| msg63792 - (view) |
Author: Steven Bethard (bethard) *  |
Date: 2008-03-17 22:10 |
| Some comparisons were changed or removed in Python 3.0. In 2.6 you could compare types (e.g. ``str < int``) and dicts supported more than just equality. These comparisons should produce Py3K warnings. |
|
|
| msg63901 - (view) |
Author: Steven Bethard (bethard) * |
Date: 2008-03-18 06:53 |
| I'm attaching a patch that handles object comparisons, type comparisons, cell comparisons, and dict comparisons. All the tests pass (including the new ones I've added) but I'd appreciate it if someone could take a second look. Other things still remaining to be done: * Someone needs to decide the correct behavior for method-wrappers (descrobject.c), implement that in Python 2.6 and forward port it to 3.0. * The following objects have a good tp_richcompare in Python 3.0: codeobject.c, methodobject.c, sliceobject.c. Those tp_richcompares should be backported to 2.6. Then a warning can be added for LE, LT, GE and GT (with no warning for EQ or NE which won't change). I may have a little time tomorrow to work on the latter task. |
|
|
| msg63948 - (view) |
Author: Steven Bethard (bethard) * |
Date: 2008-03-18 17:27 |
| Revision 61529 adds warnings for object, type, cell and dict comparisons. The code, method and slice warnings are still needed. |
|
|
| msg63954 - (view) |
Author: Benjamin Peterson (benjamin.peterson) * |
Date: 2008-03-18 17:57 |
| Your patch kicks up warnings in Objects/cellobject.c because cell_compare returns an int, your patch may return NULL. |
|
|
| msg63960 - (view) |
Author: Steven Bethard (bethard) * |
Date: 2008-03-18 18:33 |
| On Tue, Mar 18, 2008 at 11:57 AM, Benjamin Peterson <report@bugs.python.org> wrote: > Benjamin Peterson <musiccomposition@gmail.com> added the comment: > > Your patch kicks up warnings in Objects/cellobject.c because > cell_compare returns an int, your patch may return NULL. Thanks. I'll fix that. Steve |
|
|
| msg63962 - (view) |
Author: Steven Bethard (bethard) * |
Date: 2008-03-18 18:38 |
| So I believe it should be returning -2 instead of NULL. Can someone verify that -2 means raise an exception for tp_compare? |
|
|
| msg63963 - (view) |
Author: Steven Bethard (bethard) * |
Date: 2008-03-18 19:04 |
| Ok, that warning should be gone now in trunk. |
|
|
| msg63997 - (view) |
Author: Steven Bethard (bethard) * |
Date: 2008-03-18 21:48 |
| I took a closer look at sliceobject.c and it looks like both 2.6 and 3.0 compare them basically as tuples. So there don't need to be any warnings about using < and > since these are still well defined. I'll have a patch for codeobject.c and methodobject.c soon. |
|
|
| msg64000 - (view) |
Author: Steven Bethard (bethard) * |
Date: 2008-03-18 22:19 |
| Resolved in revision 61570. I can't get svnmerge block to work though. Since the code and method changes are just backports of Python 3, someone needs to run ``svnmerge.py block -r 61570``. |
|
|