Acquisition wrapped objects do not behave well on unicode call
Hi, When I have an acquisition wrapped object, e. g. my_object and call:
unicode(my_object)
The method __str__ of my_object is called even when it has an __unicode__ method. See the following bin/zopectl debug session: $ bin/zinstance debug Starting debugger (the name "app" is bound to the top-level Zope object)
import OFS.SimpleItem class A(OFS.SimpleItem.SimpleItem): ... def __str__(self): return '__str__ called' ... def __unicode__(self): return '__unicode__ called' ... a = A() print unicode(a) __unicode__ called print unicode(a.__of__(app)) __str__ called print type(a.__of__(app)) <type 'Acquisition.ImplicitAcquisitionWrapper'>
This happens for Python 2.6 and 2.7 on Zope 2.13.4. Is there a suggested work around? P.S.: I also filed a bug report: https://bugs.launchpad.net/acquisition/+bug/719335 Yours sincerely, -- Michael Howitz · mh@gocept.com · software developer gocept gmbh & co. kg · forsterstraße 29 · 06112 halle (saale) · germany http://gocept.com · tel +49 345 1229889 8 · fax +49 345 1229889 1 Zope and Plone consulting and development
On Tue, Feb 15, 2011 at 2:35 PM, Michael Howitz <mh@gocept.com> wrote:
When I have an acquisition wrapped object, e. g. my_object and call:
unicode(my_object)
The method __str__ of my_object is called even when it has an __unicode__ method.
Acquisition wrappers only fill the tp_repr and tp_str slot and as far as I can tell there's no tp_unicode slot in PyObject_HEAD_INIT. So I'm not sure how to add the C equivalent of a __unicode__ to the Wrappertype and XaqWrappertype PyExtensionClasses. It would probably have to do a lookup for a __unicode__ method on the wrapped instance, call it if it exists and otherwise call its own __str__ - essentially duplicating the logic of the string type. My C-fu is too weak to attempt this. The workaround is to call unicode(aq_base(my_object)). Hanno
On 2011-02-15 15:26:28 +0100, Hanno Schlichting said:
On Tue, Feb 15, 2011 at 2:35 PM, Michael Howitz <mh@gocept.com> wrote:
When I have an acquisition wrapped object, e. g. my_object and call:
unicode(my_object)
The method __str__ of my_object is called even when it has an __unicode__ method.
Acquisition wrappers only fill the tp_repr and tp_str slot and as far as I can tell there's no tp_unicode slot in PyObject_HEAD_INIT.
So I'm not sure how to add the C equivalent of a __unicode__ to the Wrappertype and XaqWrappertype PyExtensionClasses. It would probably have to do a lookup for a __unicode__ method on the wrapped instance, call it if it exists and otherwise call its own __str__ - essentially duplicating the logic of the string type.
Jup.
My C-fu is too weak to attempt this.
Heh. My C-fu is also rather weak but I might try it if there are no objections. (or there is somebody hacking it in 5 minutes …)
The workaround is to call unicode(aq_base(my_object)).
Yes, but that's not really an option. Regards, -- Christian Zagrodnick · cz@gocept.com gocept gmbh & co. kg · forsterstraße 29 · 06112 halle (saale) · germany http://gocept.com · tel +49 345 1229889 4 · fax +49 345 1229889 1 Zope and Plone consulting and development
Am 15.02.2011 um 14:35 schrieb Michael Howitz: [...]
print unicode(a.__of__(app)) __str__ called [...] This happens for Python 2.6 and 2.7 on Zope 2.13.4.
Zope 2.12.8 on Python 2.6 breaks the same way as 2.13. But Zope 2.11.7 on Python 2.5 works correctly. So either Python or Acquisition must have changed since than. Any ideas? Mit freundlichen Grüßen -- Michael Howitz · mh@gocept.com · software developer gocept gmbh & co. kg · forsterstraße 29 · 06112 halle (saale) · germany http://gocept.com · tel +49 345 1229889 8 · fax +49 345 1229889 1 Zope and Plone consulting and development
On Wed, Feb 16, 2011 at 9:18 AM, Michael Howitz <mh@gocept.com> wrote:
Zope 2.12.8 on Python 2.6 breaks the same way as 2.13. But Zope 2.11.7 on Python 2.5 works correctly.
