Misc/NEWS

What's New in Python 2.1 alpha 2?
=================================

Core language, builtins, and interpreter

- repr(string) is easier to read, now using hex escapes instead of octal,
  and using \t, \n and \r instead of \011, \012 and \015 (respectively):

  >>> "\texample \r\n" + chr(0) + chr(255)
  '\texample \r\n\x00\xff'         # in 2.1
  '\011example \015\012\000\377'   # in 2.0


Standard library

- random.py is now self-contained, and offers all the functionality of
  the now-deprecated whrandom.py.  See the docs for details.  random.py
  also supports new functions getstate() and setstate(), for saving
  and restoring the internal state of the generator; and jumpahead(n),
  for quickly forcing the internal state to be the same as if n calls to
  random() had been made.  The latter is particularly useful for multi-
  threaded programs, creating one instance of the random.Random() class for
  each thread, then using .jumpahead() to force each instance to use a
  non-overlapping segment of the full period.


What's New in Python 2.1 alpha 1?
=================================

Core language, builtins, and interpreter

- There is a new Unicode companion to the PyObject_Str() API
  called PyObject_Unicode(). It behaves in the same way as the
  former, but assures that the returned value is an Unicode object
  (applying the usual coercion if necessary).

- The comparison operators support "rich comparison overloading" (PEP
  207).  C extension types can provide a rich comparison function in
  the new tp_richcompare slot in the type object.  The cmp() function
  and the C function PyObject_Compare() first try the new rich
  comparison operators before trying the old 3-way comparison.  There
  is also a new C API PyObject_RichCompare() (which also falls back on
  the old 3-way comparison, but does not constrain the outcome of the
  rich comparison to a Boolean result).

  The rich comparison function takes two objects (at least one of
  which is guaranteed to have the type that provided the function) and
  an integer indicating the opcode, which can be Py_LT, Py_LE, Py_EQ,
  Py_NE, Py_GT, Py_GE (for <, <=, ==, !=, >, >=), and returns a Python
  object, which may be NotImplemented (in which case the tp_compare
  slot function is used as a fallback, if defined).

  Classes can overload individual comparison operators by defining one
  or more of the methods__lt__, __le__, __eq__, __ne__, __gt__,
  __ge__.  There are no explicit "reflected argument" versions of
  these; instead, __lt__ and __gt__ are each other's reflection,
  likewise for__le__ and __ge__; __eq__ and __ne__ are their own
  reflection (similar at the C level).  No other implications are
  made; in particular, Python does not assume that == is the Boolean
  inverse of !=, or that < is the Boolean inverse of >=.  This makes
  it possible to define types with partial orderings.

  Classes or types that want to implement (in)equality tests but not
  the ordering operators (i.e. unordered types) should implement ==
  and !=, and raise an error for the ordering operators.

  It is possible to define types whose rich comparison results are not
  Boolean; e.g. a matrix type might want to return a matrix of bits
  for A < B, giving elementwise comparisons.  Such types should ensure
  that any interpretation of their value in a Boolean context raises
  an exception, e.g. by defining __nonzero__ (or the tp_nonzero slot
  at the C level) to always raise an exception.

- Complex numbers use rich comparisons to define == and != but raise
  an exception for <, <=, > and >=.  Unfortunately, this also means
  that cmp() of two complex numbers raises an exception when the two
  numbers differ.  Since it is not mathematically meaningful to compare
  complex numbers except for equality, I hope that this doesn't break
  too much code.

- Functions and methods now support getting and setting arbitrarily
  named attributes (PEP 232).  Functions have a new __dict__
  (a.k.a. func_dict) which hold the function attributes.  Methods get
  and set attributes on their underlying im_func.  It is a TypeError
  to set an attribute on a bound method.

- The xrange() object implementation has been improved so that
  xrange(sys.maxint) can be used on 64-bit platforms.  There's still a
  limitation that in this case len(xrange(sys.maxint)) can't be
  calculated, but the common idiom "for i in xrange(sys.maxint)" will
  work fine as long as the index i doesn't actually reach 2**31.
  (Python uses regular ints for sequence and string indices; fixing
  that is much more work.)

- Two changes to from...import:

  1) "from M import X" now works even if M is not a real module; it's
     basically a getattr() operation with AttributeError exceptions
     changed into ImportError.

  2) "from M import *" now looks for M.__all__ to decide which names to
     import; if M.__all__ doesn't exist, it uses M.__dict__.keys() but
     filters out names starting with '_' as before.  Whether or not
     __all__ exists, there's no restriction on the type of M.

- File objects have a new method, xreadlines().  This is the fastest
  way to iterate over all lines in a file:

  for line in file.xreadlines():
      ...do something to line...

  See the xreadlines module (mentioned below) for how to do this for
  other file-like objects.

- Even if you don't use file.xreadlines(), you may expect a speedup on
  line-by-line input.  The file.readline() method has been optimized
  quite a bit in platform-specific ways:  on systems (like Linux) that
  support flockfile(), getc_unlocked(), and funlockfile(), those are
  used by default.  On systems (like Windows) without getc_unlocked(),
  a complicated (but still thread-safe) method using fgets() is used by
  default.

  You can force use of the fgets() method by #define'ing
  USE_FGETS_IN_GETLINE at build time (it may be faster than
  getc_unlocked()).

