehensions and lambdas could look like incorrect conditional expressions. See PEP 308 for some examples. If you put parentheses around your conditional expressions, you won’t run into this case.

PEP 309: Partial Function Application

The functools module is intended to contain tools for functional-style programming.

One useful tool in this module is the partial() function. For programs written in a functional style, you’ll sometimes want to construct variants of existing functions that have some of the parameters filled in. Consider a Python function f(a, b, c); you could create a new function g(b, c) that was equivalent to f(1, b, c). This is called “partial function application”.

partial() takes the arguments (function, arg1, arg2, ... kwarg1=value1, kwarg2=value2). The resulting object is callable, so you can just call it to invoke function with the filled-in arguments.

Here’s a small but realistic example:

import functools

def log (message, subsystem):
    "Write the contents of 'message' to the specified subsystem."
    print '%s: %s' % (subsystem, message)
    ...

server_log = functools.partial(log, subsystem='server')
server_log('Unable to open socket')

Here’s another example, from a program that uses PyGTK. Here a context- sensitive pop-up menu is being constructed dynamically. The callback provided for the menu option is a partially applied version of the open_item() method, where the first argument has been provided.

...
class Application:
    def open_item(self, path):
       ...
    def init (self):
        open_func = functools.partial(self.open_item, item_path)
        popup_menu.append( ("Open", open_func, 1) )

Another function in the functools module is the update_wrapper(wrapper, wrapped)() function that helps you write well- behaved decorators. update_wrapper() copies the name, module, and docstring attribute to a wrapper function so that tracebacks inside the wrapped function are easier to understand. For example, you might write:

def my_decorator(f):
    def wrapper(*args, **kwds):
        print 'Calling decorated function'
        return f(*args, **kwds)
    functools.update_wrapper(wrapper, f)
    return wrapper

wraps() is a decorator that can be used inside your own decorators to copy the wrapped function’s information. An alternate version of the previous example would be:

def my_decorator(f):
    @functools.wraps(f)
    def wrapper(*args, **kwds):
        print 'Calling decorated function'
        return f(*args, **kwds)
    return wrapper

PEP 314: Metadata for Python Software Packages v1.1

Some simple dependency support was added to Distutils. The setup() function now has requires, provides, and obsoletes keyword parameters. When you build a source distribution using the sdist command, the dependency information will be recorded in the PKG-INFO file.

Another new keyword parameter is download_url, which should be set to a URL for the package’s source code. This means it’s now possible to look up an entry in the package index, determine the dependencies for a package, and download the required packages.

VERSION = '1.0'
setup(name='PyPackage',
      version=VERSION,
      requires=['numarray', 'zlib (>=1.1.4)'],
      obsoletes=['OldPackage']
      download_url=('http://www.example.com/pypackage/dist/pkg-%s.tar.gz'
                    % VERSION),
     )

Another new enhancement to the Python package index at http://cheeseshop.python.org is storing source and binary archives for a package. The new upload Distutils command will upload a package to the repository.

Before a package can be uploaded, you must be able to build a distribution using the sdist Distutils command. Once that works, you can run python setup.py upload to add your package to the PyPI archive. Optionally you can GPG-sign the package by supplying the --sign and --identity options.

Package uploading was implemented by Martin von Löwis and Richard Jones.

PEP 328: Absolute and Relative Imports

The simpler part of PEP 328 was implemented in Python 2.4: parentheses could now be used to enclose the names imported from a module using the from ... import ... statement, making it easier to import many different names.

The more complicated part has been implemented in Python 2.5: importing a module can be specified to use absolute or package-relative imports. The plan is to move toward making absolute imports the default in future versions of Python.

Let’s say you have a package directory like this:

pkg/
pkg/__init__.py
pkg/main.py
pkg/string.py

This defines a package named pkg containing the pkg.main and pkg.string submodules.

Consider the code in the main.py module. What happens if it executes the statement import string? In Python 2.4 and earlier, it will first look in the package’s directory to perform a relative import, finds pkg/string.py, imports the contents of that file as the pkg.string module, and that module is bound to the name string in the pkg.main module’s namespace.

That’s fine if pkg.string was what you wanted. But what if you wanted Python’s standard string module? There’s no clean way to ignore pkg.string and look for the standard module; generally you had to look at the contents of sys.modules, which is slightly unclean. Holger Krekel’s py.std package provides a tidier way to perform imports from the standard library, import py ; py.std.string.join(), but that package isn’t available on all Python installations.

Reading code which relies on relative imports is also less clear, because a reader may be confused about which module, string or pkg.string, is intended to be used. Python users soon learned not to duplicate the names of standard library modules in the names of their packages’ submodules, but you can’t protect against having your submodule’s name being used for a new module added in a future version of Python.

In Python 2.5, you can switch import‘s behaviour to absolute imports using a from __future__ import absolute_import directive. This absolute- import behaviour will become the default in a future version (probably Python 2.7). Once absolute imports are the default, import string will always find the standard library’s version. It’s suggested that users should begin using absolute imports as much as possible, so it’s preferable to begin writing from pkg import string in your code.

Relative imports are still possible by adding a leading period to the module name when using the from ... import form:

# Import names from pkg.string
from .string import name1, name2
# Import pkg.string
from . import string

This imports the string module relative to the current package, so in pkg.main this will import name1 and name2 from pkg.string. Additional leading periods perform the relative import starting from the parent of the current package. For example, code in the A.B.C module can do:

from . import D                 # Imports A.B.D
from .. import E                # Imports A.E
from ..F import G               # Imports A.F.G

Leading periods cannot be used with the import modname form of the import statement, only the from ... import form.

PEP 338: Executing Modules as Scripts

The -m switch added in Python 2.4 to execute a module as a script gained a few more abilities. Instead of being implemented in C code inside the Python interpreter, the switch now uses an implementation in a new module, runpy.

The runpy module implements a more sophisticated import mechanism so that it’s now possible to run modules in a package such as pychecker.checker. The module also supports alternative import mechanisms such as the zipimport module. This means you can add a .zip archive’s path to sys.path and then use the -m switch to execute code from the archive.

PEP 341: Unified try/except/finally

Until Python 2.5, the try statement came in two flavours. You could use a finally block to ensure that code is always executed, or one or more except blocks to catch specific exceptions. You couldn’t combine both except blocks and a finally block, because generating the right bytecode for the combined version was complicated and it wasn’t clear what the semantics of the combined statement should be.

