cpython/Lib/distutils/version.py

344 lines
12 KiB
Python
Raw Normal View History

#
# distutils/version.py
#
# Implements multiple version numbering conventions for the
# Python Module Distribution Utilities.
#
# $Id$
#
"""Provides classes to represent module version numbers (one class for
each style of version numbering). There are currently two such classes
implemented: StrictVersion and LooseVersion.
Every version number class implements the following interface:
* the 'parse' method takes a string and parses it to some internal
representation; if the string is an invalid version number,
'parse' raises a ValueError exception
* the class constructor takes an optional string argument which,
if supplied, is passed to 'parse'
* __str__ reconstructs the string that was passed to 'parse' (or
an equivalent string -- ie. one that will generate an equivalent
version number instance)
* __repr__ generates Python code to recreate the version number instance
* _cmp compares the current instance with either another instance
of the same class or a string (which will be parsed to an instance
of the same class, thus must follow the same rules)
"""
import re
class Version:
"""Abstract base class for version numbering classes. Just provides
constructor (__init__) and reproducer (__repr__), because those
Restructure comparison dramatically. There is no longer a default *ordering* between objects; there is only a default equality test (defined by an object being equal to itself only). Read the comment in object.c. The current implementation never uses a three-way comparison to compute a rich comparison, but it does use a rich comparison to compute a three-way comparison. I'm not quite done ripping out all the calls to PyObject_Compare/Cmp, or replacing tp_compare implementations with tp_richcompare implementations; but much of that has happened (to make most unit tests pass). The following tests still fail, because I need help deciding or understanding: test_codeop -- depends on comparing code objects test_datetime -- need Tim Peters' opinion test_marshal -- depends on comparing code objects test_mutants -- need help understanding it The problem with test_codeop and test_marshal is this: these tests compare two different code objects and expect them to be equal. Is that still a feature we'd like to support? I've temporarily removed the comparison and hash code from code objects, so they use the default (equality by pointer only) comparison. For the other two tests, run them to see for yourself. (There may be more failing test with "-u all".) A general problem with getting lots of these tests to pass is the reality that for object types that have a natural total ordering, implementing __cmp__ is much more convenient than implementing __eq__, __ne__, __lt__, and so on. Should we go back to allowing __cmp__ to provide a total ordering? Should we provide some other way to implement rich comparison with a single method override? Alex proposed a __key__() method; I've considered a __richcmp__() method. Or perhaps __cmp__() just shouldn't be killed off...
2006-08-23 21:41:19 -03:00
seem to be the same for all version numbering classes; and route
rich comparisons to _cmp.
"""
def __init__ (self, vstring=None):
if vstring:
self.parse(vstring)
def __repr__ (self):
return "%s ('%s')" % (self.__class__.__name__, str(self))
Restructure comparison dramatically. There is no longer a default *ordering* between objects; there is only a default equality test (defined by an object being equal to itself only). Read the comment in object.c. The current implementation never uses a three-way comparison to compute a rich comparison, but it does use a rich comparison to compute a three-way comparison. I'm not quite done ripping out all the calls to PyObject_Compare/Cmp, or replacing tp_compare implementations with tp_richcompare implementations; but much of that has happened (to make most unit tests pass). The following tests still fail, because I need help deciding or understanding: test_codeop -- depends on comparing code objects test_datetime -- need Tim Peters' opinion test_marshal -- depends on comparing code objects test_mutants -- need help understanding it The problem with test_codeop and test_marshal is this: these tests compare two different code objects and expect them to be equal. Is that still a feature we'd like to support? I've temporarily removed the comparison and hash code from code objects, so they use the default (equality by pointer only) comparison. For the other two tests, run them to see for yourself. (There may be more failing test with "-u all".) A general problem with getting lots of these tests to pass is the reality that for object types that have a natural total ordering, implementing __cmp__ is much more convenient than implementing __eq__, __ne__, __lt__, and so on. Should we go back to allowing __cmp__ to provide a total ordering? Should we provide some other way to implement rich comparison with a single method override? Alex proposed a __key__() method; I've considered a __richcmp__() method. Or perhaps __cmp__() just shouldn't be killed off...