  You can force fgets() not to be used by #define'ing
  DONT_USE_FGETS_IN_GETLINE (this is the first thing to try if std test
  test_bufio.py fails -- and let us know if it does!).

- In addition, the fileinput module, while still slower than the other
  methods on most platforms, has been sped up too, by using
  file.readlines(sizehint).

- Support for run-time warnings has been added, including a new
  command line option (-W) to specify the disposition of warnings.
  See the description of the warnings module below.

- Extensive changes have been made to the coercion code.  This mostly
  affects extension modules (which can now implement mixed-type
  numerical operators without having to use coercion), but
  occasionally, in boundary cases the coercion semantics have changed
  subtly.  Since this was a terrible gray area of the language, this
  is considered an improvement.  Also note that __rcmp__ is no longer
  supported -- instead of calling __rcmp__, __cmp__ is called with
  reflected arguments.

- In connection with the coercion changes, a new built-in singleton
  object, NotImplemented is defined.  This can be returned for
  operations that wish to indicate they are not implemented for a
  particular combination of arguments.  From C, this is
  Py_NotImplemented.

- The interpreter accepts now bytecode files on the command line even
  if they do not have a .pyc or .pyo extension. On Linux, after executing

  echo ':pyc:M::\x87\xc6\x0d\x0a::/usr/local/bin/python:' > /proc/sys/fs/binfmt_misc/register

  any byte code file can be used as an executable (i.e. as an argument
  to execve(2)).

- %[xXo] formats of negative Python longs now produce a sign
  character.  In 1.6 and earlier, they never produced a sign,
  and raised an error if the value of the long was too large
  to fit in a Python int.  In 2.0, they produced a sign if and
  only if too large to fit in an int.  This was inconsistent
  across platforms (because the size of an int varies across
  platforms), and inconsistent with hex() and oct().  Example:

  >>> "%x" % -0x42L
  '-42'      # in 2.1
  'ffffffbe' # in 2.0 and before, on 32-bit machines
  >>> hex(-0x42L)
  '-0x42L'   # in all versions of Python

  The behavior of %d formats for negative Python longs remains
  the same as in 2.0 (although in 1.6 and before, they raised
  an error if the long didn't fit in a Python int).

  %u formats don't make sense for Python longs, but are allowed
  and treated the same as %d in 2.1.  In 2.0, a negative long
  formatted via %u produced a sign if and only if too large to
  fit in an int.  In 1.6 and earlier, a negative long formatted
  via %u raised an error if it was too big to fit in an int.

- Dictionary objects have an odd new method, popitem().  This removes
  an arbitrary item from the dictionary and returns it (in the form of
  a (key, value) pair).  This can be useful for algorithms that use a
  dictionary as a bag of "to do" items and repeatedly need to pick one
  item.  Such algorithms normally end up running in quadratic time;
  using popitem() they can usually be made to run in linear time.

Standard library

- In the time module, the time argument to the functions strftime,
  localtime, gmtime, asctime and ctime is now optional, defaulting to
  the current time (in the local timezone).

- The ftplib module now defaults to passive mode, which is deemed a
  more useful default given that clients are often inside firewalls
  these days.  Note that this could break if ftplib is used to connect
  to a *server* that is inside a firewall, from outside; this is
  expected to be a very rare situation.  To fix that, you can call
  ftp.set_pasv(0).

- The module site now treats .pth files not only for path configuration,
  but also supports extensions to the initialization code: Lines starting
  with import are executed.

- There's a new module, warnings, which implements a mechanism for
  issuing and filtering warnings.  There are some new built-in
  exceptions that serve as warning categories, and a new command line
  option, -W, to control warnings (e.g. -Wi ignores all warnings, -We
  turns warnings into errors).  warnings.warn(message[, category])
  issues a warning message; this can also be called from C as
  PyErr_Warn(category, message).

- A new module xreadlines was added.  This exports a single factory
  function, xreadlines().  The intention is that this code is the
  absolutely fastest way to iterate over all lines in an open
  file(-like) object:

  import xreadlines
  for line in xreadlines.xreadlines(file):
      ...do something to line...

  This is equivalent to the previous the speed record holder using
  file.readlines(sizehint).  Note that if file is a real file object
  (as opposed to a file-like object), this is equivalent:

  for line in file.xreadlines():
      ...do something to line...

- The bisect module has new functions bisect_left, insort_left,
  bisect_right and insort_right.  The old names bisect and insort
  are now aliases for bisect_right and insort_right.  XXX_right
  and XXX_left methods differ in what happens when the new element
  compares equal to one or more elements already in the list:  the
  XXX_left methods insert to the left, the XXX_right methods to the
  right.  Code that doesn't care where equal elements end up should
  continue to use the old, short names ("bisect" and "insort").

- The new curses.panel module wraps the panel library that forms part
  of SYSV curses and ncurses.  Contributed by Thomas Gellekum.

- The SocketServer module now sets the allow_reuse_address flag by
  default in the TCPServer class.

- A new function, sys._getframe(), returns the stack frame pointer of
  the caller.  This is intended only as a building block for
  higher-level mechanisms such as string interpolation.

Build issues

- For Unix (and Unix-compatible) builds, configuration and building of
  extension modules is now greatly automated.  Rather than having to
  edit the Modules/Setup file to indicate which modules should be
  built and where their include files and libraries are, a
  distutils-based setup.py script now takes care of building most
  extension modules.  All extension modules built this way are built
  as shared libraries.  Only a few modules that must be linked
  statically are still listed in the Setup file; you won't need to
  edit their configuration.