So either Python or Acquisition must have changed since than.
Acquisition changed a lot from 2.11 to 2.12 - it got support for parent pointers for example. The full changelog is quite long, but should be unrelated (http://pypi.python.org/pypi/Acquisition#changelog). The full diff is at: svn di svn+ssh://svn.zope.org/repos/main/Acquisition/branches/2.11/src/Acquisition/_Acquisition.c svn+ssh://svn.zope.org/repos/main/Acquisition/trunk/src/Acquisition/_Acquisition.c There's nothing obvious I can see here. Hanno
On 2011-02-16 10:10:05 +0100, Hanno Schlichting said:
On Wed, Feb 16, 2011 at 9:18 AM, Michael Howitz <mh@gocept.com> wrote:
Zope 2.12.8 on Python 2.6 breaks the same way as 2.13. But Zope 2.11.7 on Python 2.5 works correctly.
So either Python or Acquisition must have changed since than.
Acquisition changed a lot from 2.11 to 2.12 - it got support for parent pointers for example. The full changelog is quite long, but should be unrelated (http://pypi.python.org/pypi/Acquisition#changelog).
The full diff is at:
svn di svn+ssh://svn.zope.org/repos/main/Acquisition/branches/2.11/src/Acquisition/_Acquisition.c svn+ssh://svn.zope.org/repos/main/Acquisition/trunk/src/Acquisition/_Acquisition.c
There's
nothing obvious I can see here.
I suspect a change in python (which I did not follow up after getting strange feelings when looking at PyObject_Unicode). Nevertheless I've fixed that on a branch: svn+ssh://svn.zope.org/repos/main/Acquisition/branches/zagy-unicode-should-be-called Hanno: Would you review that change please? Thanks. -- Christian Zagrodnick · cz@gocept.com gocept gmbh & co. kg · forsterstraße 29 · 06112 halle (saale) · germany http://gocept.com · tel +49 345 1229889 4 · fax +49 345 1229889 1 Zope and Plone consulting and development
On Wed, Feb 16, 2011 at 10:19 PM, Christian Zagrodnick <cz@gocept.com> wrote:
I suspect a change in python (which I did not follow up after getting strange feelings when looking at PyObject_Unicode).
Nevertheless I've fixed that on a branch:
svn+ssh://svn.zope.org/repos/main/Acquisition/branches/zagy-unicode-should-be-called
Hanno:
Would you review that change please?
Sure. I'll review, merge and release. Should be sometime this week, cannot promise a day. Hanno
On 2011-02-16 22:22:53 +0100, Hanno Schlichting said:
On Wed, Feb 16, 2011 at 10:19 PM, Christian Zagrodnick <cz@gocept.com> wrote:
I suspect a change in python (which I did not follow up after getting strange feelings when looking at PyObject_Unicode).
Nevertheless I've fixed that on a branch:
svn+ssh://svn.zope.org/repos/main/Acquisition/branches/zagy-unicode-should-be-called
Hanno: Would
you review that change please?
Sure. I'll review, merge and release. Should be sometime this week, cannot promise a day.
Thanks. There is no pressure. mr.developer FTW :) Anyway, I've added another test to make myself more comfortable. -- Christian Zagrodnick · cz@gocept.com gocept gmbh & co. kg · forsterstraße 29 · 06112 halle (saale) · germany http://gocept.com · tel +49 345 1229889 4 · fax +49 345 1229889 1 Zope and Plone consulting and development
On Thu, Feb 17, 2011 at 8:27 AM, Christian Zagrodnick <cz@gocept.com> wrote:
On 2011-02-16 22:22:53 +0100, Hanno Schlichting said:
svn+ssh://svn.zope.org/repos/main/Acquisition/branches/zagy-unicode-should-be-called
Sure. I'll review, merge and release. Should be sometime this week, cannot promise a day.
