On Nov 2, 2011, at 7:59 AM, Carlos wrote: > I'm using web2py latest trunk on win7. > > As soon as I applied the following changeset: > > ce20196cd7b895029ed667ecff505987ee83d727 - (ce20196cd7b8) storage patch > > I got errors both in shell and web requests ... > > auth.settings.hmac_key = 'sha512:5fd2a4d2-d703-476c-8184-800b84e453b8' > File "C:\web2py\gluon\storage.py", line 199, in __setattr__ > raise SyntaxError, 'setting key \'%s\' does not exist' % key > SyntaxError: setting key 'hmac_key' does not exist > > If I reload the immediate prior changeset, everything works ok again. > > Based on what I see in this change, you are now removing the Storage key also > via __setitem__ (in addition to __setattr__). > > But I need a way to assign None to a Storage key, for which I was using: > mystorage['key'] = None (which did assign None and did not remove the key). > > Now I would have no way to assign a None value to a Storage key?. > > Plus the above error I get in Settings. > > Can this change be reverted?, or what do you recommend instead?. >
I'll take a look. The patch was intended to make Storage behave consistently with respect to access through attributes and through keys, mainly so that s.key and s['key'] would behave the same way--that is, both yield None when the key isn't present in the underlying dict. The behavior on storing None is along those lines: "s.key = None" deletes the key+value from the dict, and the patch makes "s['key'] = None" do the same thing. I'm frankly not sure that deletion is a great idea in any case; deletion should IMO be explicit. Finally, the settings error appears to be a bug. I'll send Massimo a patch ASAP to revert the new deletion and fix the error; I need to add a case to the unit test.