- Python should now build out of the box on Cygwin.  If it doesn't,
  mail to Jason Tishler (jlt63 at users.sourceforge.net).

- Python now always uses its own (renamed) implementation of getopt()
  -- there's too much variation among C library getopt()
  implementations.

- C++ compilers are better supported; the CXX macro is always set to a
  C++ compiler if one is found.

Windows changes

- select module:  By default under Windows, a select() call
  can specify no more than 64 sockets.  Python now boosts
  this Microsoft default to 512.  If you need even more than
  that, see the MS docs (you'll need to #define FD_SETSIZE
  and recompile Python from source).

- Support for Windows 3.1, DOS and OS/2 is gone.  The Lib/dos-8x3
  subdirectory is no more!


What's New in Python 2.0?
=========================

Below is a list of all relevant changes since release 1.6.  Older
changes are in the file HISTORY.  If you are making the jump directly
from Python 1.5.2 to 2.0, make sure to read the section for 1.6 in the
HISTORY file!  Many important changes listed there.

Alternatively, a good overview of the changes between 1.5.2 and 2.0 is
the document "What's New in Python 2.0" by Kuchling and Moshe Zadka:
http://starship.python.net/crew/amk/python/writing/new-python/.

--Guido van Rossum (home page: http://www.pythonlabs.com/~guido/)

======================================================================

What's new in 2.0 (since release candidate 1)?
==============================================

Standard library

- The copy_reg module was modified to clarify its intended use: to
  register pickle support for extension types, not for classes.
  pickle() will raise a TypeError if it is passed a class.

- Fixed a bug in gettext's "normalize and expand" code that prevented
  it from finding an existing .mo file.

- Restored support for HTTP/0.9 servers in httplib.

- The math module was changed to stop raising OverflowError in case of
  underflow, and return 0 instead in underflow cases.  Whether Python
  used to raise OverflowError in case of underflow was platform-
  dependent (it did when the platform math library set errno to ERANGE
  on underflow).

- Fixed a bug in StringIO that occurred when the file position was not
  at the end of the file and write() was called with enough data to
  extend past the end of the file.

- Fixed a bug that caused Tkinter error messages to get lost on
  Windows.  The bug was fixed by replacing direct use of
  interp->result with Tcl_GetStringResult(interp).

- Fixed bug in urllib2 that caused it to fail when it received an HTTP
  redirect response.

- Several changes were made to distutils: Some debugging code was
  removed from util.  Fixed the installer used when an external zip
  program (like WinZip) is not found; the source code for this
  installer is in Misc/distutils.  check_lib() was modified to behave
  more like AC_CHECK_LIB by add other_libraries() as a parameter.  The
  test for whether installed modules are on sys.path was changed to
  use both normcase() and normpath().

- Several minor bugs were fixed in the xml package (the minidom,
  pulldom, expatreader, and saxutils modules).

- The regression test driver (regrtest.py) behavior when invoked with
  -l changed: It now reports a count of objects that are recognized as
  garbage but not freed by the garbage collector.

- The regression test for the math module was changed to test
  exceptional behavior when the test is run in verbose mode.  Python
  cannot yet guarantee consistent exception behavior across platforms,
  so the exception part of test_math is run only in verbose mode, and
  may fail on your platform.

Internals

- PyOS_CheckStack() has been disabled on Win64, where it caused
  test_sre to fail.

Build issues

- Changed compiler flags, so that gcc is always invoked with -Wall and
  -Wstrict-prototypes.  Users compiling Python with GCC should see
  exactly one warning, except if they have passed configure the
  --with-pydebug flag.  The expected warning is for getopt() in
  Modules/main.c.  This warning will be fixed for Python 2.1.

- Fixed configure to add -threads argument during linking on OSF1.

Tools and other miscellany

- The compiler in Tools/compiler was updated to support the new
  language features introduced in 2.0: extended print statement, list
  comprehensions, and augmented assignments.  The new compiler should
  also be backwards compatible with Python 1.5.2; the compiler will
  always generate code for the version of the interpreter it runs
  under.

What's new in 2.0 release candidate 1 (since beta 2)?
=====================================================

What is release candidate 1?

We believe that release candidate 1 will fix all known bugs that we
intend to fix for the 2.0 final release.  This release should be a bit
more stable than the previous betas.  We would like to see even more
widespread testing before the final release, so we are producing this
release candidate.  The final release will be exactly the same unless
any show-stopping (or brown bag) bugs are found by testers of the
release candidate.

All the changes since the last beta release are bug fixes or changes
to support building Python for specific platforms.

Core language, builtins, and interpreter

- A bug that caused crashes when __coerce__ was used with augmented
  assignment, e.g. +=, was fixed.

- Raise ZeroDivisionError when raising zero to a negative number,
  e.g. 0.0 ** -2.0.  Note that math.pow is unrelated to the builtin
  power operator and the result of math.pow(0.0, -2.0) will vary by
  platform.  On Linux, it raises a ValueError.

- A bug in Unicode string interpolation was fixed that occasionally
  caused errors with formats including "%%".  For example, the
  following expression "%% %s" % u"abc" no longer raises a TypeError.

- Compilation of deeply nested expressions raises MemoryError instead
  of SyntaxError, e.g. eval("[" * 50 + "]" * 50).