2006-08-23 21:41:19 -03:00
def __eq__(self, other):
c = self._cmp(other)
Restructure comparison dramatically. There is no longer a default *ordering* between objects; there is only a default equality test (defined by an object being equal to itself only). Read the comment in object.c. The current implementation never uses a three-way comparison to compute a rich comparison, but it does use a rich comparison to compute a three-way comparison. I'm not quite done ripping out all the calls to PyObject_Compare/Cmp, or replacing tp_compare implementations with tp_richcompare implementations; but much of that has happened (to make most unit tests pass). The following tests still fail, because I need help deciding or understanding: test_codeop -- depends on comparing code objects test_datetime -- need Tim Peters' opinion test_marshal -- depends on comparing code objects test_mutants -- need help understanding it The problem with test_codeop and test_marshal is this: these tests compare two different code objects and expect them to be equal. Is that still a feature we'd like to support? I've temporarily removed the comparison and hash code from code objects, so they use the default (equality by pointer only) comparison. For the other two tests, run them to see for yourself. (There may be more failing test with "-u all".) A general problem with getting lots of these tests to pass is the reality that for object types that have a natural total ordering, implementing __cmp__ is much more convenient than implementing __eq__, __ne__, __lt__, and so on. Should we go back to allowing __cmp__ to provide a total ordering? Should we provide some other way to implement rich comparison with a single method override? Alex proposed a __key__() method; I've considered a __richcmp__() method. Or perhaps __cmp__() just shouldn't be killed off...
2006-08-23 21:41:19 -03:00
if c is NotImplemented:
return c
return c == 0
def __lt__(self, other):
c = self._cmp(other)
Restructure comparison dramatically. There is no longer a default *ordering* between objects; there is only a default equality test (defined by an object being equal to itself only). Read the comment in object.c. The current implementation never uses a three-way comparison to compute a rich comparison, but it does use a rich comparison to compute a three-way comparison. I'm not quite done ripping out all the calls to PyObject_Compare/Cmp, or replacing tp_compare implementations with tp_richcompare implementations; but much of that has happened (to make most unit tests pass). The following tests still fail, because I need help deciding or understanding: test_codeop -- depends on comparing code objects test_datetime -- need Tim Peters' opinion test_marshal -- depends on comparing code objects test_mutants -- need help understanding it The problem with test_codeop and test_marshal is this: these tests compare two different code objects and expect them to be equal. Is that still a feature we'd like to support? I've temporarily removed the comparison and hash code from code objects, so they use the default (equality by pointer only) comparison. For the other two tests, run them to see for yourself. (There may be more failing test with "-u all".) A general problem with getting lots of these tests to pass is the reality that for object types that have a natural total ordering, implementing __cmp__ is much more convenient than implementing __eq__, __ne__, __lt__, and so on. Should we go back to allowing __cmp__ to provide a total ordering? Should we provide some other way to implement rich comparison with a single method override? Alex proposed a __key__() method; I've considered a __richcmp__() method. Or perhaps __cmp__() just shouldn't be killed off...
2006-08-23 21:41:19 -03:00
if c is NotImplemented:
return c
return c < 0
def __le__(self, other):
c = self._cmp(other)
Restructure comparison dramatically. There is no longer a default *ordering* between objects; there is only a default equality test (defined by an object being equal to itself only). Read the comment in object.c. The current implementation never uses a three-way comparison to compute a rich comparison, but it does use a rich comparison to compute a three-way comparison. I'm not quite done ripping out all the calls to PyObject_Compare/Cmp, or replacing tp_compare implementations with tp_richcompare implementations; but much of that has happened (to make most unit tests pass). The following tests still fail, because I need help deciding or understanding: test_codeop -- depends on comparing code objects test_datetime -- need Tim Peters' opinion test_marshal -- depends on comparing code objects test_mutants -- need help understanding it The problem with test_codeop and test_marshal is this: these tests compare two different code objects and expect them to be equal. Is that still a feature we'd like to support? I've temporarily removed the comparison and hash code from code objects, so they use the default (equality by pointer only) comparison. For the other two tests, run them to see for yourself. (There may be more failing test with "-u all".) A general problem with getting lots of these tests to pass is the reality that for object types that have a natural total ordering, implementing __cmp__ is much more convenient than implementing __eq__, __ne__, __lt__, and so on. Should we go back to allowing __cmp__ to provide a total ordering? Should we provide some other way to implement rich comparison with a single method override? Alex proposed a __key__() method; I've considered a __richcmp__() method. Or perhaps __cmp__() just shouldn't be killed off...
