0

Suppose I have a class:

class Data:
  data: np.ndarray
  def __init__(self):
      self.data = np.zeros((10,10))
      self.str2col = { str(i):i for i in range(0,10) }
  def __get_col(self,i):
      """
      @param i: STR or INT
      """
      if isinstance(i,int):
        return i 
      if isinstance(i,str):
        return self.str2col[i]
      raise Exception("Invalid index")

Assuming I haven't made any errors, I can index this class as follows: 

i = '4'
j = 4
obj = Data()
point = obj.data[obj.__get_col(i),obj.__get_col(j)]

The question is: how does one, in python, overload the "[]" such that I can write, in full numpy indexing syntax, 

data = obj[:,:]
data = obj['1',:]
data = obj[:,'2']
Chris
  • 28,822
  • 27
  • 83
  • 158
  • 1
    `obj[x:y:z]` is equivalent to `obj.__getitem__(slice(x,y,z))`, and note, the commas are just tuples, so something like `obj[x0:y0:z0, x1,y1,z1]` is equivalent to `obj.__getitem__((slice(x0,y0,z0),slice(x1,y1,z1)))` – juanpa.arrivillaga Jan 06 '20 at 21:32
  • 1
    The `:` triggers the creation of a `slice` object; `obj[x0:y0:z0, x1,y1,z1]` is equivalent to `obj.__getitem__((slice(x0,y0,z0), x1, y1, z1)`. – chepner Jan 06 '20 at 21:47
  • 1
    Technically, I think the commas are part of the syntax, not a tuple expression, though the end result is the same (`__getitem__` receiving a tuple). I'm not sure there is any way to observe the difference at the Python level, though. – chepner Jan 06 '20 at 21:48
  • The numpy `index_tricks` file has a bunch of classes with custom indexing code. It may give you ideas. – hpaulj Jan 06 '20 at 21:55

0 Answers0