- In 2.0b2 on Windows, the interpreter wrote .pyc files in text mode,
  rendering them useless.  They are now written in binary mode again.

Standard library

- Keyword arguments are now accepted for most pattern and match object
  methods in SRE, the standard regular expression engine.

- In SRE, fixed error with negative lookahead and lookbehind that
  manifested itself as a runtime error in patterns like "(?<!abc)(def)".

- Several bugs in the Unicode handling and error handling in _tkinter
  were fixed.

- Fix memory management errors in Merge() and Tkapp_Call() routines.

- Several changes were made to cStringIO to make it compatible with
  the file-like object interface and with StringIO.  If operations are
  performed on a closed object, an exception is raised.  The truncate
  method now accepts a position argument and readline accepts a size
  argument.

- There were many changes made to the linuxaudiodev module and its
  test suite; as a result, a short, unexpected audio sample should now
  play when the regression test is run.

  Note that this module is named poorly, because it should work
  correctly on any platform that supports the Open Sound System
  (OSS).

  The module now raises exceptions when errors occur instead of
  crashing.  It also defines the AFMT_A_LAW format (logarithmic A-law
  audio) and defines a getptr() method that calls the
  SNDCTL_DSP_GETxPTR ioctl defined in the OSS Programmer's Guide.

- The library_version attribute, introduced in an earlier beta, was
  removed because it can not be supported with early versions of the C
  readline library, which provides no way to determine the version at
  compile-time.

- The binascii module is now enabled on Win64.

- tokenize.py no longer suffers "recursion depth" errors when parsing
  programs with very long string literals.

Internals

- Fixed several buffer overflow vulnerabilities in calculate_path(),
  which is called when the interpreter starts up to determine where
  the standard library is installed.  These vulnerabilities affect all
  previous versions of Python and can be exploited by setting very
  long values for PYTHONHOME or argv[0].  The risk is greatest for a
  setuid Python script, although use of the wrapper in
  Misc/setuid-prog.c will eliminate the vulnerability.

- Fixed garbage collection bugs in instance creation that were
  triggered when errors occurred during initialization.  The solution,
  applied in cPickle and in PyInstance_New(), is to call
  PyObject_GC_Init() after the initialization of the object's
  container attributes is complete.

- pyexpat adds definitions of PyModule_AddStringConstant and
  PyModule_AddObject if the Python version is less than 2.0, which
  provides compatibility with PyXML on Python 1.5.2.

- If the platform has a bogus definition for LONG_BIT (the number of
  bits in a long), an error will be reported at compile time.

- Fix bugs in _PyTuple_Resize() which caused hard-to-interpret garbage
  collection crashes and possibly other, unreported crashes.

- Fixed a memory leak in _PyUnicode_Fini().

Build issues

- configure now accepts a --with-suffix option that specifies the
  executable suffix.  This is useful for builds on Cygwin and Mac OS
  X, for example.

- The mmap.PAGESIZE constant is now initialized using sysconf when
  possible, which eliminates a dependency on -lucb for Reliant UNIX.

- The md5 file should now compile on all platforms.

- The select module now compiles on platforms that do not define
  POLLRDNORM and related constants.

- Darwin (Mac OS X):  Initial support for static builds on this
  platform.

- BeOS: A number of changes were made to the build and installation
  process.  ar-fake now operates on a directory of object files.
  dl_export.h is gone, and its macros now appear on the mwcc command
  line during build on PPC BeOS.

- Platform directory in lib/python2.0 is "plat-beos5" (or
  "plat-beos4", if building on BeOS 4.5), rather than "plat-beos".

- Cygwin: Support for shared libraries, Tkinter, and sockets.

- SunOS 4.1.4_JL: Fix test for directory existence in configure.

Tools and other miscellany

- Removed debugging prints from main used with freeze.

- IDLE auto-indent no longer crashes when it encounters Unicode
  characters.

What's new in 2.0 beta 2 (since beta 1)?
========================================

Core language, builtins, and interpreter

- Add support for unbounded ints in %d,i,u,x,X,o formats; for example
  "%d" % 2L**64 == "18446744073709551616".

- Add -h and -V command line options to print the usage message and
  Python version number and exit immediately.

- eval() and exec accept Unicode objects as code parameters.

- getattr() and setattr() now also accept Unicode objects for the
  attribute name, which are converted to strings using the default
  encoding before lookup.

- Multiplication on string and Unicode now does proper bounds
  checking; e.g. 'a' * 65536 * 65536 will raise ValueError, "repeated
  string is too long."

- Better error message when continue is found in try statement in a
  loop.


Standard library and extensions

- array: reverse() method of array now works.  buffer_info() now does
  argument checking; it still takes no arguments.

- asyncore/asynchat: Included most recent version from Sam Rushing.

- cgi: Accept '&' or ';' as separator characters when parsing form data.

- CGIHTTPServer: Now works on Windows (and perhaps even Mac).

- ConfigParser: When reading the file, options spelled in upper case
  letters are now correctly converted to lowercase.

- copy: Copy Unicode objects atomically.

- cPickle: Fail gracefully when copy_reg can't be imported.

- cStringIO: Implemented readlines() method.

- dbm: Add get() and setdefault() methods to dbm object.  Add constant
  `library' to module that names the library used.  Added doc strings
  and method names to error messages.  Uses configure to determine
  which ndbm.h file to include; Berkeley DB's nbdm and GDBM's ndbm is
  now available options.

