Abstract

This PEP, previously known as PEP 3000, describes smaller scale changes and new features for which no separate PEP is written yet, all targeted for Python 3000.

The list of features included in this document is subject to change and isn't binding on the Python development community; features may be added, removed, and modified at any time. The purpose of this list is to focus our language development effort on changes that are steps to 3.0, and to encourage people to invent ways to smooth the transition.

This document is not a wish-list that anyone can extend. While there are two authors of this PEP, we're just supplying the text; the decisions for which changes are listed in this document are made by Guido van Rossum, who has chosen them as goals for Python 3.0.

Guido's pronouncements on things that will not change in Python 3.0 are recorded in PEP 3099. [43]

General goals

A general goal is to reduce feature duplication by removing old ways of doing things. A general principle of the design will be that one obvious way of doing something is enough. [1]

Influencing PEPs

• PEP 238 (Changing the Division Operator) [34]
• PEP 328 (Imports: Multi-Line and Absolute/Relative) [39]
• PEP 343 (The "with" Statement) [40]
• PEP 352 (Required Superclass for Exceptions) [41]

Style changes

• The C style guide will be updated to use 4-space indents, never tabs. This style should be used for all new files; existing files can be updated only if there is no hope to ever merge a particular file from the Python 2 HEAD. Within a file, the indentation style should be consistent. No other style guide changes are planned ATM.

Core language

• True division becomes default behavior [34] [done]

• exec as a statement is not worth it -- make it a function [done]

• Add optional declarations for static typing [45] [10] [done]

• Support only new-style classes; classic classes will be gone [1] [done]

• Replace print by a function [14] [44] [done]

• The softspace attribute of files goes away. [done]

• Use except E1, E2, E3 as err: if you want the error variable. [3] [done]

• None becomes a keyword [4]; also True and False [done]

• ... to become a general expression element [16] [done]

• as becomes a keyword [5] (starting in 2.6 already) [done]

• Have list comprehensions be syntactic sugar for passing an equivalent generator expression to list(); as a consequence the loop variable will no longer be exposed [36] [done]

• Comparisons other than == and != between disparate types will raise an exception unless explicitly supported by the type [6] [done]

• floats will not be acceptable as arguments in place of ints for operations where floats are inadvertantly accepted (PyArg_ParseTuple() i & l formats)

• Remove from ... import * at function scope. [done] This means that functions can always be optimized and support for unoptimized functions can go away.

• Imports [39]

  • Imports will be absolute by default. [done]
  • Relative imports must be explicitly specified. [done]
  • Indirection entries in sys.modules (i.e., a value of None for A.string means to use the top-level string module) will not be supported.

• __init__.py might become optional in sub-packages? __init__.py will still be required for top-level packages.

• Cleanup the Py_InitModule() variants {,3,4} (also import and parser APIs)

• Cleanup the APIs exported in pythonrun, etc.

• Some expressions will require parentheses that didn't in 2.x:

  • List comprehensions will require parentheses around the iterables. This will make list comprehensions more similar to generator comprehensions. [x for x in 1, 2] will need to be: [x for x in (1, 2)] [done]
  • Lambdas may have to be parenthesized [38] [NO]

• In order to get rid of the confusion between __builtin__ and __builtins__, it was decided to rename __builtin__ (the module) to builtins, and to leave __builtins__ (the sandbox hook) alone. [47] [48] [done]

• Attributes on functions of the form func_whatever will be renamed __whatever__ [17] [done]

• Set literals and comprehensions [19] [20] [done] {x} means set([x]); {x, y} means set([x, y]). {F(x) for x in S if P(x)} means set(F(x) for x in S if P(x)). NB. {range(x)} means set([range(x)]), NOT set(range(x)). There's no literal for an empty set; use set() (or {1}&{2} :-). There's no frozenset literal; they are too rarely needed.

• The __nonzero__ special method will be renamed to __bool__ and have to return a bool. The typeobject slot will be called tp_bool [23] [done]

• Dict comprehensions, as first proposed in [35] [done] {K(x): V(x) for x in S if P(x)} means dict((K(x), V(x)) for x in S if P(x)).

To be removed:

• String exceptions: use instances of an Exception class [2] [done]

• raise Exception, "message": use raise Exception("message") [12] [done]

• `x`: use repr(x) [2] [done]

• The <> operator: use != instead [3] [done]

• The __mod__ and __divmod__ special methods on float. [they should stay] [21]

• Drop unbound methods [7] [25] [done]

• METH_OLDARGS [done]

• WITH_CYCLE_GC [done]

• __getslice__, __setslice__, __delslice__ [32]; remove slice opcodes and use slice objects. [done]

• __oct__, __hex__: use __index__ in oct() and hex() instead. [done]

• __methods__ and __members__ [done]

• C APIs (see code): PyFloat_AsString, PyFloat_AsReprString, PyFloat_AsStringEx, PySequence_In, PyEval_EvalFrame, PyEval_CallObject, _PyObject_Del, _PyObject_GC_Del, _PyObject_GC_Track, _PyObject_GC_UnTrack PyString_AsEncodedString, PyString_AsDecodedString PyArg_NoArgs, PyArg_GetInt, intargfunc, intintargfunc

  PyImport_ReloadModule ?

