Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.8/site-packages/bottleneck/slow/nonreduce_axis.py: 12%

1import numpy as np 

2from numpy import partition, argpartition 

3 

4__all__ = ["rankdata", "nanrankdata", "partition", "argpartition", "push"] 

5 

6 

7def rankdata(a, axis=None): 

8 "Slow rankdata function used for unaccelerated dtypes." 

9 return _rank(scipy_rankdata, a, axis) 

10 

11 

12def nanrankdata(a, axis=None): 

13 "Slow nanrankdata function used for unaccelerated dtypes." 

14 return _rank(_nanrankdata_1d, a, axis) 

15 

16 

17def _rank(func1d, a, axis): 

18 a = np.array(a, copy=False) 

19 if axis is None: 

20 a = a.ravel() 

21 axis = 0 

22 if a.size == 0: 

23 y = a.astype(np.float64, copy=True) 

24 else: 

25 y = np.apply_along_axis(func1d, axis, a) 

26 if a.dtype != np.float64: 

27 y = y.astype(np.float64) 

28 return y 

29 

30 

31def _nanrankdata_1d(a): 

32 y = np.empty(a.shape, dtype=np.float64) 

33 y.fill(np.nan) 

34 idx = ~np.isnan(a) 

35 y[idx] = scipy_rankdata(a[idx]) 

36 return y 

37 

38 

39def push(a, n=None, axis=-1): 

40 "Slow push used for unaccelerated dtypes." 

41 if n is None: 

42 n = np.inf 

43 y = np.array(a) 

44 ndim = y.ndim 

45 if axis != -1 or axis != ndim - 1: 

46 y = np.rollaxis(y, axis, ndim) 

47 if ndim == 1: 

48 y = y[None, :] 

49 elif ndim == 0: 

50 return y 

51 fidx = ~np.isnan(y) 

52 recent = np.empty(y.shape[:-1]) 

53 count = np.empty(y.shape[:-1]) 

54 recent.fill(np.nan) 

55 count.fill(np.nan) 

56 with np.errstate(invalid="ignore"): 

57 for i in range(y.shape[-1]): 

58 idx = (i - count) > n 

59 recent[idx] = np.nan 

60 idx = ~fidx[..., i] 

61 y[idx, i] = recent[idx] 

62 idx = fidx[..., i] 

63 count[idx] = i 

64 recent[idx] = y[idx, i] 

65 if axis != -1 or axis != ndim - 1: 

66 y = np.rollaxis(y, ndim - 1, axis) 

67 if ndim == 1: 

68 return y[0] 

69 return y 

70 

71 

72# --------------------------------------------------------------------------- 

73# 

74# SciPy 

75# 

76# Local copy of SciPy's rankdata to avoid a SciPy dependency. The SciPy 

77# license is included in the Bottleneck license file, which is distributed 

78# with Bottleneck. 

79# 

80# Code taken from scipy master branch on Aug 31, 2016. 

81 

82 

83def scipy_rankdata(a, method="average"): 

84 """ 

85 rankdata(a, method='average') 

86 Assign ranks to data, dealing with ties appropriately. 

87 Ranks begin at 1. The `method` argument controls how ranks are assigned 

88 to equal values. See [1]_ for further discussion of ranking methods. 

89 Parameters 

90 ---------- 

91 a : array_like 

92 The array of values to be ranked. The array is first flattened. 

93 method : str, optional 

94 The method used to assign ranks to tied elements. 

95 The options are 'average', 'min', 'max', 'dense' and 'ordinal'. 

96 'average': 

97 The average of the ranks that would have been assigned to 

98 all the tied values is assigned to each value. 

99 'min': 

100 The minimum of the ranks that would have been assigned to all 

101 the tied values is assigned to each value. (This is also 

102 referred to as "competition" ranking.) 

103 'max': 

104 The maximum of the ranks that would have been assigned to all 

105 the tied values is assigned to each value. 

106 'dense': 

107 Like 'min', but the rank of the next highest element is assigned 

108 the rank immediately after those assigned to the tied elements. 

109 'ordinal': 

110 All values are given a distinct rank, corresponding to the order 

111 that the values occur in `a`. 

112 The default is 'average'. 

113 Returns 

114 ------- 

115 ranks : ndarray 

116 An array of length equal to the size of `a`, containing rank 

117 scores. 

118 References 

119 ---------- 

120 .. [1] "Ranking", http://en.wikipedia.org/wiki/Ranking 

121 Examples 

122 -------- 

123 >>> from scipy.stats import rankdata 

124 >>> rankdata([0, 2, 3, 2]) 

125 array([ 1. , 2.5, 4. , 2.5]) 

126 >>> rankdata([0, 2, 3, 2], method='min') 

127 array([ 1, 2, 4, 2]) 

128 >>> rankdata([0, 2, 3, 2], method='max') 

129 array([ 1, 3, 4, 3]) 

130 >>> rankdata([0, 2, 3, 2], method='dense') 

131 array([ 1, 2, 3, 2]) 

132 >>> rankdata([0, 2, 3, 2], method='ordinal') 

133 array([ 1, 2, 4, 3]) 

134 """ 

135 if method not in ("average", "min", "max", "dense", "ordinal"): 

136 raise ValueError('unknown method "{0}"'.format(method)) 

137 

138 a = np.ravel(np.asarray(a)) 

139 algo = "mergesort" if method == "ordinal" else "quicksort" 

140 sorter = np.argsort(a, kind=algo) 

141 

142 inv = np.empty(sorter.size, dtype=np.intp) 

143 inv[sorter] = np.arange(sorter.size, dtype=np.intp) 

144 

145 if method == "ordinal": 

146 return inv + 1 

147 

148 a = a[sorter] 

149 obs = np.r_[True, a[1:] != a[:-1]] 

150 dense = obs.cumsum()[inv] 

151 

152 if method == "dense": 

153 return dense 

154 

155 # cumulative counts of each unique value 

156 count = np.r_[np.nonzero(obs)[0], len(obs)] 

157 

158 if method == "max": 

159 return count[dense] 

160 

161 if method == "min": 

162 return count[dense - 1] + 1 

163 

164 # average method 

165 return 0.5 * (count[dense] + count[dense - 1] + 1)