Branch reviewed, merged and released in Acquisition 2.13.6. Could you remove the merged branch once you updated your buildout config? Hanno
On 2011-02-19 17:17:44 +0100, Hanno Schlichting said:
On Thu, Feb 17, 2011 at 8:27 AM, Christian Zagrodnick <cz@gocept.com> wrote:
On 2011-02-16 22:22:53 +0100, Hanno Schlichting said:
svn+ssh://svn.zope.org/repos/main/Acquisition/branches/zagy-unicode-should-be-called
Sure.
I'll review, merge and release. Should be sometime this week, cannot promise a day.
Branch reviewed, merged and released in Acquisition 2.13.6.
Could you remove the merged branch once you updated your buildout config?
Done. Thanks for releasing! -- Christian Zagrodnick · cz@gocept.com gocept gmbh & co. kg · forsterstraße 29 · 06112 halle (saale) · germany http://gocept.com · tel +49 345 1229889 4 · fax +49 345 1229889 1 Zope and Plone consulting and development
On 2/20/11 1:32 AM, Christian Zagrodnick wrote:
On 2011-02-19 17:17:44 +0100, Hanno Schlichting said:
On Thu, Feb 17, 2011 at 8:27 AM, Christian Zagrodnick<cz@gocept.com> wrote:
On 2011-02-16 22:22:53 +0100, Hanno Schlichting said:
svn+ssh://svn.zope.org/repos/main/Acquisition/branches/zagy-unicode-should-be-called Sure. I'll review, merge and release. Should be sometime this week, cannot promise a day. Branch reviewed, merged and released in Acquisition 2.13.6.
Could you remove the merged branch once you updated your buildout config? Done.
Thanks for releasing! This change introduces a regression when calling unicode on wrapped objects that implement __str__ but not __unicode__. Essentially it is now doing the equivalent of str(aq_base(obj)) ... __str__ used to get a wrapped object as 'self', but now it is unwrapped.
Here's a failing test that can be added to Acquisition's TestUnicode test case to demonstrate the issue: def test_str_fallback_is_still_wrapped(self): class A(Acquisition.Implicit): def __str__(self): return str(len(Acquisition.aq_chain(self))) wrapped = A().__of__(A()) self.assertEqual(u'2', unicode(wrapped)) This is currently causing some regressions in Plone tests. David ---------- David Glick Web Developer davidglick@groundwire.org 206.286.1235x32 Groundwire: You Are Connected http://groundwire.org Online tools and strategies for the environmental movement. Sign up for Groundwire News! http://groundwire.org/email-capture
On 2011-02-25 21:56:49 +0100, David Glick said:
On 2/20/11 1:32 AM, Christian Zagrodnick wrote:
On 2011-02-19 17:17:44 +0100, Hanno Schlichting said:
On Thu, Feb 17, 2011 at 8:27 AM, Christian Zagrodnick<cz@gocept.com> wrote:
On 2011-02-16 22:22:53 +0100, Hanno Schlichting said:
svn+ssh://svn.zope.org/repos/main/Acquisition/branches/zagy-unicode-should-be-called Sure. I'll
review, merge and release. Should be sometime this week, cannot promise a day. Branch reviewed, merged and released in Acquisition 2.13.6.
Could you remove the merged branch once you updated your buildout config? Done.
Thanks for releasing! This change introduces a regression when calling unicode on wrapped objects that implement __str__ but not __unicode__. Essentially it is now doing the equivalent of str(aq_base(obj)) ... __str__ used to get a wrapped object as 'self', but now it is unwrapped.