- distutils: Update to version 0.9.3.

- dl: Add several dl.RTLD_ constants.

- fpectl: Now supported on FreeBSD.

- gc: Add DEBUG_SAVEALL option.  When enabled all garbage objects
  found by the collector will be saved in gc.garbage.  This is useful
  for debugging a program that creates reference cycles.

- httplib: Three changes: Restore support for set_debuglevel feature
  of HTTP class.  Do not close socket on zero-length response.  Do not
  crash when server sends invalid content-length header.

- mailbox: Mailbox class conforms better to qmail specifications.

- marshal: When reading a short, sign-extend on platforms where shorts
  are bigger than 16 bits.  When reading a long, repair the unportable
  sign extension that was being done for 64-bit machines.  (It assumed
  that signed right shift sign-extends.)

- operator: Add contains(), invert(), __invert__() as aliases for
  __contains__(), inv(), and __inv__() respectively.

- os: Add support for popen2() and popen3() on all platforms where
  fork() exists.  (popen4() is still in the works.)

- os: (Windows only:) Add startfile() function that acts like double-
  clicking on a file in Explorer (or passing the file name to the
  DOS "start" command).

- os.path: (Windows, DOS:) Treat trailing colon correctly in
  os.path.join.  os.path.join("a:", "b") yields "a:b".

- pickle: Now raises ValueError when an invalid pickle that contains
  a non-string repr where a string repr was expected.  This behavior
  matches cPickle.

- posixfile: Remove broken __del__() method.

- py_compile: support CR+LF line terminators in source file.

- readline: Does not immediately exit when ^C is hit when readline and
  threads are configured.  Adds definition of rl_library_version.  (The
  latter addition requires GNU readline 2.2 or later.)

- rfc822: Domain literals returned by AddrlistClass method
  getdomainliteral() are now properly wrapped in brackets.

- site: sys.setdefaultencoding() should only be called in case the
  standard default encoding ("ascii") is changed. This saves quite a
  few cycles during startup since the first call to
  setdefaultencoding() will initialize the codec registry and the
  encodings package.

- socket: Support for size hint in readlines() method of object returned
  by makefile().

- sre: Added experimental expand() method to match objects.  Does not
  use buffer interface on Unicode strings.  Does not hang if group id
  is followed by whitespace.

- StringIO: Size hint in readlines() is now supported as documented.

- struct: Check ranges for bytes and shorts.

- urllib: Improved handling of win32 proxy settings. Fixed quote and
  quote_plus functions so that the always encode a comma.

- Tkinter: Image objects are now guaranteed to have unique ids.  Set
  event.delta to zero if Tk version doesn't support mousewheel.
  Removed some debugging prints.

- UserList: now implements __contains__().

- webbrowser: On Windows, use os.startfile() instead of os.popen(),
  which works around a bug in Norton AntiVirus 2000 that leads directly
  to a Blue Screen freeze.

- xml: New version detection code allows PyXML to override standard
  XML package if PyXML version is greater than 0.6.1.

- xml.dom: DOM level 1 support for basic XML.  Includes xml.dom.minidom
  (conventional DOM), and xml.dom.pulldom, which allows building the DOM
  tree only for nodes which are sufficiently interesting to a specific
  application.  Does not provide the HTML-specific extensions.  Still
  undocumented.

- xml.sax: SAX 2 support for Python, including all the handler
  interfaces needed to process XML 1.0 compliant XML.  Some
  documentation is already available.

- pyexpat: Renamed to xml.parsers.expat since this is part of the new,
  packagized XML support.


C API

- Add three new convenience functions for module initialization --
  PyModule_AddObject(), PyModule_AddIntConstant(), and
  PyModule_AddStringConstant().

- Cleaned up definition of NULL in C source code; all definitions were
  removed and add #error to Python.h if NULL isn't defined after
  #include of stdio.h.

- Py_PROTO() macros that were removed in 2.0b1 have been restored for
  backwards compatibility (at the source level) with old extensions.

- A wrapper API was added for signal() and sigaction().  Instead of
  either function, always use PyOS_getsig() to get a signal handler
  and PyOS_setsig() to set one.  A new convenience typedef
  PyOS_sighandler_t is defined for the type of signal handlers.

- Add PyString_AsStringAndSize() function that provides access to the
  internal data buffer and size of a string object -- or the default
  encoded version of a Unicode object.

- PyString_Size() and PyString_AsString() accept Unicode objects.

- The standard header <limits.h> is now included by Python.h (if it
  exists).  INT_MAX and LONG_MAX will always be defined, even if
  <limits.h> is not available.

- PyFloat_FromString takes a second argument, pend, that was
  effectively useless.  It is now officially useless but preserved for
  backwards compatibility.  If the pend argument is not NULL, *pend is
  set to NULL.

- PyObject_GetAttr() and PyObject_SetAttr() now accept Unicode objects
  for the attribute name.  See note on getattr() above.

- A few bug fixes to argument processing for Unicode.
  PyArg_ParseTupleAndKeywords() now accepts "es#" and "es".
  PyArg_Parse() special cases "s#" for Unicode objects; it returns a
  pointer to the default encoded string data instead of to the raw
  UTF-16.

- Py_BuildValue accepts B format (for bgen-generated code).