Guido van Rossum spent some time working with Java, which does support the equivalent of combining except blocks and a finally block, and this clarified what the statement should mean. In Python 2.5, you can now write:

try:
    block-1 ...
except Exception1:
    handler-1 ...
except Exception2:
    handler-2 ...
else:
    else-block
finally:
    final-block

The code in block-1 is executed. If the code raises an exception, the various except blocks are tested: if the exception is of class Exception1, handler-1 is executed; otherwise if it’s of class Exception2, handler-2 is executed, and so forth. If no exception is raised, the else-block is executed.

No matter what happened previously, the final-block is executed once the code block is complete and any raised exceptions handled. Even if there’s an error in an exception handler or the else-block and a new exception is raised, the code in the final-block is still run.

PEP 342: New Generator Features

Python 2.5 adds a simple way to pass values into a generator. As introduced in Python 2.3, generators only produce output; once a generator’s code was invoked to create an iterator, there was no way to pass any new information into the function when its execution is resumed. Sometimes the ability to pass in some information would be useful. Hackish solutions to this include making the generator’s code look at a global variable and then changing the global variable’s value, or passing in some mutable object that callers then modify.

To refresh your memory of basic generators, here’s a simple example:

def counter (maximum):
    i = 0
    while i < maximum:
        yield i
        i += 1

When you call counter(10), the result is an iterator that returns the values from 0 up to 9. On encountering the yield statement, the iterator returns the provided value and suspends the function’s execution, preserving the local variables. Execution resumes on the following call to the iterator’s next() method, picking up after the yield statement.

In Python 2.3, yield was a statement; it didn’t return any value. In 2.5, yield is now an expression, returning a value that can be assigned to a variable or otherwise operated on:

val = (yield i)

I recommend that you always put parentheses around a yield expression when you’re doing something with the returned value, as in the above example. The parentheses aren’t always necessary, but it’s easier to always add them instead of having to remember when they’re needed.

(PEP 342 explains the exact rules, which are that a yield-expression must always be parenthesized except when it occurs at the top-level expression on the right-hand side of an assignment. This means you can write val = yield i but have to use parentheses when there’s an operation, as in val = (yield i) + 12.)

Values are sent into a generator by calling its send(value)() method. The generator’s code is then resumed and the yield expression returns the specified value. If the regular next() method is called, the yield returns None.

Here’s the previous example, modified to allow changing the value of the internal counter.

def counter (maximum):
    i = 0
    while i < maximum:
        val = (yield i)
        # If value provided, change counter
        if val is not None:
            i = val
        else:
            i += 1

And here’s an example of changing the counter:

>>> it = counter(10)
>>> print it.next()
0
>>> print it.next()
1
>>> print it.send(8)
8
>>> print it.next()
9
>>> print it.next()
Traceback (most recent call last):
  File "t.py", line 15, in ?
    print it.next()
StopIteration

yield will usually return None, so you should always check for this case. Don’t just use its value in expressions unless you’re sure that the send() method will be the only method used to resume your generator function.

In addition to send(), there are two other new methods on generators:

• throw(type, value=None, traceback=None)() is used to raise an exception inside the generator; the exception is raised by the yield expression where the generator’s execution is paused.

• close() raises a new GeneratorExit exception inside the generator to terminate the iteration. On receiving this exception, the generator’s code must either raise GeneratorExit or StopIteration. Catching the GeneratorExit exception and returning a value is illegal and will trigger a RuntimeError; if the function raises some other exception, that exception is propagated to the caller. close() will also be called by Python’s garbage collector when the generator is garbage-collected.

  If you need to run cleanup code when a GeneratorExit occurs, I suggest using a try: ... finally: suite instead of catching GeneratorExit.

The cumulative effect of these changes is to turn generators from one-way producers of information into both producers and consumers.

Generators also become coroutines, a more generalized form of subroutines. Subroutines are entered at one point and exited at another point (the top of the function, and a return statement), but coroutines can be entered, exited, and resumed at many different points (the yield statements). We’ll have to figure out patterns for using coroutines effectively in Python.

The addition of the close() method has one side effect that isn’t obvious. close() is called when a generator is garbage-collected, so this means the generator’s code gets one last chance to run before the generator is destroyed. This last chance means that try...finally statements in generators can now be guaranteed to work; the finally clause will now always get a chance to run. The syntactic restriction that you couldn’t mix yield statements with a try...finally suite has therefore been removed. This seems like a minor bit of language trivia, but using generators and try...finally is actually necessary in order to implement the with statement described by PEP 343. I’ll look at this new statement in the following section.

Another even more esoteric effect of this change: previously, the gi_frame attribute of a generator was always a frame object. It’s now possible for gi_frame to be None once the generator has been exhausted.

PEP 343: The ‘with’ statement

The ‘with‘ statement clarifies code that previously would use try...finally blocks to ensure that clean-up code is executed. In this section, I’ll discuss the statement as it will commonly be used. In the next section, I’ll examine the implementation details and show how to write objects for use with this statement.

The ‘with‘ statement is a new control-flow structure whose basic structure is:

with expression [as variable]:
    with-block

The expression is evaluated, and it should result in an object that supports the context management protocol (that is, has __enter__() and __exit__() methods.

The object’s __enter__() is called before with-block is executed and therefore can run set-up code. It also may return a value that is bound to the name variable, if given. (Note carefully that variable is not assigned the result of expression.)

After execution of the with-block is finished, the object’s __exit__() method is called, even if the block raised an exception, and can therefore run clean-up code.

To enable the statement in Python 2.5, you need to add the following directive to your module:

from __future__ import with_statement

The statement will always be enabled in Python 2.6.

Some standard Python objects now support the context management protocol and can be used with the ‘with‘ statement. File objects are one example:

with open('/etc/passwd', 'r') as f:
    for line in f:
        print line
        ... more processing code ...

After this statement has executed, the file object in f will have been automatically closed, even if the for loop raised an exception part- way through the block.

The threading module’s locks and condition variables also support the ‘with‘ statement:

lock = threading.Lock()
with lock:
    # Critical section of code
    ...

The lock is acquired before the block is executed and always released once the block is complete.