Atomic Types

• Remove distinction between int and long types; 'long' built-in type and literals with 'L' or 'l' suffix disappear [1] [done]
• Make all strings be Unicode, and have a separate bytes() type [1] The new string type will be called 'str'. See PEP 3137. [done]
• Return iterable views instead of lists where appropriate for atomic type methods (e.g. dict.keys(), dict.values(), dict.items(), etc.); iter* methods will be removed. [done]
• Make string.join() stringify its arguments? [18] [NO]
• Fix open() so it returns a ValueError if the mode is bad rather than IOError. [done]

To be removed:

• basestring.find() and basestring.rfind(); use basestring.index() or basestring.[r]partition() or or basestring.rindex() in a try/except block??? [13] [UNLIKELY]
• file.xreadlines() method [31] [done]
• dict.setdefault()? [15] [UNLIKELY]
• dict.has_key() method; use in operator [done]
• list.sort() and builtin.sorted() methods: eliminate cmp parameter [27] [done]

Built-in Namespace

• Make built-ins return an iterator where appropriate (e.g. range(), zip(), map(), filter(), etc.) [done]
• Remove input() and rename raw_input() to input(). If you need the old input(), use eval(input()). [done]
• Introduce trunc(), which would call the __trunc__() method on its argument; suggested use is for objects like float where calling __int__() has data loss, but an integral representation is still desired? [8] [done]
• Exception hierarchy changes [41] [done]
• Add a bin() function for a binary representation of integers [done]

To be removed:

• apply(): use f(*args, **kw) instead [2] [done]

• buffer(): must die (use a bytes() type instead) (?) [2] [done]

• callable(): just use isinstance(x, collections.Callable) (?) [2] [done]

• compile(): put in sys (or perhaps in a module of its own) [2]

• coerce(): no longer needed [2] [done]

• execfile(), reload(): use exec() [2] [done]

• intern(): put in sys [2], [22] [done]

• reduce(): put in functools, a loop is more readable most of the times [2], [9] [done]

• xrange(): use range() instead [1] [See range() above] [done]

• StandardError: this is a relic from the original exception hierarchy;

  subclass Exception instead. [done]

Standard library

• Reorganize the standard library to not be as shallow?
• Move test code to where it belongs, there will be no more test() functions in the standard library
• Convert all tests to use either doctest or unittest.
• For the procedures of standard library improvement, see PEP 3001 [42]

To be removed:

• The sets module. [done]

• stdlib modules to be removed

  • see docstrings and comments in the source

    • macfs [to do]
    • new, reconvert, stringold, xmllib, pcre, pypcre, strop [all done]

  • see PEP 4 [33]

    • buildtools, mimetools, multifile, rfc822, [to do]
    • mpz, posixfile, regsub, rgbimage, sha, statcache, sv, TERMIOS, timing [done]
    • cfmfile, gopherlib, md5, MimeWriter, mimify [done]
    • cl, sets, xreadlines, rotor, whrandom [done]

  • Everything in lib-old [33] [done]

    • Para, addpack, cmp, cmpcache, codehack, dircmp, dump, find, fmt, grep, lockfile, newdir, ni, packmail, poly, rand, statcache, tb, tzparse, util, whatsound, whrandom, zmod

• sys.exitfunc: use atexit module instead [28], [49] [done]

• sys.exc_type, sys.exc_values, sys.exc_traceback: not thread-safe; use sys.exc_info() or an attribute of the exception [2] [11] [28] [done]

• sys.exc_clear: Python 3's except statements provide the same functionality [24] [46] [28] [done]

• array.read, array.write [30]

• operator.isCallable : callable() built-in is being removed [29] [50] [done]

• operator.sequenceIncludes : redundant thanks to operator.contains [29] [50] [done]

• In the thread module, the aquire_lock() and release_lock() aliases for the acquire() and release() methods on lock objects. (Probably also just remove the thread module as a public API, in favor of always using threading.py.)

• UserXyz classes, in favour of XyzMixins.

• Remove the unreliable empty() and full() methods from Queue.py?

• Remove jumpahead() from the random API?

• Make the primitive for random be something generating random bytes rather than random floats?

• Get rid of Cookie.SerialCookie and Cookie.SmartCookie?

• Modify the heapq.heapreplace() API to compare the new value to the top of the heap?

Outstanding Issues

• Require C99, so we can use // comments, named initializers, declare variables without introducing a new scope, among other benefits. (Also better support for IEEE floating point issues like NaN and infinities?)
• Remove support for old systems, including: BeOS, RISCOS, (SGI) Irix, Tru64

References

Copyright

This document has been placed in the public domain.