Internals

- On Unix, fix code for finding Python installation directory so that
  it works when argv[0] is a relative path.

- Added a true unicode_internal_encode() function and fixed the
  unicode_internal_decode function() to support Unicode objects directly
  rather than by generating a copy of the object.

- Several of the internal Unicode tables are much smaller now, and
  the source code should be much friendlier to weaker compilers.

- In the garbage collector: Fixed bug in collection of tuples.  Fixed
  bug that caused some instances to be removed from the container set
  while they were still live.  Fixed parsing in gc.set_debug() for
  platforms where sizeof(long) > sizeof(int).

- Fixed refcount problem in instance deallocation that only occurred
  when Py_REF_DEBUG was defined and Py_TRACE_REFS was not.

- On Windows, getpythonregpath is now protected against null data in
  registry key.

- On Unix, create .pyc/.pyo files with O_EXCL flag to avoid a race
  condition.


Build and platform-specific issues

- Better support of GNU Pth via --with-pth configure option.

- Python/C API now properly exposed to dynamically-loaded extension
  modules on Reliant UNIX.

- Changes for the benefit of SunOS 4.1.4 (really!).  mmapmodule.c:
  Don't define MS_SYNC to be zero when it is undefined.  Added missing
  prototypes in posixmodule.c.

- Improved support for HP-UX build.  Threads should now be correctly
  configured (on HP-UX 10.20 and 11.00).

- Fix largefile support on older NetBSD systems and OpenBSD by adding
  define for TELL64.


Tools and other miscellany

- ftpmirror: Call to main() is wrapped in if __name__ == "__main__".

- freeze: The modulefinder now works with 2.0 opcodes.

- IDLE:
  Move hackery of sys.argv until after the Tk instance has been
  created, which allows the application-specific Tkinter
  initialization to be executed if present; also pass an explicit
  className parameter to the Tk() constructor.


What's new in 2.0 beta 1?
=========================

Source Incompatibilities
------------------------

None.  Note that 1.6 introduced several incompatibilities with 1.5.2,
such as single-argument append(), connect() and bind(), and changes to
str(long) and repr(float).


Binary Incompatibilities
------------------------

- Third party extensions built for Python 1.5.x or 1.6 cannot be used
with Python 2.0; these extensions will have to be rebuilt for Python
2.0.

- On Windows, attempting to import a third party extension built for
Python 1.5.x or 1.6 results in an immediate crash; there's not much we
can do about this.  Check your PYTHONPATH environment variable!

- Python bytecode files (*.pyc and *.pyo) are not compatible between
releases.


Overview of Changes Since 1.6
-----------------------------

There are many new modules (including brand new XML support through
the xml package, and i18n support through the gettext module); a list
of all new modules is included below.  Lots of bugs have been fixed.

The process for making major new changes to the language has changed
since Python 1.6.  Enhancements must now be documented by a Python
Enhancement Proposal (PEP) before they can be accepted.

There are several important syntax enhancements, described in more
detail below:

  - Augmented assignment, e.g. x += 1

  - List comprehensions, e.g. [x**2 for x in range(10)]

  - Extended import statement, e.g. import Module as Name

  - Extended print statement, e.g. print >> file, "Hello"

Other important changes:

  - Optional collection of cyclical garbage

Python Enhancement Proposal (PEP)
---------------------------------

PEP stands for Python Enhancement Proposal.  A PEP is a design
document providing information to the Python community, or describing
a new feature for Python.  The PEP should provide a concise technical
specification of the feature and a rationale for the feature.

We intend PEPs to be the primary mechanisms for proposing new
features, for collecting community input on an issue, and for
documenting the design decisions that have gone into Python.  The PEP
author is responsible for building consensus within the community and
documenting dissenting opinions.

The PEPs are available at http://python.sourceforge.net/peps/.

Augmented Assignment
--------------------

This must have been the most-requested feature of the past years!
Eleven new assignment operators were added:

    += -= *= /= %= **= <<= >>= &= ^= |=

For example,

    A += B

is similar to

    A = A + B

except that A is evaluated only once (relevant when A is something
like dict[index].attr).

However, if A is a mutable object, A may be modified in place.  Thus,
if A is a number or a string, A += B has the same effect as A = A+B
(except A is only evaluated once); but if a is a list, A += B has the
same effect as A.extend(B)!

Classes and built-in object types can override the new operators in
order to implement the in-place behavior; the not-in-place behavior is
used automatically as a fallback when an object doesn't implement the
in-place behavior.  For classes, the method name is derived from the
method name for the corresponding not-in-place operator by inserting
an 'i' in front of the name, e.g. __iadd__ implements in-place
__add__.

Augmented assignment was implemented by Thomas Wouters.


List Comprehensions
-------------------

This is a flexible new notation for lists whose elements are computed
from another list (or lists).  The simplest form is:

    [<expression> for <variable> in <sequence>]

For example, [i**2 for i in range(4)] yields the list [0, 1, 4, 9].
This is more efficient than a for loop with a list.append() call.

You can also add a condition:

    [<expression> for <variable> in <sequence> if <condition>]

For example, [w for w in words if w == w.lower()] would yield the list
of words that contain no uppercase characters.  This is more efficient
than a for loop with an if statement and a list.append() call.