The new localcontext() function in the decimal module makes it easy to save and restore the current decimal context, which encapsulates the desired precision and rounding characteristics for computations:

from decimal import Decimal, Context, localcontext

# Displays with default precision of 28 digits
v = Decimal('578')
print v.sqrt()

with localcontext(Context(prec=16)):
    # All code in this block uses a precision of 16 digits.
    # The original context is restored on exiting the block.
    print v.sqrt()

db_connection = DatabaseConnection()
with db_connection as cursor:
    cursor.execute('insert into ...')
    cursor.execute('delete from ...')
    # ... more operations ...

class DatabaseConnection:
    # Database interface
    def cursor (self):
        "Returns a cursor object and starts a new transaction"
    def commit (self):
        "Commits current transaction"
    def rollback (self):
        "Rolls back current transaction"

class DatabaseConnection:
    ...
    def __enter__ (self):
        # Code to start a new transaction
        cursor = self.cursor()
        return cursor

class DatabaseConnection:
    ...
    def __exit__ (self, type, value, tb):
        if tb is None:
            # No exception, so commit
            self.commit()
        else:
            # Exception occurred, so rollback.
            self.rollback()
            # return False

from contextlib import contextmanager

@contextmanager
def db_transaction (connection):
    cursor = connection.cursor()
    try:
        yield cursor
    except:
        connection.rollback()
        raise
    else:
        connection.commit()

db = DatabaseConnection()
with db_transaction(db) as cursor:
    ...

lock = threading.Lock()
with nested (db_transaction(db), lock) as (cursor, locked):
    ...

import urllib, sys
from contextlib import closing

with closing(urllib.urlopen('http://www.yahoo.com')) as f:
    for line in f:
        sys.stdout.write(line)

PEP 352: Exceptions as New-Style Classes

Exception classes can now be new-style classes, not just classic classes, and the built-in Exception class and all the standard built-in exceptions (NameError, ValueError, etc.) are now new-style classes.

The inheritance hierarchy for exceptions has been rearranged a bit. In 2.5, the inheritance relationships are:

BaseException       # New in Python 2.5
|- KeyboardInterrupt
|- SystemExit
|- Exception
   |- (all other current built-in exceptions)

This rearrangement was done because people often want to catch all exceptions that indicate program errors. KeyboardInterrupt and SystemExit aren’t errors, though, and usually represent an explicit action such as the user hitting Control-C or code calling sys.exit(). A bare except: will catch all exceptions, so you commonly need to list KeyboardInterrupt and SystemExit in order to re-raise them. The usual pattern is:

try:
    ...
except (KeyboardInterrupt, SystemExit):
    raise
except:
    # Log error...
    # Continue running program...

In Python 2.5, you can now write except Exception to achieve the same result, catching all the exceptions that usually indicate errors but leaving KeyboardInterrupt and SystemExit alone. As in previous versions, a bare except: still catches all exceptions.

The goal for Python 3.0 is to require any class raised as an exception to derive from BaseException or some descendant of BaseException, and future releases in the Python 2.x series may begin to enforce this constraint. Therefore, I suggest you begin making all your exception classes derive from Exception now. It’s been suggested that the bare except: form should be removed in Python 3.0, but Guido van Rossum hasn’t decided whether to do this or not.

Raising of strings as exceptions, as in the statement raise "Error occurred", is deprecated in Python 2.5 and will trigger a warning. The aim is to be able to remove the string-exception feature in a few releases.

PEP 353: Using ssize_t as the index type

A wide-ranging change to Python’s C API, using a new Py_ssize_t type definition instead of int, will permit the interpreter to handle more data on 64-bit platforms. This change doesn’t affect Python’s capacity on 32-bit platforms.

Various pieces of the Python interpreter used C’s int type to store sizes or counts; for example, the number of items in a list or tuple were stored in an int. The C compilers for most 64-bit platforms still define int as a 32-bit type, so that meant that lists could only hold up to 2**31 - 1 = 2147483647 items. (There are actually a few different programming models that 64-bit C compilers can use – see http://www.unix.org/version2/whatsnew/lp64_wp.html for a discussion – but the most commonly available model leaves int as 32 bits.)

A limit of 2147483647 items doesn’t really matter on a 32-bit platform because you’ll run out of memory before hitting the length limit. Each list item requires space for a pointer, which is 4 bytes, plus space for a PyObject representing the item. 2147483647*4 is already more bytes than a 32-bit address space can contain.

It’s possible to address that much memory on a 64-bit platform, however. The pointers for a list that size would only require 16 GiB of space, so it’s not unreasonable that Python programmers might construct lists that large. Therefore, the Python interpreter had to be changed to use some type other than int, and this will be a 64-bit type on 64-bit platforms. The change will cause incompatibilities on 64-bit machines, so it was deemed worth making the transition now, while the number of 64-bit users is still relatively small. (In 5 or 10 years, we may all be on 64-bit machines, and the transition would be more painful then.)

This change most strongly affects authors of C extension modules. Python strings and container types such as lists and tuples now use Py_ssize_t to store their size. Functions such as PyList_Size() now return Py_ssize_t. Code in extension modules may therefore need to have some variables changed to Py_ssize_t.

The PyArg_ParseTuple() and Py_BuildValue() functions have a new conversion code, n, for Py_ssize_t. PyArg_ParseTuple()‘s s# and t# still output int by default, but you can define the macro PY_SSIZE_T_CLEAN before including Python.h to make them return Py_ssize_t.

PEP 353 has a section on conversion guidelines that extension authors should read to learn about supporting 64-bit platforms.

PEP 357: The ‘__index__’ method

The NumPy developers had a problem that could only be solved by adding a new special method, __index__(). When using slice notation, as in [start:stop:step], the values of the start, stop, and step indexes must all be either integers or long integers. NumPy defines a variety of specialized integer types corresponding to unsigned and signed integers of 8, 16, 32, and 64 bits, but there was no way to signal that these types could be used as slice indexes.

Slicing can’t just use the existing __int__() method because that method is also used to implement coercion to integers. If slicing used __int__(), floating-point numbers would also become legal slice indexes and that’s clearly an undesirable behaviour.

Instead, a new special method called __index__() was added. It takes no arguments and returns an integer giving the slice index to use. For example:

class C:
    def __index__ (self):
        return self.value

The return value must be either a Python integer or long integer. The interpreter will check that the type returned is correct, and raises a TypeError if this requirement isn’t met.

A corresponding nb_index slot was added to the C-level PyNumberMethods structure to let C extensions implement this protocol. PyNumber_Index(obj)() can be used in extension code to call the __index__() function and retrieve its result.

Other Language Changes

Here are all of the changes that Python 2.5 makes to the core Python language.

• The dict type has a new hook for letting subclasses provide a default value when a key isn’t contained in the dictionary. When a key isn’t found, the dictionary’s __missing__(key)() method will be called. This hook is used to implement the new defaultdict class in the collections module. The following example defines a dictionary that returns zero for any missing key:

  class zerodict (dict):
      def __missing__ (self, key):
          return 0
  
  d = zerodict({1:1, 2:2})
  print d[1], d[2]   # Prints 1, 2
  print d[3], d[4]   # Prints 0, 0
  

• Both 8-bit and Unicode strings have new partition(sep)() and rpartition(sep)() methods that simplify a common use case.

