Re: ZCatalog regression: indexing None values

Hi Hanno,

thanks for your reply.

Hanno Schlichting wrote:
> yuppie <y.2015 <at> wcm-solutions.de> writes:
>> I'd say it is no longer possible to store None values in the index, but
>> it's a different question if the attempt to index them is silently
>> ignored or if an error is raised.
>>
>> AFAICS migrating existing code and persistent objects is much easier and
>> in most cases unnecessary if None values are silently ignored.
> 
> As the one who wrote that patch: My idea was only to raise a more meaningful
> TypeError from the catalog code, instead of a the very general "object has
> default comparision" TypeError raised by BTrees.

Yes. The regression was caused by upgrading to BTrees 4, not by your
patch. Your patch just blessed it somehow as intended behavior.

> Silently ignoring None values would also be a good approach. Or maybe one
> could change the 3.1.x releases to log a deprecation warning when None is
> indexed, and create a 4.0 release series where None is silently ignored.
> 
> That way people have a chance to find the places where None is indexed
> and review whether or not this makes a difference in each specific case.

Not sure if I did get you right: Products.ZCatalog 3.1 raises an error,
so people will notice if None is indexed. An additional deprecation
warning doesn't make sense to me.

Changing the behavior would make a regression in 3.x less disruptive, so
I think a 3.2 release would be fine. But I don't care much how it is called.

Anyway I have no ambitions to implement the change I proposed. The issue
is now in the issue tracker and I hope someone else will resolve it.

Cheers,

	Yuppie

_______________________________________________
Zope-Dev maillist  -  [email protected]
https://mail.zope.org/mailman/listinfo/zope-dev
**  No cross posts or HTML encoding!  **
(Related lists -
 https://mail.zope.org/mailman/listinfo/zope-announce
 https://mail.zope.org/mailman/listinfo/zope )