Instantiating a kind of NumPy array

I am creating a class object which must inherit from numpy ndarrays. I perform an isinstance check and an array.shape check with the __new__ method. Is it alright to have it here, or should it be elsewhere? I saw suggestions to create an exceptions class to accompany it, but it doesn't seem necessary... maybe preferable.

'''
pnts_instantiation.py
'''
import numpy as np
import sys
from types import *

class Points(np.ndarray):
  '''ndarray required,info='array info',name='pnts'')
  '''
  def __new__(cls,arr=None,info=None,name='pnts'):
    '''create Points from existing data'''
    err = 'Failed...Requires an ndarray...\nProvided: {}'.format(type(arr))
    if isinstance(arr,(NoneType,StringType,UnicodeType,ListType,TupleType)):
      return err
    if arr.shape < (4,2):
      return ('Failed...Requires array shape > (4,2)\nProvided: {}'.format(arr.shape))
    self = np.asarray(arr).view(cls)   # view as Points class
    self.info = info          # set info
    self.name = name          # set name
    self.applied = None
    # other properties not pertinent to discussion removed for simplicity
    return self

  def __array_finalize__(new_arr, src_arr):
    '''new_arr: new Points object...housecleaning takes place
         for explicit, view casting or new from template...
       src_arr: None, any subclass of ndarray including our own OR another
         instance of our own array (see docs)'''
    if src_arr is None: return
    new_arr.applied = getattr(src_arr,'applied',None) # provide a default
    new_arr.name = getattr(src_arr,'name',None)

  def __array_wrap__(self,out_arr,context=None):
    '''wrap it up'''
    return np.ndarray.__array_wrap__(self, out_arr, context)

  def __repr__(self):
    '''return point info, shape and dtype'''
    s = self
    st = '==> {}'.format(s)
    if (hasattr(s,'name')) and (hasattr(s,'info')):
      st = 'name:  {}\nshape: {}\ninfo:  {}\nvalues:\n{}'.format(s.name,s.shape,s.info,s)
    elif (hasattr(s,'name')) and (hasattr(s, 'applied')):
      st = '{}.{}:  {}'.format(s.name,s.applied, s)
    else:
      st = '{}: {}'.format(s.applied,s)
    return st

def test_cases():
  '''conditional case check'''
  cases = [None,
           'string',
          [[1,2],[3,4]],
          np.asarray([[1,2],[3,4]],dtype='float64',),
          np.ma.asarray([[1,2],[3,4],[5,6]],dtype='float64'),
          np.asarray([[1,2],[3,4],[5,6]],dtype='float64'),
          np.asarray(zip(np.arange(5),np.arange(5)),dtype='float64')
          ]
  counter = 0
  for a_case in cases:
    print('\nCase: {}\nIn: {}\nOut: {}'.format(counter,a_case,Points(a_case)))
    counter += 1

#-------------------
if __name__ == '__main__':
  test_cases()