You can also have nested for loops and more than one 'if' clause.  For
example, here's a function that flattens a sequence of sequences::

    def flatten(seq):
        return [x for subseq in seq for x in subseq]

    flatten([[0], [1,2,3], [4,5], [6,7,8,9], []])

This prints

    [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

List comprehensions originated as a patch set from Greg Ewing; Skip
Montanaro and Thomas Wouters also contributed.  Described by PEP 202.


Extended Import Statement
-------------------------

Many people have asked for a way to import a module under a different
name.  This can be accomplished like this:

    import foo
    bar = foo
    del foo

but this common idiom gets old quickly.  A simple extension of the
import statement now allows this to be written as follows:

    import foo as bar

There's also a variant for 'from ... import':

    from foo import bar as spam

This also works with packages; e.g. you can write this:

    import test.regrtest as regrtest

Note that 'as' is not a new keyword -- it is recognized only in this
context (this is only possible because the syntax for the import
statement doesn't involve expressions).

Implemented by Thomas Wouters.  Described by PEP 221.


Extended Print Statement
------------------------

Easily the most controversial new feature, this extension to the print
statement adds an option to make the output go to a different file
than the default sys.stdout.

For example, to write an error message to sys.stderr, you can now
write:

    print >> sys.stderr, "Error: bad dog!"

As a special feature, if the expression used to indicate the file
evaluates to None, the current value of sys.stdout is used.  Thus:

    print >> None, "Hello world"

is equivalent to

    print "Hello world"

Design and implementation by Barry Warsaw.  Described by PEP 214.


Optional Collection of Cyclical Garbage
---------------------------------------

Python is now equipped with a garbage collector that can hunt down
cyclical references between Python objects.  It's no replacement for
reference counting; in fact, it depends on the reference counts being
correct, and decides that a set of objects belong to a cycle if all
their reference counts can be accounted for from their references to
each other.  This devious scheme was first proposed by Eric Tiedemann,
and brought to implementation by Neil Schemenauer.

There's a module "gc" that lets you control some parameters of the
garbage collection.  There's also an option to the configure script
that lets you enable or disable the garbage collection.  In 2.0b1,
it's on by default, so that we (hopefully) can collect decent user
experience with this new feature.  There are some questions about its
performance.  If it proves to be too much of a problem, we'll turn it
off by default in the final 2.0 release.


Smaller Changes
---------------

A new function zip() was added.  zip(seq1, seq2, ...) is equivalent to
map(None, seq1, seq2, ...) when the sequences have the same length;
i.e. zip([1,2,3], [10,20,30]) returns [(1,10), (2,20), (3,30)].  When
the lists are not all the same length, the shortest list wins:
zip([1,2,3], [10,20]) returns [(1,10), (2,20)].  See PEP 201.

sys.version_info is a tuple (major, minor, micro, level, serial).

Dictionaries have an odd new method, setdefault(key, default).
dict.setdefault(key, default) returns dict[key] if it exists; if not,
it sets dict[key] to default and returns that value.  Thus:

    dict.setdefault(key, []).append(item)

does the same work as this common idiom:

    if not dict.has_key(key):
        dict[key] = []
    dict[key].append(item)

There are two new variants of SyntaxError that are raised for
indentation-related errors: IndentationError and TabError.

Changed \x to consume exactly two hex digits; see PEP 223.  Added \U
escape that consumes exactly eight hex digits.

The limits on the size of expressions and file in Python source code
have been raised from 2**16 to 2**32.  Previous versions of Python
were limited because the maximum argument size the Python VM accepted
was 2**16.  This limited the size of object constructor expressions,
e.g. [1,2,3] or {'a':1, 'b':2}, and the size of source files.  This
limit was raised thanks to a patch by Charles Waldman that effectively
fixes the problem.  It is now much more likely that you will be
limited by available memory than by an arbitrary limit in Python.

The interpreter's maximum recursion depth can be modified by Python
programs using sys.getrecursionlimit and sys.setrecursionlimit.  This
limit is the maximum number of recursive calls that can be made by
Python code.  The limit exists to prevent infinite recursion from
overflowing the C stack and causing a core dump.  The default value is
1000.  The maximum safe value for a particular platform can be found
by running Misc/find_recursionlimit.py.

New Modules and Packages
------------------------

atexit - for registering functions to be called when Python exits.

imputil - Greg Stein's alternative API for writing custom import
hooks.

pyexpat - an interface to the Expat XML parser, contributed by Paul
Prescod.

xml - a new package with XML support code organized (so far) in three
subpackages: xml.dom, xml.sax, and xml.parsers.  Describing these
would fill a volume.  There's a special feature whereby a
user-installed package named _xmlplus overrides the standard
xmlpackage; this is intended to give the XML SIG a hook to distribute
backwards-compatible updates to the standard xml package.

webbrowser - a platform-independent API to launch a web browser.


Changed Modules
---------------

array -- new methods for array objects: count, extend, index, pop, and
remove

binascii -- new functions b2a_hex and a2b_hex that convert between
binary data and its hex representation

calendar -- Many new functions that support features including control
over which day of the week is the first day, returning strings instead
of printing them.  Also new symbolic constants for days of week,
e.g. MONDAY, ..., SUNDAY.

cgi -- FieldStorage objects have a getvalue method that works like a
dictionary's get method and returns the value attribute of the object.