  The find(S)() method is often used to get an index which is then used to slice the string and obtain the pieces that are before and after the separator. partition(sep)() condenses this pattern into a single method call that returns a 3-tuple containing the substring before the separator, the separator itself, and the substring after the separator. If the separator isn’t found, the first element of the tuple is the entire string and the other two elements are empty. rpartition(sep)() also returns a 3-tuple but starts searching from the end of the string; the r stands for ‘reverse’.

  Some examples:

  >>> ('http://www.python.org').partition('://')
  ('http', '://', 'www.python.org')
  >>> ('file:/usr/share/doc/index.html').partition('://')
  ('file:/usr/share/doc/index.html', '', '')
  >>> (u'Subject: a quick question').partition(':')
  (u'Subject', u':', u' a quick question')
  >>> 'www.python.org'.rpartition('.')
  ('www.python', '.', 'org')
  >>> 'www.python.org'.rpartition(':')
  ('', '', 'www.python.org')
  

  (Implemented by Fredrik Lundh following a suggestion by Raymond Hettinger.)

• The startswith() and endswith() methods of string types now accept tuples of strings to check for.

  def is_image_file (filename):
      return filename.endswith(('.gif', '.jpg', '.tiff'))
  

  (Implemented by Georg Brandl following a suggestion by Tom Lynn.)

• The min() and max() built-in functions gained a key keyword parameter analogous to the key argument for sort(). This parameter supplies a function that takes a single argument and is called for every value in the list; min()/max() will return the element with the smallest/largest return value from this function. For example, to find the longest string in a list, you can do:

  L = ['medium', 'longest', 'short']
  # Prints 'longest'
  print max(L, key=len)
  # Prints 'short', because lexicographically 'short' has the largest value
  print max(L)
  

  (Contributed by Steven Bethard and Raymond Hettinger.)

• Two new built-in functions, any() and all(), evaluate whether an iterator contains any true or false values. any() returns True if any value returned by the iterator is true; otherwise it will return False. all() returns True only if all of the values returned by the iterator evaluate as true. (Suggested by Guido van Rossum, and implemented by Raymond Hettinger.)

• The result of a class’s __hash__() method can now be either a long integer or a regular integer. If a long integer is returned, the hash of that value is taken. In earlier versions the hash value was required to be a regular integer, but in 2.5 the id() built-in was changed to always return non-negative numbers, and users often seem to use id(self) in __hash__() methods (though this is discouraged).

• ASCII is now the default encoding for modules. It’s now a syntax error if a module contains string literals with 8-bit characters but doesn’t have an encoding declaration. In Python 2.4 this triggered a warning, not a syntax error. See PEP 263 for how to declare a module’s encoding; for example, you might add a line like this near the top of the source file:

  # -*- coding: latin1 -*-
  

• A new warning, UnicodeWarning, is triggered when you attempt to compare a Unicode string and an 8-bit string that can’t be converted to Unicode using the default ASCII encoding. The result of the comparison is false:

  >>> chr(128) == unichr(128)   # Can't convert chr(128) to Unicode
  __main__:1: UnicodeWarning: Unicode equal comparison failed
    to convert both arguments to Unicode - interpreting them
    as being unequal
  False
  >>> chr(127) == unichr(127)   # chr(127) can be converted
  True
  

  Previously this would raise a UnicodeDecodeError exception, but in 2.5 this could result in puzzling problems when accessing a dictionary. If you looked up unichr(128) and chr(128) was being used as a key, you’d get a UnicodeDecodeError exception. Other changes in 2.5 resulted in this exception being raised instead of suppressed by the code in dictobject.c that implements dictionaries.

  Raising an exception for such a comparison is strictly correct, but the change might have broken code, so instead UnicodeWarning was introduced.

  (Implemented by Marc-André Lemburg.)

• One error that Python programmers sometimes make is forgetting to include an __init__.py module in a package directory. Debugging this mistake can be confusing, and usually requires running Python with the -v switch to log all the paths searched. In Python 2.5, a new ImportWarning warning is triggered when an import would have picked up a directory as a package but no __init__.py was found. This warning is silently ignored by default; provide the -Wd option when running the Python executable to display the warning message. (Implemented by Thomas Wouters.)

• The list of base classes in a class definition can now be empty. As an example, this is now legal:

  class C():
      pass
  

  (Implemented by Brett Cannon.)

>>> quit
'Use Ctrl-D (i.e. EOF) to exit.'

New, Improved, and Removed Modules

The standard library received many enhancements and bug fixes in Python 2.5. Here’s a partial list of the most notable changes, sorted alphabetically by module name. Consult the Misc/NEWS file in the source tree for a more complete list of changes, or look through the SVN logs for all the details.

• The audioop module now supports the a-LAW encoding, and the code for u-LAW encoding has been improved. (Contributed by Lars Immisch.)

• The codecs module gained support for incremental codecs. The codec.lookup() function now returns a CodecInfo instance instead of a tuple. CodecInfo instances behave like a 4-tuple to preserve backward compatibility but also have the attributes encode, decode, incrementalencoder, incrementaldecoder, streamwriter, and streamreader. Incremental codecs can receive input and produce output in multiple chunks; the output is the same as if the entire input was fed to the non-incremental codec. See the codecs module documentation for details. (Designed and implemented by Walter Dörwald.)

• The collections module gained a new type, defaultdict, that subclasses the standard dict type. The new type mostly behaves like a dictionary but constructs a default value when a key isn’t present, automatically adding it to the dictionary for the requested key value.

  The first argument to defaultdict‘s constructor is a factory function that gets called whenever a key is requested but not found. This factory function receives no arguments, so you can use built-in type constructors such as list() or int(). For example, you can make an index of words based on their initial letter like this:

  words = """Nel mezzo del cammin di nostra vita
  mi ritrovai per una selva oscura
  che la diritta via era smarrita""".lower().split()
  
  index = defaultdict(list)
  
  for w in words:
      init_letter = w[0]
      index[init_letter].append(w)
  

  Printing index results in the following output:

  defaultdict(<type 'list'>, {'c': ['cammin', 'che'], 'e': ['era'],
          'd': ['del', 'di', 'diritta'], 'm': ['mezzo', 'mi'],
          'l': ['la'], 'o': ['oscura'], 'n': ['nel', 'nostra'],
          'p': ['per'], 's': ['selva', 'smarrita'],
          'r': ['ritrovai'], 'u': ['una'], 'v': ['vita', 'via']}
  

  (Contributed by Guido van Rossum.)

