Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.8/site-packages/toolz/sandbox/core.py: 59%

1from toolz.itertoolz import getter, cons, pluck 

2from itertools import tee, starmap 

3 

4 

5# See #166: https://github.com/pytoolz/toolz/issues/166 

6# See #173: https://github.com/pytoolz/toolz/pull/173 

7class EqualityHashKey(object): 

8 """ Create a hash key that uses equality comparisons between items. 

9 

10 This may be used to create hash keys for otherwise unhashable types: 

11 

12 >>> from toolz import curry 

13 >>> EqualityHashDefault = curry(EqualityHashKey, None) 

14 >>> set(map(EqualityHashDefault, [[], (), [1], [1]])) # doctest: +SKIP 

15 {=[]=, =()=, =[1]=} 

16 

17 **Caution:** adding N ``EqualityHashKey`` items to a hash container 

18 may require O(N**2) operations, not O(N) as for typical hashable types. 

19 Therefore, a suitable key function such as ``tuple`` or ``frozenset`` 

20 is usually preferred over using ``EqualityHashKey`` if possible. 

21 

22 The ``key`` argument to ``EqualityHashKey`` should be a function or 

23 index that returns a hashable object that effectively distinguishes 

24 unequal items. This helps avoid the poor scaling that occurs when 

25 using the default key. For example, the above example can be improved 

26 by using a key function that distinguishes items by length or type: 

27 

28 >>> EqualityHashLen = curry(EqualityHashKey, len) 

29 >>> EqualityHashType = curry(EqualityHashKey, type) # this works too 

30 >>> set(map(EqualityHashLen, [[], (), [1], [1]])) # doctest: +SKIP 

31 {=[]=, =()=, =[1]=} 

32 

33 ``EqualityHashKey`` is convenient to use when a suitable key function 

34 is complicated or unavailable. For example, the following returns all 

35 unique values based on equality: 

36 

37 >>> from toolz import unique 

38 >>> vals = [[], [], (), [1], [1], [2], {}, {}, {}] 

39 >>> list(unique(vals, key=EqualityHashDefault)) 

40 [[], (), [1], [2], {}] 

41 

42 **Warning:** don't change the equality value of an item already in a hash 

43 container. Unhashable types are unhashable for a reason. For example: 

44 

45 >>> L1 = [1] ; L2 = [2] 

46 >>> s = set(map(EqualityHashDefault, [L1, L2])) 

47 >>> s # doctest: +SKIP 

48 {=[1]=, =[2]=} 

49 

50 >>> L1[0] = 2 # Don't do this! ``s`` now has duplicate items! 

51 >>> s # doctest: +SKIP 

52 {=[2]=, =[2]=} 

53 

54 Although this may appear problematic, immutable data types is a common 

55 idiom in functional programming, and``EqualityHashKey`` easily allows 

56 the same idiom to be used by convention rather than strict requirement. 

57 

58 See Also: 

59 identity 

60 """ 

61 __slots__ = ['item', 'key'] 

62 _default_hashkey = '__default__hashkey__' 

63 

64 def __init__(self, key, item): 

65 if key is None: 

66 self.key = self._default_hashkey 

67 elif not callable(key): 

68 self.key = getter(key) 

69 else: 

70 self.key = key 

71 self.item = item 

72 

73 def __hash__(self): 

74 if self.key == self._default_hashkey: 

75 val = self.key 

76 else: 

77 val = self.key(self.item) 

78 return hash(val) 

79 

80 def __eq__(self, other): 

81 try: 

82 return (self._default_hashkey == other._default_hashkey and 

83 self.item == other.item) 

84 except AttributeError: 

85 return False 

86 

87 def __ne__(self, other): 

88 return not self.__eq__(other) 

89 

90 def __str__(self): 

91 return '=%s=' % str(self.item) 

92 

93 def __repr__(self): 

94 return '=%s=' % repr(self.item) 

95 

96 

97# See issue #293: https://github.com/pytoolz/toolz/issues/239 

98def unzip(seq): 

99 """Inverse of ``zip`` 

100 

101 >>> a, b = unzip([('a', 1), ('b', 2)]) 

102 >>> list(a) 

103 ['a', 'b'] 

104 >>> list(b) 

105 [1, 2] 

106 

107 Unlike the naive implementation ``def unzip(seq): zip(*seq)`` this 

108 implementation can handle an infinite sequence ``seq``. 

109 

110 Caveats: 

111 

112 * The implementation uses ``tee``, and so can use a significant amount 

113 of auxiliary storage if the resulting iterators are consumed at 

114 different times. 

115 

116 * The inner sequence cannot be infinite. In Python 3 ``zip(*seq)`` can be 

117 used if ``seq`` is a finite sequence of infinite sequences. 

118 

119 """ 

120 

121 seq = iter(seq) 

122 

123 # Check how many iterators we need 

124 try: 

125 first = tuple(next(seq)) 

126 except StopIteration: 

127 return tuple() 

128 

129 # and create them 

130 niters = len(first) 

131 seqs = tee(cons(first, seq), niters) 

132 

133 return tuple(starmap(pluck, enumerate(seqs)))