ConfigParser -- The parser object has new methods has_option,
remove_section, remove_option, set, and write.  They allow the module
to be used for writing config files as well as reading them.

ftplib -- ntransfercmd(), transfercmd(), and retrbinary() all now
optionally support the RFC 959 REST command.

gzip -- readline and readlines now accept optional size arguments

httplib -- New interfaces and support for HTTP/1.1 by Greg Stein.  See
the module doc strings for details.

locale -- implement getdefaultlocale for Win32 and Macintosh

marshal -- no longer dumps core when marshaling deeply nested or
recursive data structures

os -- new functions isatty, seteuid, setegid, setreuid, setregid

os/popen2 -- popen2/popen3/popen4 support under Windows.  popen2/popen3
support under Unix.

os/pty -- support for openpty and forkpty

os.path -- fix semantics of os.path.commonprefix

smtplib -- support for sending very long messages

socket -- new function getfqdn()

readline -- new functions to read, write and truncate history files.
The readline section of the library reference manual contains an
example.

select -- add interface to poll system call

shutil -- new copyfileobj function

SimpleHTTPServer, CGIHTTPServer -- Fix problems with buffering in the
HTTP server.

Tkinter -- optimization of function flatten

urllib -- scans environment variables for proxy configuration,
e.g. http_proxy.

whichdb -- recognizes dumbdbm format


Obsolete Modules
----------------

None.  However note that 1.6 made a whole slew of modules obsolete:
stdwin, soundex, cml, cmpcache, dircache, dump, find, grep, packmail,
poly, zmod, strop, util, whatsound.


Changed, New, Obsolete Tools
----------------------------

None.


C-level Changes
---------------

Several cleanup jobs were carried out throughout the source code.

All C code was converted to ANSI C; we got rid of all uses of the
Py_PROTO() macro, which makes the header files a lot more readable.

Most of the portability hacks were moved to a new header file,
pyport.h; several other new header files were added and some old
header files were removed, in an attempt to create a more rational set
of header files.  (Few of these ever need to be included explicitly;
they are all included by Python.h.)

Trent Mick ensured portability to 64-bit platforms, under both Linux
and Win64, especially for the new Intel Itanium processor.  Mick also
added large file support for Linux64 and Win64.

The C APIs to return an object's size have been update to consistently
use the form PyXXX_Size, e.g. PySequence_Size and PyDict_Size.  In
previous versions, the abstract interfaces used PyXXX_Length and the
concrete interfaces used PyXXX_Size.  The old names,
e.g. PyObject_Length, are still available for backwards compatibility
at the API level, but are deprecated.

The PyOS_CheckStack function has been implemented on Windows by
Fredrik Lundh.  It prevents Python from failing with a stack overflow
on Windows.

The GC changes resulted in creation of two new slots on object,
tp_traverse and tp_clear.  The augmented assignment changes result in
the creation of a new slot for each in-place operator.

The GC API creates new requirements for container types implemented in
C extension modules.  See Include/objimpl.h for details.

PyErr_Format has been updated to automatically calculate the size of
the buffer needed to hold the formatted result string.  This change
prevents crashes caused by programmer error.

New C API calls: PyObject_AsFileDescriptor, PyErr_WriteUnraisable.

PyRun_AnyFileEx, PyRun_SimpleFileEx, PyRun_FileEx -- New functions
that are the same as their non-Ex counterparts except they take an
extra flag argument that tells them to close the file when done.

XXX There were other API changes that should be fleshed out here.


Windows Changes
---------------

New popen2/popen3/peopen4 in os module (see Changed Modules above).

os.popen is much more usable on Windows 95 and 98.  See Microsoft
Knowledge Base article Q150956.  The Win9x workaround described there
is implemented by the new w9xpopen.exe helper in the root of your
Python installation.  Note that Python uses this internally; it is not
a standalone program.

Administrator privileges are no longer required to install Python
on Windows NT or Windows 2000.  If you have administrator privileges,
Python's registry info will be written under HKEY_LOCAL_MACHINE.
Otherwise the installer backs off to writing Python's registry info
under HKEY_CURRENT_USER.  The latter is sufficient for all "normal"
uses of Python, but will prevent some advanced uses from working
(for example, running a Python script as an NT service, or possibly
from CGI).

[This was new in 1.6] The installer no longer runs a separate Tcl/Tk
installer; instead, it installs the needed Tcl/Tk files directly in the
Python directory.  If you already have a Tcl/Tk installation, this
wastes some disk space (about 4 Megs) but avoids problems with
conflicting Tcl/Tk installations, and makes it much easier for Python
to ensure that Tcl/Tk can find all its files.

[This was new in 1.6] The Windows installer now installs by default in
\Python20\ on the default volume, instead of \Program Files\Python-2.0\.


Updates to the changes between 1.5.2 and 1.6
--------------------------------------------

The 1.6 NEWS file can't be changed after the release is done, so here
is some late-breaking news:

New APIs in locale.py: normalize(), getdefaultlocale(), resetlocale(),
and changes to getlocale() and setlocale().

The new module is now enabled per default.

It is not true that the encodings codecs cannot be used for normal
strings: the string.encode() (which is also present on 8-bit strings
!) allows using them for 8-bit strings too, e.g. to convert files from
cp1252 (Windows) to latin-1 or vice-versa.

Japanese codecs are available from Tamito KAJIYAMA:
http://pseudo.grad.sccs.chukyo-u.ac.jp/~kajiyama/python/


======================================================================