• The deque double-ended queue type supplied by the collections module now has a remove(value)() method that removes the first occurrence of value in the queue, raising ValueError if the value isn’t found. (Contributed by Raymond Hettinger.)

• New module: The contextlib module contains helper functions for use with the new ‘with‘ statement. See section The contextlib module for more about this module.

• New module: The cProfile module is a C implementation of the existing profile module that has much lower overhead. The module’s interface is the same as profile: you run cProfile.run('main()') to profile a function, can save profile data to a file, etc. It’s not yet known if the Hotshot profiler, which is also written in C but doesn’t match the profile module’s interface, will continue to be maintained in future versions of Python. (Contributed by Armin Rigo.)

  Also, the pstats module for analyzing the data measured by the profiler now supports directing the output to any file object by supplying a stream argument to the Stats constructor. (Contributed by Skip Montanaro.)

• The csv module, which parses files in comma-separated value format, received several enhancements and a number of bugfixes. You can now set the maximum size in bytes of a field by calling the csv.field_size_limit(new_limit)() function; omitting the new_limit argument will return the currently-set limit. The reader class now has a line_num attribute that counts the number of physical lines read from the source; records can span multiple physical lines, so line_num is not the same as the number of records read.

  The CSV parser is now stricter about multi-line quoted fields. Previously, if a line ended within a quoted field without a terminating newline character, a newline would be inserted into the returned field. This behavior caused problems when reading files that contained carriage return characters within fields, so the code was changed to return the field without inserting newlines. As a consequence, if newlines embedded within fields are important, the input should be split into lines in a manner that preserves the newline characters.

  (Contributed by Skip Montanaro and Andrew McNamara.)

• The datetime class in the datetime module now has a strptime(string, format)() method for parsing date strings, contributed by Josh Spoerri. It uses the same format characters as time.strptime() and time.strftime():

  from datetime import datetime
  
  ts = datetime.strptime('10:13:15 2006-03-07',
                         '%H:%M:%S %Y-%m-%d')
  

• The SequenceMatcher.get_matching_blocks() method in the difflib module now guarantees to return a minimal list of blocks describing matching subsequences. Previously, the algorithm would occasionally break a block of matching elements into two list entries. (Enhancement by Tim Peters.)

• The doctest module gained a SKIP option that keeps an example from being executed at all. This is intended for code snippets that are usage examples intended for the reader and aren’t actually test cases.

  An encoding parameter was added to the testfile() function and the DocFileSuite class to specify the file’s encoding. This makes it easier to use non-ASCII characters in tests contained within a docstring. (Contributed by Bjorn Tillenius.)

• The email package has been updated to version 4.0. (Contributed by Barry Warsaw.)

• The fileinput module was made more flexible. Unicode filenames are now supported, and a mode parameter that defaults to "r" was added to the input() function to allow opening files in binary or universal-newline mode. Another new parameter, openhook, lets you use a function other than open() to open the input files. Once you’re iterating over the set of files, the FileInput object’s new fileno() returns the file descriptor for the currently opened file. (Contributed by Georg Brandl.)

• In the gc module, the new get_count() function returns a 3-tuple containing the current collection counts for the three GC generations. This is accounting information for the garbage collector; when these counts reach a specified threshold, a garbage collection sweep will be made. The existing gc.collect() function now takes an optional generation argument of 0, 1, or 2 to specify which generation to collect. (Contributed by Barry Warsaw.)

• The nsmallest() and nlargest() functions in the heapq module now support a key keyword parameter similar to the one provided by the min()/max() functions and the sort() methods. For example:

  >>> import heapq
  >>> L = ["short", 'medium', 'longest', 'longer still']
  >>> heapq.nsmallest(2, L)  # Return two lowest elements, lexicographically
  ['longer still', 'longest']
  >>> heapq.nsmallest(2, L, key=len)   # Return two shortest elements
  ['short', 'medium']
  

  (Contributed by Raymond Hettinger.)

• The itertools.islice() function now accepts None for the start and step arguments. This makes it more compatible with the attributes of slice objects, so that you can now write the following:

  s = slice(5)     # Create slice object
  itertools.islice(iterable, s.start, s.stop, s.step)
  

  (Contributed by Raymond Hettinger.)

• The format() function in the locale module has been modified and two new functions were added, format_string() and currency().

  The format() function’s val parameter could previously be a string as long as no more than one %char specifier appeared; now the parameter must be exactly one %char specifier with no surrounding text. An optional monetary parameter was also added which, if True, will use the locale’s rules for formatting currency in placing a separator between groups of three digits.

  To format strings with multiple %char specifiers, use the new format_string() function that works like format() but also supports mixing %char specifiers with arbitrary text.

  A new currency() function was also added that formats a number according to the current locale’s settings.

  (Contributed by Georg Brandl.)

• The mailbox module underwent a massive rewrite to add the capability to modify mailboxes in addition to reading them. A new set of classes that include mbox, MH, and Maildir are used to read mailboxes, and have an add(message)() method to add messages, remove(key)() to remove messages, and lock()/unlock() to lock/unlock the mailbox. The following example converts a maildir-format mailbox into an mbox-format one:

  import mailbox
  
  # 'factory=None' uses email.Message.Message as the class representing
  # individual messages.
  src = mailbox.Maildir('maildir', factory=None)
  dest = mailbox.mbox('/tmp/mbox')
  
  for msg in src:
      dest.add(msg)
  

  (Contributed by Gregory K. Johnson. Funding was provided by Google’s 2005 Summer of Code.)

• New module: the msilib module allows creating Microsoft Installer .msi files and CAB files. Some support for reading the .msi database is also included. (Contributed by Martin von Löwis.)

• The nis module now supports accessing domains other than the system default domain by supplying a domain argument to the nis.match() and nis.maps() functions. (Contributed by Ben Bell.)

• The operator module’s itemgetter() and attrgetter() functions now support multiple fields. A call such as operator.attrgetter('a', 'b') will return a function that retrieves the a and b attributes. Combining this new feature with the sort() method’s key parameter lets you easily sort lists using multiple fields. (Contributed by Raymond Hettinger.)

• The optparse module was updated to version 1.5.1 of the Optik library. The OptionParser class gained an epilog attribute, a string that will be printed after the help message, and a destroy() method to break reference cycles created by the object. (Contributed by Greg Ward.)

