plone.reload and super()

[hannosch]

MJ told me at some point that methods using super() aren't reloaded properly.

[d-maurer]

The reason: for methods using super, Python adds a "closure variable" with name __class__. The _closure_changed check detects that the closures are different (one contains the old, the other the new class) and therefore, the method is not updated in place but overridden. The new method definition contains the wrong class reference and therefore the isinstance check of super fails.

Recommendation: replace the simplistic _closure_changed function by something more intelligent: it should ignore for the comparison cells in the closure that correspond to the __class__ cell. To be able to do this, it must get the full function and not only its closures as parameter. <func>.__code__.co_freevars gives in modern Python versions the names for the closure cells. This way, the check can ignore differences for the __class__ cell.

[d-maurer]

Note: I have just released dm.plonepatches.reload with 2 patches to improve on plone.reload. Under Python 3, the patches allow proper handling of super (this issue) and classmethod (--> #11 ). In addition, for Python 2 and 3, they allow a reload of top level modules (--> #10 ) and the update in place of simple decorated functions.

dm.plonepatches.reload is released under the GPL. Thus, you can take inspiration or use pieces of code from it without fear to infringe my intellectual property rights.

[pbauer]

@d-maurer please make a pull-request with your fixes

[d-maurer]

Philip Bauer wrote at 2021-3-18 08:25 -0700: ***@***.*** please make a pull-request with your fixes I am no Plone contributor and do not wish to become one. I have released `dm.plonepatches.reload` under the GPL - thus, it is open source. In addition, I specifically granted you the right to use ideas and code pieces from it to improve `plone.reload`.