[issue3051] heapq change breaking compatibility
Éric Araujo mer...@netwok.org added the comment: The global docs index has one entry for “comparison”, which is http://docs.python.org/dev/reference/expressions#not-in This other page says that “in general, __lt__() and __eq__() are sufficient, if you want the conventional meanings of the comparison operators”: http://docs.python.org/dev/library/stdtypes.html#comparisons Other useful bits: http://docs.python.org/dev/reference/datamodel#object.__lt__ http://docs.python.org/dev/library/functions#sorted http://docs.python.org/dev/library/functools#functools.cmp_to_key http://docs.python.org/dev/howto/sorting#odd-and-ends It may be useful to add more cross-links between those places (especially pointing to the first link). -- nosy: +eric.araujo ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Roundup Robot devnull@devnull added the comment: New changeset 103a2eb61069 by Raymond Hettinger in branch '2.7': Issue 3051: make pure python code pass the same tests as the C version. http://hg.python.org/cpython/rev/103a2eb61069 -- nosy: +python-dev ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Raymond Hettinger raymond.hettin...@gmail.com added the comment: Maciek, I added the compatability code to the Python version as requested. Now the tests pass for both versions. There is still work to be done to automatically run both versions against the test suite. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Roundup Robot devnull@devnull added the comment: New changeset 83e4765ec4cb by Raymond Hettinger in branch '3.2': Issue 3051: make pure python code pass the same tests as the C version. http://hg.python.org/cpython/rev/83e4765ec4cb -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
R. David Murray rdmur...@bitdance.com added the comment: For what it's worth, I agree with Fijal. I think the python version and the C version should behave the same, so that other implementations of Python can use the Python version and be compatible wtih CPython. -- nosy: +r.david.murray ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Raymond Hettinger rhettin...@users.sourceforge.net added the comment: All six of the rich comparisons need to be implemented or the result is undefined. This module never made guarantees for objects defining only one of the six. We could change the pure python code to handle both __lt__ and __le__ but that would make it much harder to read and understand. The C version supports and that is what runs by default. -- priority: normal - low ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Maciek Fijalkowski fi...@genesilico.pl added the comment: I cannot honestly make much sense from what you said. My concern is whether python and C version behaves the same or not. It seems that in current version they intentionally behave differently, for simplicity and it's against policy of having the same functionality. I agree that it's an obscure corner case, but still. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Raymond Hettinger rhettin...@users.sourceforge.net added the comment: Am closing this. It would make no sense to change simple, pure python code to support objects implementing only one of the rich comparison methods. People implementing rich comparisons need to implement all six if they want to guarantee total ordering and to be usable by various modules that need to be able to make comparisons. FWIW, the C code is not guaranteed to be exactly the same in terms of implementation details, only the published API should be the same. And, for this module, a decision was made for the C code to support only lists eventhough the pure python version supports any sequence. -- status: open - closed ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Maciek Fijalkowski fi...@genesilico.pl added the comment: Hello. I would like to complain. It was decided at some point some time ago that both pure-python and C version should run against the same test suite and should not have any differencies. The reasoning behind it is that other python implementations might choose to use pure-python version and we should avoid surprises with random code crashing in obscure ways. Please don't divert deliberately those sources. Cheers, fijal -- nosy: +fijal ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Changes by Jean-Paul Calderone exar...@twistedmatrix.com: -- status: closed - open ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Changes by Raymond Hettinger [EMAIL PROTECTED]: -- status: open - closed ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Raymond Hettinger [EMAIL PROTECTED] added the comment: Thomas, please let me know if r64116 works for you. ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Thomas Herve [EMAIL PROTECTED] added the comment: Yes, the last commit did the trick. Thanks. ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Raymond Hettinger [EMAIL PROTECTED] added the comment: Would like to make the 3.0 code use __lt__ only. Any objections? -- priority: high - normal ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Thomas Herve [EMAIL PROTECTED] added the comment: Sure, that's fine with me. ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Jean-Paul Calderone [EMAIL PROTECTED] added the comment: I tried this too and then wrote a couple unit tests for this. The one for the Python implementation which tests the case where only __le__ is defined fails, though. Diff attached. -- keywords: +patch Added file: http://bugs.python.org/file10582/test_heapq.diff ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Raymond Hettinger [EMAIL PROTECTED] added the comment: I saw no need to complicate the pure python code for this. Really, the client code should use __cmp__ or define all six rich comparisons. ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Jean-Paul Calderone [EMAIL PROTECTED] added the comment: Thanks for the explanation. Unfortunately, even if we change our code to work with the new requirements, all the old code is still out there. Maybe this doesn't matter, since there are so many other incompatibilities between Python 2.5 and Python 2.6. And there aren't many cases where the extension module isn't available, anyway. It will be surprising and probably hard to debug if anyone runs into this, but I suppose it's possible that no one will. ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Raymond Hettinger [EMAIL PROTECTED] added the comment: There should be no cases where the pure python code runs instead of the C code. ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Changes by Raymond Hettinger [EMAIL PROTECTED]: -- status: closed - open ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Thomas Herve [EMAIL PROTECTED] added the comment: Unfortunately, the modification didn't fix the problem. ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Changes by Georg Brandl [EMAIL PROTECTED]: -- assignee: - rhettinger nosy: +rhettinger ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Thomas Herve [EMAIL PROTECTED] added the comment: Okay then. At least the issue is recorded somewhere, if someone has the same problem. ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Jean-Paul Calderone [EMAIL PROTECTED] added the comment: The heapq documentation isn't very clear about its requirements. It does explicitly say that Heaps are arrays for which heap[k] = heap[2*k+1] and heap[k] = heap[2*k+2] for all k, counting elements from zero. (this in the module reference for the heapq module, both in the Python 2.5 version and the in-development version) which might lead one to believe that = (__le__) is the important operation. I don't know where it is documented that heapq behaves the same as sort(). I think the documentation needs some improvement to avoid this kind of confusion. It's very hard, often impossible, to know what is an accidental and erroneous implementation detail and what is a stable, public API. Also, I'm not sure why the code is being changed to accomodate newly written applications which never could have worked, breaking existing applications which did work, but I suppose that's just the decision CPython developers want to make. -- nosy: +exarkun ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Raymond Hettinger [EMAIL PROTECTED] added the comment: I'll fix this to accommodate both cases, __lt__ and __le__. After trying xy and finding the comparison isn't defined, it can try (not y=x) instead. Also, will document that either __cmp__ or all six rich comparisons should be defined for code that wants to run through sort, bisect, min/max, or heapq. The rich comparison PEP is clear on this point, but I don't think the affirmative statement ever made it to main docs: The reflexivity rules *are* assumed by Python. Thus, the interpreter may swap yx with xy, y=x with x=y, and may swap the arguments of x==y and x!=y. -- PEP 207 -- priority: - high ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue3051] heapq change breaking compatibility
Raymond Hettinger [EMAIL PROTECTED] added the comment: Fixed in r63998. -- resolution: - fixed status: open - closed ___ Python tracker [EMAIL PROTECTED] http://bugs.python.org/issue3051 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