• The os module underwent several changes. The stat_float_times variable now defaults to true, meaning that os.stat() will now return time values as floats. (This doesn’t necessarily mean that os.stat() will return times that are precise to fractions of a second; not all systems support such precision.)

  Constants named os.SEEK_SET, os.SEEK_CUR, and os.SEEK_END have been added; these are the parameters to the os.lseek() function. Two new constants for locking are os.O_SHLOCK and os.O_EXLOCK.

  Two new functions, wait3() and wait4(), were added. They’re similar the waitpid() function which waits for a child process to exit and returns a tuple of the process ID and its exit status, but wait3() and wait4() return additional information. wait3() doesn’t take a process ID as input, so it waits for any child process to exit and returns a 3-tuple of process-id, exit-status, resource-usage as returned from the resource.getrusage() function. wait4(pid)() does take a process ID. (Contributed by Chad J. Schroeder.)

  On FreeBSD, the os.stat() function now returns times with nanosecond resolution, and the returned object now has st_gen and st_birthtime. The st_flags member is also available, if the platform supports it. (Contributed by Antti Louko and Diego Pettenò.)

• The Python debugger provided by the pdb module can now store lists of commands to execute when a breakpoint is reached and execution stops. Once breakpoint #1 has been created, enter commands 1 and enter a series of commands to be executed, finishing the list with end. The command list can include commands that resume execution, such as continue or next. (Contributed by Grégoire Dooms.)

• The pickle and cPickle modules no longer accept a return value of None from the __reduce__() method; the method must return a tuple of arguments instead. The ability to return None was deprecated in Python 2.4, so this completes the removal of the feature.

• The pkgutil module, containing various utility functions for finding packages, was enhanced to support PEP 302’s import hooks and now also works for packages stored in ZIP-format archives. (Contributed by Phillip J. Eby.)

• The pybench benchmark suite by Marc-André Lemburg is now included in the Tools/pybench directory. The pybench suite is an improvement on the commonly used pystone.py program because pybench provides a more detailed measurement of the interpreter’s speed. It times particular operations such as function calls, tuple slicing, method lookups, and numeric operations, instead of performing many different operations and reducing the result to a single number as pystone.py does.

• The pyexpat module now uses version 2.0 of the Expat parser. (Contributed by Trent Mick.)

• The Queue class provided by the Queue module gained two new methods. join() blocks until all items in the queue have been retrieved and all processing work on the items have been completed. Worker threads call the other new method, task_done(), to signal that processing for an item has been completed. (Contributed by Raymond Hettinger.)

• The old regex and regsub modules, which have been deprecated ever since Python 2.0, have finally been deleted. Other deleted modules: statcache, tzparse, whrandom.

• Also deleted: the lib-old directory, which includes ancient modules such as dircmp and ni, was removed. lib-old wasn’t on the default sys.path, so unless your programs explicitly added the directory to sys.path, this removal shouldn’t affect your code.

• The rlcompleter module is no longer dependent on importing the readline module and therefore now works on non-Unix platforms. (Patch from Robert Kiendl.)

• The SimpleXMLRPCServer and DocXMLRPCServer classes now have a rpc_paths attribute that constrains XML-RPC operations to a limited set of URL paths; the default is to allow only '/' and '/RPC2'. Setting rpc_paths to None or an empty tuple disables this path checking.

• The socket module now supports AF_NETLINK sockets on Linux, thanks to a patch from Philippe Biondi. Netlink sockets are a Linux-specific mechanism for communications between a user-space process and kernel code; an introductory article about them is at http://www.linuxjournal.com/article/7356. In Python code, netlink addresses are represented as a tuple of 2 integers, (pid, group_mask).

  Two new methods on socket objects, recv_into(buffer)() and recvfrom_into(buffer)(), store the received data in an object that supports the buffer protocol instead of returning the data as a string. This means you can put the data directly into an array or a memory-mapped file.

  Socket objects also gained getfamily(), gettype(), and getproto() accessor methods to retrieve the family, type, and protocol values for the socket.

• New module: the spwd module provides functions for accessing the shadow password database on systems that support shadow passwords.

• The struct is now faster because it compiles format strings into Struct objects with pack() and unpack() methods. This is similar to how the re module lets you create compiled regular expression objects. You can still use the module-level pack() and unpack() functions; they’ll create Struct objects and cache them. Or you can use Struct instances directly:

  s = struct.Struct('ih3s')
  
  data = s.pack(1972, 187, 'abc')
  year, number, name = s.unpack(data)
  

  You can also pack and unpack data to and from buffer objects directly using the pack_into(buffer, offset, v1, v2, ...)() and unpack_from(buffer, offset)() methods. This lets you store data directly into an array or a memory- mapped file.

  (Struct objects were implemented by Bob Ippolito at the NeedForSpeed sprint. Support for buffer objects was added by Martin Blais, also at the NeedForSpeed sprint.)

• The Python developers switched from CVS to Subversion during the 2.5 development process. Information about the exact build version is available as the sys.subversion variable, a 3-tuple of (interpreter-name, branch-name, revision-range). For example, at the time of writing my copy of 2.5 was reporting ('CPython', 'trunk', '45313:45315').

  This information is also available to C extensions via the Py_GetBuildInfo() function that returns a string of build information like this: "trunk:45355:45356M, Apr 13 2006, 07:42:19". (Contributed by Barry Warsaw.)

• Another new function, sys._current_frames(), returns the current stack frames for all running threads as a dictionary mapping thread identifiers to the topmost stack frame currently active in that thread at the time the function is called. (Contributed by Tim Peters.)

• The TarFile class in the tarfile module now has an extractall() method that extracts all members from the archive into the current working directory. It’s also possible to set a different directory as the extraction target, and to unpack only a subset of the archive’s members.

  The compression used for a tarfile opened in stream mode can now be autodetected using the mode 'r|*'. (Contributed by Lars Gustäbel.)

• The threading module now lets you set the stack size used when new threads are created. The stack_size([*size*])() function returns the currently configured stack size, and supplying the optional size parameter sets a new value. Not all platforms support changing the stack size, but Windows, POSIX threading, and OS/2 all do. (Contributed by Andrew MacIntyre.)

• The unicodedata module has been updated to use version 4.1.0 of the Unicode character database. Version 3.2.0 is required by some specifications, so it’s still available as unicodedata.ucd_3_2_0.

• New module: the uuid module generates universally unique identifiers (UUIDs) according to RFC 4122. The RFC defines several different UUID versions that are generated from a starting string, from system properties, or purely randomly. This module contains a UUID class and functions named uuid1(), uuid3(), uuid4(), and uuid5() to generate different versions of UUID. (Version 2 UUIDs are not specified in RFC 4122 and are not supported by this module.)

