I got this problem writing a little GUI lib that maps classes to simple table views. Every class member of a certain type = column, and order of columns is important. But...
It's possible in Python3, through the use of PEP3115 which allows you to override the dict type in the metaclass while the class is being constructed (eg. to use an OrderedDict which tracks the insertion order). Here's an implementation of this approach:
class OrderedMeta(type):
@classmethod
def __prepare__(metacls, name, bases):
return OrderedDict()
def __new__(cls, name, bases, clsdict):
c = type.__new__(cls, name, bases, clsdict)
c._orderedKeys = clsdict.keys()
return c
class Person(metaclass=OrderedMeta):
name = None
date_of_birth = None
nationality = None
gender = None
address = None
comment = None
for member in Person._orderedKeys:
if not getattr(Person, member):
print(member)
In Python2, it's a lot more tricky. It would be achievable with something fairly hacky like introspecting the source, and working out the definition order from the AST, but that's probably a lot more trouble than it's worth.
If all you need is an aggregate of variables, perhaps you should use a namedtuple instead. It also maintains the order of the fields (as it's a tuple).
from collections import namedtuple
Person = namedtuple('Person', ('name',
'data_of_birth',
'nationality',
'gender',
'address',
'comment'))
print Person._fields
Ok, it's not an answer per se but a workaround that only works in the context of original question ("Every class member [which is an instance] of a certain type = column, and order of columns is important"). The solution is to introduce a class variable _count
into the CertainType
class, and increment it at every instantiation. After that all class members of the CertainType are packed into a list which is sorted using key=attrgetter('_count')
P.S. Omitting that "which is an instance" part was a mistake on my part and it has limited range of solutions considerably. Sorry for that.