2006-08-23 21:41:19 -03:00
if c is NotImplemented:
return c
return c <= 0
def __gt__(self, other):
c = self._cmp(other)
Restructure comparison dramatically. There is no longer a default *ordering* between objects; there is only a default equality test (defined by an object being equal to itself only). Read the comment in object.c. The current implementation never uses a three-way comparison to compute a rich comparison, but it does use a rich comparison to compute a three-way comparison. I'm not quite done ripping out all the calls to PyObject_Compare/Cmp, or replacing tp_compare implementations with tp_richcompare implementations; but much of that has happened (to make most unit tests pass). The following tests still fail, because I need help deciding or understanding: test_codeop -- depends on comparing code objects test_datetime -- need Tim Peters' opinion test_marshal -- depends on comparing code objects test_mutants -- need help understanding it The problem with test_codeop and test_marshal is this: these tests compare two different code objects and expect them to be equal. Is that still a feature we'd like to support? I've temporarily removed the comparison and hash code from code objects, so they use the default (equality by pointer only) comparison. For the other two tests, run them to see for yourself. (There may be more failing test with "-u all".) A general problem with getting lots of these tests to pass is the reality that for object types that have a natural total ordering, implementing __cmp__ is much more convenient than implementing __eq__, __ne__, __lt__, and so on. Should we go back to allowing __cmp__ to provide a total ordering? Should we provide some other way to implement rich comparison with a single method override? Alex proposed a __key__() method; I've considered a __richcmp__() method. Or perhaps __cmp__() just shouldn't be killed off...
2006-08-23 21:41:19 -03:00
if c is NotImplemented:
return c
return c > 0
def __ge__(self, other):
c = self._cmp(other)
Restructure comparison dramatically. There is no longer a default *ordering* between objects; there is only a default equality test (defined by an object being equal to itself only). Read the comment in object.c. The current implementation never uses a three-way comparison to compute a rich comparison, but it does use a rich comparison to compute a three-way comparison. I'm not quite done ripping out all the calls to PyObject_Compare/Cmp, or replacing tp_compare implementations with tp_richcompare implementations; but much of that has happened (to make most unit tests pass). The following tests still fail, because I need help deciding or understanding: test_codeop -- depends on comparing code objects test_datetime -- need Tim Peters' opinion test_marshal -- depends on comparing code objects test_mutants -- need help understanding it The problem with test_codeop and test_marshal is this: these tests compare two different code objects and expect them to be equal. Is that still a feature we'd like to support? I've temporarily removed the comparison and hash code from code objects, so they use the default (equality by pointer only) comparison. For the other two tests, run them to see for yourself. (There may be more failing test with "-u all".) A general problem with getting lots of these tests to pass is the reality that for object types that have a natural total ordering, implementing __cmp__ is much more convenient than implementing __eq__, __ne__, __lt__, and so on. Should we go back to allowing __cmp__ to provide a total ordering? Should we provide some other way to implement rich comparison with a single method override? Alex proposed a __key__() method; I've considered a __richcmp__() method. Or perhaps __cmp__() just shouldn't be killed off...
2006-08-23 21:41:19 -03:00
if c is NotImplemented:
return c
return c >= 0
# Interface for version-number classes -- must be implemented
# by the following classes (the concrete ones -- Version should
# be treated as an abstract class).
# __init__ (string) - create and take same action as 'parse'
# (string parameter is optional)
# parse (string) - convert a string representation to whatever
# internal representation is appropriate for
# this style of version numbering
# __str__ (self) - convert back to a string; should be very similar
# (if not identical to) the string supplied to parse
# __repr__ (self) - generate Python code to recreate
# the instance
# _cmp (self, other) - compare two version numbers ('other' may
# be an unparsed version string, or another
# instance of your version class)
class StrictVersion (Version):
"""Version numbering for anal retentives and software idealists.