  >>> import uuid
  >>> # make a UUID based on the host ID and current time
  >>> uuid.uuid1()
  UUID('a8098c1a-f86e-11da-bd1a-00112444be1e')
  
  >>> # make a UUID using an MD5 hash of a namespace UUID and a name
  >>> uuid.uuid3(uuid.NAMESPACE_DNS, 'python.org')
  UUID('6fa459ea-ee8a-3ca4-894e-db77e160355e')
  
  >>> # make a random UUID
  >>> uuid.uuid4()
  UUID('16fd2706-8baf-433b-82eb-8c7fada847da')
  
  >>> # make a UUID using a SHA-1 hash of a namespace UUID and a name
  >>> uuid.uuid5(uuid.NAMESPACE_DNS, 'python.org')
  UUID('886313e1-3b8a-5372-9b90-0c9aee199e5d')
  

  (Contributed by Ka-Ping Yee.)

• The weakref module’s WeakKeyDictionary and WeakValueDictionary types gained new methods for iterating over the weak references contained in the dictionary. iterkeyrefs() and keyrefs() methods were added to WeakKeyDictionary, and itervaluerefs() and valuerefs() were added to WeakValueDictionary. (Contributed by Fred L. Drake, Jr.)

• The webbrowser module received a number of enhancements. It’s now usable as a script with python -m webbrowser, taking a URL as the argument; there are a number of switches to control the behaviour (-n for a new browser window, -t for a new tab). New module-level functions, open_new() and open_new_tab(), were added to support this. The module’s open() function supports an additional feature, an autoraise parameter that signals whether to raise the open window when possible. A number of additional browsers were added to the supported list such as Firefox, Opera, Konqueror, and elinks. (Contributed by Oleg Broytmann and Georg Brandl.)

• The xmlrpclib module now supports returning datetime objects for the XML-RPC date type. Supply use_datetime=True to the loads() function or the Unmarshaller class to enable this feature. (Contributed by Skip Montanaro.)

• The zipfile module now supports the ZIP64 version of the format, meaning that a .zip archive can now be larger than 4 GiB and can contain individual files larger than 4 GiB. (Contributed by Ronald Oussoren.)

• The zlib module’s Compress and Decompress objects now support a copy() method that makes a copy of the object’s internal state and returns a new Compress or Decompress object. (Contributed by Chris AtLee.)

import ctypes

libc = ctypes.CDLL('libc.so.6')
result = libc.printf("Line of output\n")

s = "this is a string"
buf = ctypes.create_string_buffer(s)
libc.strfry(buf)

>>> libc.atof('2.71828')
-1783957616
>>> libc.atof.restype = ctypes.c_double
>>> libc.atof('2.71828')
2.71828

import ctypes

d = {}
ctypes.pythonapi.PyObject_SetItem(ctypes.py_object(d),
          ctypes.py_object("abc"),  ctypes.py_object(1))
# d is now {'abc', 1}.

from xml.etree import ElementTree as ET

tree = ET.parse('ex-1.xml')

feed = urllib.urlopen(
          'http://planet.python.org/rss10.xml')
tree = ET.parse(feed)

svg = ET.XML("""<svg width="10px" version="1.0">
             </svg>""")
svg.set('height', '320px')
svg.append(elem1)

if elem.tag is ET.Comment:
    ...
elif elem.tag is ET.ProcessingInstruction:
    ...

# Encoding is US-ASCII
tree.write('output.xml')

# Encoding is UTF-8
f = open('output.xml', 'w')
tree.write(f, encoding='utf-8')

# Old versions
h = md5.md5()
h = md5.new()

# New version
h = hashlib.md5()

# Old versions
h = sha.sha()
h = sha.new()

# New version
h = hashlib.sha1()

# Hash that weren't previously available
h = hashlib.sha224()
h = hashlib.sha256()
h = hashlib.sha384()
h = hashlib.sha512()

# Alternative form
h = hashlib.new('md5')          # Provide algorithm as a string

conn = sqlite3.connect('/tmp/example')

c = conn.cursor()

# Create table
c.execute('''create table stocks
(date text, trans text, symbol text,
 qty real, price real)''')

# Insert a row of data
c.execute("""insert into stocks
          values ('2006-01-05','BUY','RHAT',100,35.14)""")

# Never do this -- insecure!
symbol = 'IBM'
c.execute("... where symbol = '%s'" % symbol)

# Do this instead
t = (symbol,)
c.execute('select * from stocks where symbol=?', t)

# Larger example
for t in (('2006-03-28', 'BUY', 'IBM', 1000, 45.00),
          ('2006-04-05', 'BUY', 'MSOFT', 1000, 72.00),
          ('2006-04-06', 'SELL', 'IBM', 500, 53.00),
         ):
    c.execute('insert into stocks values (?,?,?,?,?)', t)

>>> c = conn.cursor()
>>> c.execute('select * from stocks order by price')
>>> for row in c:
...    print row
...
(u'2006-01-05', u'BUY', u'RHAT', 100, 35.140000000000001)
(u'2006-03-28', u'BUY', u'IBM', 1000, 45.0)
(u'2006-04-06', u'SELL', u'IBM', 500, 53.0)
(u'2006-04-05', u'BUY', u'MSOFT', 1000, 72.0)
>>>

from wsgiref import simple_server

wsgi_app = ...

host = ''
port = 8000
httpd = simple_server.make_server(host, port, wsgi_app)
httpd.serve_forever()

Build and C API Changes

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

• The Python source tree was converted from CVS to Subversion, in a complex migration procedure that was supervised and flawlessly carried out by Martin von Löwis. The procedure was developed as PEP 347.

• Coverity, a company that markets a source code analysis tool called Prevent, provided the results of their examination of the Python source code. The analysis found about 60 bugs that were quickly fixed. Many of the bugs were refcounting problems, often occurring in error-handling code. See http://scan.coverity.com for the statistics.

• The largest change to the C API came from PEP 353, which modifies the interpreter to use a Py_ssize_t type definition instead of int. See the earlier section PEP 353: Using ssize_t as the index type for a discussion of this change.

• The design of the bytecode compiler has changed a great deal, no longer generating bytecode by traversing the parse tree. Instead the parse tree is converted to an abstract syntax tree (or AST), and it is the abstract syntax tree that’s traversed to produce the bytecode.

