7.15. Operator Accessors¶

obj(x) - call
obj[x] - getitem
obj[x] - missing
obj[x] = 10 - setitem
del obj[x] - delitem

7.15.1. About¶

Table 7.9. Operator Overload¶
Operator	Method	Remarks
`obj(x)`	`obj.__call__(x)`
`obj[x]`	`obj.__getitem__(x)`
`obj[x]`	`obj.__missing__(x)`	(when `x` is not in `obj`)
`obj[x] = 10`	`obj.__setitem__(x, 10)`
`del obj[x]`	`obj.__delitem__(x)`

7.15.2. Example¶

>>> def echo(text):
...     return text
>>>
>>>
>>> echo('Mark Watney')
'Mark Watney'
>>>
>>> echo.__call__('Mark Watney')
'Mark Watney'

7.15.3. Use Case - 0x01¶

>>> data = dict()
>>>
>>> data['a'] = 10  # data.__setitem__('a', 10) -> None
>>> data['a']       # data.__getitem__('a') -> 10
10
>>>
>>> data['x']       # data.__getitem__('x') -> data.__missing__() -> KeyError: 'x'
Traceback (most recent call last):
KeyError: 'x'
>>>
>>> data()          # data.__call__() -> TypeError: 'dict' object is not callable
Traceback (most recent call last):
TypeError: 'dict' object is not callable

7.15.4. Use Case - 0x02¶

Recap information about slice:

>>> data = slice(1, 2, 3)
>>>
>>>
>>> data.start
1
>>>
>>> data.stop
2
>>>
>>> data.step
3

Let's define a class with getitem and it's instance:

>>> class MyClass:
...     def __getitem__(self, item):
...         print(type(item))
>>>
>>>
>>> my = MyClass()

>>> my[1]
<class 'int'>
>>>
>>> my[1.0]
<class 'float'>
>>>
>>> my[1:2]
<class 'slice'>
>>>
>>> my[1,2]
<class 'tuple'>
>>>
>>> my[[1,2]]
<class 'list'>
>>>

>>> my['1969-07-20':'1969-07-21']
<class 'slice'>
>>>
>>> my[['Evening','Morning']]
<class 'list'>
>>>
>>> my['1969-07-20':'1969-07-21', ['Evening','Morning']]
<class 'tuple'>

7.15.5. Use Case - 0x03¶

Getitem in numpy:

>>> import numpy as np
>>>
>>>
>>> data = np.array([[1, 2, 3],
...                  [4, 5, 6],
...                  [7, 8, 9]])
>>>
>>> data[1][2]
6
>>>
>>> data[1,2]
6
>>>
>>> data[1:2]
array([[4, 5, 6]])
>>>
>>> data[1:2, 0]
array([4])
>>>
>>> data[1:2, 1:]
array([[5, 6]])

data[1]:

>>> data.__getitem__(1)
array([4, 5, 6])

data[1,2]:

>>> data.__getitem__((1,2))
6

data[1:2]:

>>> data.__getitem__(slice(1,2))
array([[4, 5, 6]])

data[:, 2]:

>>> data.__getitem__((slice(None,None,None), 2))
array([3, 6, 9])

Intuitive implementation of numpy array[row,col] accessor:

>>> class array(np.ndarray):
...     def __getitem__(key):
...         if isinstance(key, int):
...             return super().__getitem__(key)
...
...         if isinstance(key, tuple):
...             row = key[0]
...             col = key[1]
...             return super().__getitem__(row).__getitem__(col)
...
...         if isinstance(key, slice):
...             start = key[0] if key[0] else 0
...             stop = key[1] if key[0] else len(self)
...             step = key[2] if key[2] else 1
...             return ...

7.15.6. Use Case - 0x04¶

Cache

>>> class Cache(dict):
...     def __init__(self, func):
...         self.func = func
...
...     def __call__(self, *args):
...         return self[args]
...
...     def __missing__(self, key):
...         self[key] = self.func(*key)
...         return self[key]
>>>
>>>
>>> @Cache
... def add(a, b):
...     return a + b
>>>
>>>
>>> _ = add(1,2)  # computed
>>> _ = add(1,2)  # fetched from cache
>>> _ = add(1,2)  # fetched from cache
>>> _ = add(1,2)  # fetched from cache
>>> _ = add(2,1)  # computed
>>> _ = add(2,1)  # fetched from cache
>>>
>>> add  
{(1, 2): 3,
 (2, 1): 3}