Implements the standard interface for version number classes as
described above. A version number consists of two or three
dot-separated numeric components, with an optional "pre-release" tag
on the end. The pre-release tag consists of the letter 'a' or 'b'
followed by a number. If the numeric components of two version
numbers are equal, then one with a pre-release tag will always
be deemed earlier (lesser) than one without.
The following are valid version numbers (shown in the order that
would be obtained by sorting according to the supplied cmp function):
0.4 0.4.0 (these two are equivalent)
0.4.1
0.5a1
0.5b3
0.5
0.9.6
1.0
1.0.4a3
1.0.4b1
1.0.4
The following are examples of invalid version numbers:
1
2.7.2.2
1.3.a4
1.3pl1
1.3c4
The rationale for this version numbering system will be explained
in the distutils documentation.
"""
2001-12-06 16:51:35 -04:00
version_re = re.compile(r'^(\d+) \. (\d+) (\. (\d+))? ([ab](\d+))?$',
re.VERBOSE | re.ASCII)
def parse (self, vstring):
match = self.version_re.match(vstring)
if not match:
raise ValueError("invalid version number '%s'" % vstring)
(major, minor, patch, prerelease, prerelease_num) = \
match.group(1, 2, 4, 5, 6)
if patch:
self.version = tuple(map(int, [major, minor, patch]))
else:
Merged revisions 56125-56153 via svnmerge from svn+ssh://pythondev@svn.python.org/python/branches/p3yk ........ r56127 | georg.brandl | 2007-06-30 09:32:49 +0200 (Sat, 30 Jun 2007) | 2 lines Fix a place where floor division would be in order. ........ r56135 | guido.van.rossum | 2007-07-01 06:13:54 +0200 (Sun, 01 Jul 2007) | 28 lines Make map() and filter() identical to itertools.imap() and .ifilter(), respectively. I fixed two bootstrap issues, due to the dynamic import of itertools: 1. Starting python requires that map() and filter() are not used until site.py has added build/lib.<arch> to sys.path. 2. Building python requires that setup.py and distutils and everything they use is free of map() and filter() calls. Beyond this, I only fixed the tests in test_builtin.py. Others, please help fixing the remaining tests that are now broken! The fixes are usually simple: a. map(None, X) -> list(X) b. map(F, X) -> list(map(F, X)) c. map(lambda x: F(x), X) -> [F(x) for x in X] d. filter(F, X) -> list(filter(F, X)) e. filter(lambda x: P(x), X) -> [x for x in X if P(x)] Someone, please also contribute a fixer for 2to3 to do this. It can leave map()/filter() calls alone that are already inside a list() or sorted() call or for-loop. Only in rare cases have I seen code that depends on map() of lists of different lengths going to the end of the longest, or on filter() of a string or tuple returning an object of the same type; these will need more thought to fix. ........ r56136 | guido.van.rossum | 2007-07-01 06:22:01 +0200 (Sun, 01 Jul 2007) | 3 lines Make it so that test_decimal fails instead of hangs, to help automated test runners. ........ r56139 | georg.brandl | 2007-07-01 18:20:58 +0200 (Sun, 01 Jul 2007) | 2 lines Fix a few test cases after the map->imap change. ........ r56142 | neal.norwitz | 2007-07-02 06:38:12 +0200 (Mon, 02 Jul 2007) | 1 line Get a bunch more tests passing after converting map/filter to return iterators. ........ r56147 | guido.van.rossum | 2007-07-02 15:32:02 +0200 (Mon, 02 Jul 2007) | 4 lines Fix the remaining failing unit tests (at least on OSX). Also tweaked urllib2 so it doesn't raise socket.gaierror when all network interfaces are turned off. ........
2007-07-03 05:25:58 -03:00
self.version = tuple(map(int, [major, minor])) + (0,)
if prerelease:
self.prerelease = (prerelease[0], int(prerelease_num))
else:
self.prerelease = None
def __str__ (self):
2001-12-06 16:51:35 -04:00
if self.version[2] == 0:
vstring = '.'.join(map(str, self.version[0:2]))
else:
vstring = '.'.join(map(str, self.version))
if self.prerelease:
vstring = vstring + self.prerelease[0] + str(self.prerelease[1])
return vstring
2001-12-06 16:51:35 -04:00
def _cmp (self, other):
if isinstance(other, str):
other = StrictVersion(other)
if self.version != other.version:
# numeric versions don't match
# prerelease stuff doesn't matter
if self.version < other.version:
return -1
else:
return 1
# have to compare prerelease
# case 1: neither has prerelease; they're equal
# case 2: self has prerelease, other doesn't; other is greater
# case 3: self doesn't have prerelease, other does: self is greater
# case 4: both have prerelease: must compare them!
if (not self.prerelease and not other.prerelease):
return 0
elif (self.prerelease and not other.prerelease):
return -1
elif (not self.prerelease and other.prerelease):
return 1
elif (self.prerelease and other.prerelease):
if self.prerelease == other.prerelease:
return 0
elif self.prerelease < other.prerelease:
return -1
else:
return 1
else:
assert False, "never get here"
# end class StrictVersion
# The rules according to Greg Stein:
Merged revisions 70578,70599,70641-70642,70650,70660-70661,70674,70691,70697-70698,70700,70704 via svnmerge from svn+ssh://pythondev@svn.python.org/python/trunk ........ r70578 | benjamin.peterson | 2009-03-23 22:24:56 -0500 (Mon, 23 Mar 2009) | 1 line this is better written using assertRaises ........ r70599 | benjamin.peterson | 2009-03-25 16:42:51 -0500 (Wed, 25 Mar 2009) | 1 line this can be slightly less ugly ........ r70641 | guilherme.polo | 2009-03-27 16:43:08 -0500 (Fri, 27 Mar 2009) | 3 lines Adjusted _tkinter to compile without warnings when WITH_THREAD is not defined (part of issue #5035) ........ r70642 | georg.brandl | 2009-03-27 19:48:48 -0500 (Fri, 27 Mar 2009) | 1 line Fix typo. ........ r70650 | benjamin.peterson | 2009-03-28 14:16:10 -0500 (Sat, 28 Mar 2009) | 1 line give os.symlink and os.link() better parameter names #5564 ........ r70660 | georg.brandl | 2009-03-28 14:52:58 -0500 (Sat, 28 Mar 2009) | 1 line Switch to fixed Sphinx version. ........ r70661 | georg.brandl | 2009-03-28 14:57:36 -0500 (Sat, 28 Mar 2009) | 2 lines Add section numbering to some of the larger subdocuments. ........ r70674 | guilherme.polo | 2009-03-29 05:19:05 -0500 (Sun, 29 Mar 2009) | 1 line Typo fix. ........ r70691 | raymond.hettinger | 2009-03-29 13:51:11 -0500 (Sun, 29 Mar 2009) | 1 line Make life easier for non-CPython implementations. ........ r70697 | benjamin.peterson | 2009-03-29 16:22:35 -0500 (Sun, 29 Mar 2009) | 1 line this has been fixed since 2.6 (I love removing these) ........ r70698 | benjamin.peterson | 2009-03-29 16:31:05 -0500 (Sun, 29 Mar 2009) | 1 line thanks to guido's bytecode verifier, this is fixed ........ r70700 | benjamin.peterson | 2009-03-29 16:50:14 -0500 (Sun, 29 Mar 2009) | 1 line use the awesome new status iterator ........ r70704 | benjamin.peterson | 2009-03-29 21:49:32 -0500 (Sun, 29 Mar 2009) | 1 line there's actually three methods here #5600 ........
