python - Numpy subclass attribute slicing -

myclass numpy.ndarray subclass, intended represent set of images change on time. each image there set of metadata, time, ambient temperature , camera temperature. i've stored these metadata in list of dictionaries, each dictionary corresponds layer in array (myclass.metadata[0] dictionary corresponds image in myclass[0]).

i've overloaded getattr() make items in dictionary accessible key, myclass.etemp yields e.g. [24.9, 25.0, 25.1].

when slice myclass-object, how achieve attribute-array gets sliced same way?

now if myobject[1].etemp, [24.9, 25.0, 25.1], want [25.0].

this class:

class stack(numpy.ndarray):     props= [             'version',             'width',             'height',             'shotrange',             'calibrange',             'unit',             'irb_fname',             'fidx',             'distance',             'etemp',             'zoom',             'ctemp',             'date',             'recdate',             'rectime',             ]      def __new__(cls, input_array, mdata=none):         obj = numpy.asarray(input_array).view(cls)         if isinstance(mdata, collections.iterable): # when reading text file             obj.mdata = mdata         else:             obj.mdata = [arr.mdata[0] arr in input_array] # when combining stack-type objects         return obj      def __array_finalize__(self, obj):         if obj none: return         self.mdata = getattr(obj, 'mdata', none)      def __getattr__(self, name):         print(numpy.shape(self))         if numpy.rank(self) < 3: # we're looking @ single slice             pass         if name == 'starttime':             return self.mdata[0]['date']         elif name == 'time':             return [(item['date'] - self.mdata[0]['date']).total_seconds() item in self.mdata]         elif name in stack.props:             return [item[name] item in self.mdata]         else:             raise attributeerror 

what need implement behavior? or there other better way store metadata?

you need override __getitem__ method.

class foo(object):     def __getitem__(self,items):         print items  f = foo() f[1,2,3] f[1:3] f[1,1:3,2:3] 

this returns:

1 (1, 2, 3) slice(1, 3, none) (1, slice(1, 3, none), slice(2, 3, none)) 

within getitem, you'll need slice attributes appropriately handling above cases.