  It’s possible for Python code to obtain AST objects by using the compile() built-in and specifying _ast.PyCF_ONLY_AST as the value of the flags parameter:

  from _ast import PyCF_ONLY_AST
  ast = compile("""a=0
  for i in range(10):
      a += i
  """, "<string>", 'exec', PyCF_ONLY_AST)
  
  assignment = ast.body[0]
  for_loop = ast.body[1]
  

  No official documentation has been written for the AST code yet, but PEP 339 discusses the design. To start learning about the code, read the definition of the various AST nodes in Parser/Python.asdl. A Python script reads this file and generates a set of C structure definitions in Include/Python-ast.h. The PyParser_ASTFromString() and PyParser_ASTFromFile(), defined in Include/pythonrun.h, take Python source as input and return the root of an AST representing the contents. This AST can then be turned into a code object by PyAST_Compile(). For more information, read the source code, and then ask questions on python-dev.

  The AST code was developed under Jeremy Hylton’s management, and implemented by (in alphabetical order) Brett Cannon, Nick Coghlan, Grant Edwards, John Ehresman, Kurt Kaiser, Neal Norwitz, Tim Peters, Armin Rigo, and Neil Schemenauer, plus the participants in a number of AST sprints at conferences such as PyCon.

• Evan Jones’s patch to obmalloc, first described in a talk at PyCon DC 2005, was applied. Python 2.4 allocated small objects in 256K-sized arenas, but never freed arenas. With this patch, Python will free arenas when they’re empty. The net effect is that on some platforms, when you allocate many objects, Python’s memory usage may actually drop when you delete them and the memory may be returned to the operating system. (Implemented by Evan Jones, and reworked by Tim Peters.)

  Note that this change means extension modules must be more careful when allocating memory. Python’s API has many different functions for allocating memory that are grouped into families. For example, PyMem_Malloc(), PyMem_Realloc(), and PyMem_Free() are one family that allocates raw memory, while PyObject_Malloc(), PyObject_Realloc(), and PyObject_Free() are another family that’s supposed to be used for creating Python objects.

  Previously these different families all reduced to the platform’s malloc() and free() functions. This meant it didn’t matter if you got things wrong and allocated memory with the PyMem() function but freed it with the PyObject() function. With 2.5’s changes to obmalloc, these families now do different things and mismatches will probably result in a segfault. You should carefully test your C extension modules with Python 2.5.

• The built-in set types now have an official C API. Call PySet_New() and PyFrozenSet_New() to create a new set, PySet_Add() and PySet_Discard() to add and remove elements, and PySet_Contains() and PySet_Size() to examine the set’s state. (Contributed by Raymond Hettinger.)

• C code can now obtain information about the exact revision of the Python interpreter by calling the Py_GetBuildInfo() function that returns a string of build information like this: "trunk:45355:45356M, Apr 13 2006, 07:42:19". (Contributed by Barry Warsaw.)

• Two new macros can be used to indicate C functions that are local to the current file so that a faster calling convention can be used. Py_LOCAL(type)() declares the function as returning a value of the specified type and uses a fast-calling qualifier. Py_LOCAL_INLINE(type)() does the same thing and also requests the function be inlined. If PY_LOCAL_AGGRESSIVE() is defined before python.h is included, a set of more aggressive optimizations are enabled for the module; you should benchmark the results to find out if these optimizations actually make the code faster. (Contributed by Fredrik Lundh at the NeedForSpeed sprint.)

• PyErr_NewException(name, base, dict)() can now accept a tuple of base classes as its base argument. (Contributed by Georg Brandl.)

• The PyErr_Warn() function for issuing warnings is now deprecated in favour of PyErr_WarnEx(category, message, stacklevel)() which lets you specify the number of stack frames separating this function and the caller. A stacklevel of 1 is the function calling PyErr_WarnEx(), 2 is the function above that, and so forth. (Added by Neal Norwitz.)

• The CPython interpreter is still written in C, but the code can now be compiled with a C++ compiler without errors. (Implemented by Anthony Baxter, Martin von Löwis, Skip Montanaro.)

• The PyRange_New() function was removed. It was never documented, never used in the core code, and had dangerously lax error checking. In the unlikely case that your extensions were using it, you can replace it by something like the following:

  range = PyObject_CallFunction((PyObject*) &PyRange_Type, "lll",
                                start, stop, step);
  

Porting to Python 2.5

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

• ASCII is now the default encoding for modules. It’s now a syntax error if a module contains string literals with 8-bit characters but doesn’t have an encoding declaration. In Python 2.4 this triggered a warning, not a syntax error.
• Previously, the gi_frame attribute of a generator was always a frame object. Because of the PEP 342 changes described in section PEP 342: New Generator Features, it’s now possible for gi_frame to be None.
• A new warning, UnicodeWarning, is triggered when you attempt to compare a Unicode string and an 8-bit string that can’t be converted to Unicode using the default ASCII encoding. Previously such comparisons would raise a UnicodeDecodeError exception.
• Library: the csv module is now stricter about multi-line quoted fields. If your files contain newlines embedded within fields, the input should be split into lines in a manner which preserves the newline characters.
• Library: the locale module’s format() function’s would previously accept any string as long as no more than one %char specifier appeared. In Python 2.5, the argument must be exactly one %char specifier with no surrounding text.
• Library: The pickle and cPickle modules no longer accept a return value of None from the __reduce__() method; the method must return a tuple of arguments instead. The modules also no longer accept the deprecated bin keyword parameter.
• Library: The SimpleXMLRPCServer and DocXMLRPCServer classes now have a rpc_paths attribute that constrains XML-RPC operations to a limited set of URL paths; the default is to allow only '/' and '/RPC2'. Setting rpc_paths to None or an empty tuple disables this path checking.
• C API: Many functions now use Py_ssize_t instead of int to allow processing more data on 64-bit machines. Extension code may need to make the same change to avoid warnings and to support 64-bit machines. See the earlier section PEP 353: Using ssize_t as the index type for a discussion of this change.
• C API: The obmalloc changes mean that you must be careful to not mix usage of the PyMem_*() and PyObject_*() families of functions. Memory allocated with one family’s *_Malloc() must be freed with the corresponding family’s *_Free() function.

Acknowledgements

The author would like to thank the following people for offering suggestions, corrections and assistance with various drafts of this article: Georg Brandl, Nick Coghlan, Phillip J. Eby, Lars Gustäbel, Raymond Hettinger, Ralf W. Grosse- Kunstleve, Kent Johnson, Iain Lowe, Martin von Löwis, Fredrik Lundh, Andrew McNamara, Skip Montanaro, Gustavo Niemeyer, Paul Prescod, James Pryor, Mike Rovner, Scott Weikart, Barry Warsaw, Thomas Wouters.