Here's a failing test that can be added to Acquisition's TestUnicode test case to demonstrate the issue:
def test_str_fallback_is_still_wrapped(self): class A(Acquisition.Implicit): def __str__(self): return str(len(Acquisition.aq_chain(self))) wrapped = A().__of__(A()) self.assertEqual(u'2', unicode(wrapped))
This is currently causing some regressions in Plone tests.
I'll have a look at it. Thanks for spotting that. -- Christian Zagrodnick · cz@gocept.com gocept gmbh & co. kg · forsterstraße 29 · 06112 halle (saale) · germany http://gocept.com · tel +49 345 1229889 4 · fax +49 345 1229889 1 Zope and Plone consulting and development
On 2011-02-28 07:43:55 +0100, Christian Zagrodnick said:
On 2011-02-25 21:56:49 +0100, David Glick said:
On 2/20/11 1:32 AM, Christian Zagrodnick wrote:
On 2011-02-19 17:17:44 +0100, Hanno Schlichting said:
On Thu, Feb 17, 2011 at 8:27 AM, Christian Zagrodnick<cz@gocept.com> wrote:
On 2011-02-16 22:22:53 +0100, Hanno Schlichting said:
> svn+ssh://svn.zope.org/repos/main/Acquisition/branches/zagy-unicode-should-be-called Sure. I'll>>>>>>>>>>> > review, merge and release. Should be sometime this week, cannot promise a day. Branch reviewed, merged and released in Acquisition 2.13.6.
Could you remove the merged branch once you updated your buildout config? Done.
Thanks for releasing! This change introduces a regression when calling unicode on wrapped objects that implement __str__ but not __unicode__. Essentially it is now doing the equivalent of str(aq_base(obj)) ... __str__ used to get a wrapped object as 'self', but now it is unwrapped.
Here's a failing test that can be added to Acquisition's TestUnicode test case to demonstrate the issue:
def test_str_fallback_is_still_wrapped(self): class A(Acquisition.Implicit): def __str__(self): return str(len(Acquisition.aq_chain(self))) wrapped = A().__of__(A()) self.assertEqual(u'2', unicode(wrapped))
This is currently causing some regressions in Plone tests.
I'll have a look at it. Thanks for spotting that.
Fixed in r120651 (trunk) - Fixed bug: When an object did not implement ``__unicode__``, calling ``unicode(wrapped)`` was calling ``__str__`` with an unwrapped ``self``. Is the situation with Plone better now? -- Christian Zagrodnick · cz@gocept.com gocept gmbh & co. kg · forsterstraße 29 · 06112 halle (saale) · germany http://gocept.com · tel +49 345 1229889 4 · fax +49 345 1229889 1 Zope and Plone consulting and development
On 3/1/11 11:57 AM, Christian Zagrodnick wrote:
This change introduces a regression when calling unicode on wrapped
objects that implement __str__ but not __unicode__. Essentially it is now doing the equivalent of str(aq_base(obj)) ... __str__ used to get a wrapped object as 'self', but now it is unwrapped.
Here's a failing test that can be added to Acquisition's TestUnicode test case to demonstrate the issue:
def test_str_fallback_is_still_wrapped(self): class A(Acquisition.Implicit): def __str__(self): return str(len(Acquisition.aq_chain(self))) wrapped = A().__of__(A()) self.assertEqual(u'2', unicode(wrapped))
This is currently causing some regressions in Plone tests. I'll have a look at it. Thanks for spotting that. Fixed in r120651 (trunk)
- Fixed bug: When an object did not implement ``__unicode__``, calling ``unicode(wrapped)`` was calling ``__str__`` with an unwrapped ``self``.
Is the situation with Plone better now?
Yes, this fixes the test failures we were seeing before. Thanks! David ---------- David Glick Web Developer davidglick@groundwire.org 206.286.1235x32 Groundwire: You Are Connected http://groundwire.org Online tools and strategies for the environmental movement. Sign up for Groundwire News! http://groundwire.org/email-capture
participants (4)
-
Christian Zagrodnick -
David Glick -
Hanno Schlichting -
Michael Howitz