I've run against table inheritance problems (documented in http://code.google.com/p/web2py/issues/detail?id=649 and http://code.google.com/p/web2py/issues/detail?id=648) and others have too (http://code.google.com/p/web2py/issues/detail?id=353). Patches are provided with specific solutions.
However, I think there's a general underlying solution that should be adopted: Right now, when table s inherits from table p, it copies all the "real" fields, with their applied validators etc -- which is not what is tentatively expected, and leads to issues described above. I think a better solution would be to store in the table a copy of all the fields passed in, *before* validators and primary key decisions are made, and when table inheritance is required, copy the original fields, rather than the applied fields, and then compute their validators / keys properly with respect to the inheriting table s. This would also allow introducing a field type for self reference (e.g. 'self') which would always refer to its own table, and solve issue 649. If I understand correctly, any instance in which the behaviour of this solution would deviate from how web2py behaves today is an ambiguous or unspecified or buggy behaviour today -- but I might have missed something.