2009-03-30 11:51:56 -03:00
# 1) a version number has 1 or more numbers separated by a period or by
# sequences of letters. If only periods, then these are compared
# left-to-right to determine an ordering.
# 2) sequences of letters are part of the tuple for comparison and are
# compared lexicographically
# 3) recognize the numeric components may have leading zeroes
2001-12-06 16:51:35 -04:00
#
# The LooseVersion class below implements these rules: a version number
# string is split up into a tuple of integer and string components, and
# comparison is a simple tuple comparison. This means that version
# numbers behave in a predictable and obvious way, but a way that might
# not necessarily be how people *want* version numbers to behave. There
# wouldn't be a problem if people could stick to purely numeric version
# numbers: just split on period and compare the numbers as tuples.
# However, people insist on putting letters into their version numbers;
# the most common purpose seems to be:
# - indicating a "pre-release" version
# ('alpha', 'beta', 'a', 'b', 'pre', 'p')
# - indicating a post-release patch ('p', 'pl', 'patch')
# but of course this can't cover all version number schemes, and there's
# no way to know what a programmer means without asking him.
2001-12-06 16:51:35 -04:00
#
# The problem is what to do with letters (and other non-numeric
# characters) in a version number. The current implementation does the
# obvious and predictable thing: keep them as strings and compare
# lexically within a tuple comparison. This has the desired effect if
# an appended letter sequence implies something "post-release":
# eg. "0.99" < "0.99pl14" < "1.0", and "5.001" < "5.001m" < "5.002".
#
# However, if letters in a version number imply a pre-release version,
# the "obvious" thing isn't correct. Eg. you would expect that
# "1.5.1" < "1.5.2a2" < "1.5.2", but under the tuple/lexical comparison
# implemented here, this just isn't so.
#
# Two possible solutions come to mind. The first is to tie the
# comparison algorithm to a particular set of semantic rules, as has
# been done in the StrictVersion class above. This works great as long
# as everyone can go along with bondage and discipline. Hopefully a
# (large) subset of Python module programmers will agree that the
# particular flavour of bondage and discipline provided by StrictVersion
# provides enough benefit to be worth using, and will submit their
# version numbering scheme to its domination. The free-thinking
# anarchists in the lot will never give in, though, and something needs
2000-06-28 11:48:01 -03:00
# to be done to accommodate them.
2001-12-06 16:51:35 -04:00
#
# Perhaps a "moderately strict" version class could be implemented that
# lets almost anything slide (syntactically), and makes some heuristic
# assumptions about non-digits in version number strings. This could
# sink into special-case-hell, though; if I was as talented and
# idiosyncratic as Larry Wall, I'd go ahead and implement a class that
# somehow knows that "1.2.1" < "1.2.2a2" < "1.2.2" < "1.2.2pl3", and is
# just as happy dealing with things like "2g6" and "1.13++". I don't
# think I'm smart enough to do it right though.
2001-12-06 16:51:35 -04:00
#
# In any case, I've coded the test suite for this module (see
# ../test/test_version.py) specifically to fail on things like comparing
# "1.2a2" and "1.2". That's not because the *code* is doing anything
# wrong, it's because the simple, obvious design doesn't match my
# complicated, hairy expectations for real-world version numbers. It
# would be a snap to fix the test suite to say, "Yep, LooseVersion does
# the Right Thing" (ie. the code matches the conception). But I'd rather
# have a conception that matches common notions about version numbers.
class LooseVersion (Version):
"""Version numbering for anarchists and software realists.
Implements the standard interface for version number classes as
described above. A version number consists of a series of numbers,
separated by either periods or strings of letters. When comparing
version numbers, the numeric components will be compared
numerically, and the alphabetic components lexically. The following
are all valid version numbers, in no particular order:
1.5.1
1.5.2b2
161
3.10a
8.02
3.4j
1996.07.12
3.2.pl0
3.1.1.6
2g6
11g
0.960923
2.2beta29
1.13++
5.5.kw
2.0b1pl0
In fact, there is no such thing as an invalid version number under
this scheme; the rules for comparison are simple and predictable,
but may not always give the results you want (for some definition
of "want").
"""
component_re = re.compile(r'(\d+ | [a-z]+ | \.)', re.VERBOSE)
def __init__ (self, vstring=None):
if vstring:
self.parse(vstring)
def parse (self, vstring):
# I've given up on thinking I can reconstruct the version string
# from the parsed tuple -- so I just store the string here for
# use by __str__
self.vstring = vstring
components = [x for x in self.component_re.split(vstring)
if x and x != '.']
for i, obj in enumerate(components):
try:
components[i] = int(obj)
except ValueError:
pass
self.version = components
def __str__ (self):
return self.vstring
def __repr__ (self):
return "LooseVersion ('%s')" % str(self)
def _cmp (self, other):
if isinstance(other, str):
other = LooseVersion(other)
2009-02-16 14:22:15 -04:00
if self.version == other.version:
return 0
if self.version < other.version:
return -1
if self.version > other.version:
return 1
2001-12-06 16:51:35 -04:00
# end class LooseVersion