1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
// ferray-ma: Masked reductions (REQ-4)
//
// mean, sum, min, max, var, std, count — all skip masked elements.
use ferray_core::dimension::Dimension;
use ferray_core::dtype::Element;
use ferray_core::error::{FerrayError, FerrayResult};
use num_traits::Float;
use crate::MaskedArray;
impl<T, D> MaskedArray<T, D>
where
T: Element + Copy,
D: Dimension,
{
/// Count the number of unmasked (valid) elements.
///
/// # Errors
/// This function does not currently error but returns `Result` for API
/// consistency.
pub fn count(&self) -> FerrayResult<usize> {
let n = self
.data()
.iter()
.zip(self.mask().iter())
.filter(|(_, m)| !**m)
.count();
Ok(n)
}
}
impl<T, D> MaskedArray<T, D>
where
T: Element + Float,
D: Dimension,
{
/// Compute the sum of unmasked elements.
///
/// Returns zero if all elements are masked.
///
/// # Errors
/// Returns an error only for internal failures.
pub fn sum(&self) -> FerrayResult<T> {
let zero = num_traits::zero::<T>();
let s = self
.data()
.iter()
.zip(self.mask().iter())
.filter(|(_, m)| !**m)
.fold(zero, |acc, (v, _)| acc + *v);
Ok(s)
}
/// Compute the mean of unmasked elements.
///
/// Returns `NaN` if no elements are unmasked.
///
/// # Errors
/// Returns an error only for internal failures.
pub fn mean(&self) -> FerrayResult<T> {
let zero = num_traits::zero::<T>();
let one: T = num_traits::one();
let (sum, count) = self
.data()
.iter()
.zip(self.mask().iter())
.filter(|(_, m)| !**m)
.fold((zero, 0usize), |(s, c), (v, _)| (s + *v, c + 1));
if count == 0 {
return Ok(T::nan());
}
Ok(sum / T::from(count).unwrap_or(one))
}
/// Compute the minimum of unmasked elements.
///
/// # Errors
/// Returns `FerrayError::InvalidValue` if no elements are unmasked.
pub fn min(&self) -> FerrayResult<T> {
self.data()
.iter()
.zip(self.mask().iter())
.filter(|(_, m)| !**m)
.map(|(v, _)| *v)
.fold(None, |acc: Option<T>, v| {
Some(match acc {
Some(a) => {
if v < a {
v
} else {
a
}
}
None => v,
})
})
.ok_or_else(|| FerrayError::invalid_value("min: all elements are masked"))
}
/// Compute the maximum of unmasked elements.
///
/// # Errors
/// Returns `FerrayError::InvalidValue` if no elements are unmasked.
pub fn max(&self) -> FerrayResult<T> {
self.data()
.iter()
.zip(self.mask().iter())
.filter(|(_, m)| !**m)
.map(|(v, _)| *v)
.fold(None, |acc: Option<T>, v| {
Some(match acc {
Some(a) => {
if v > a {
v
} else {
a
}
}
None => v,
})
})
.ok_or_else(|| FerrayError::invalid_value("max: all elements are masked"))
}
/// Compute the variance of unmasked elements (population variance, ddof=0).
///
/// Returns `NaN` if no elements are unmasked.
///
/// # Errors
/// Returns an error only for internal failures.
pub fn var(&self) -> FerrayResult<T> {
let mean = self.mean()?;
if mean.is_nan() {
return Ok(T::nan());
}
let zero = num_traits::zero::<T>();
let one: T = num_traits::one();
let (sum_sq, count) = self
.data()
.iter()
.zip(self.mask().iter())
.filter(|(_, m)| !**m)
.fold((zero, 0usize), |(s, c), (v, _)| {
let d = *v - mean;
(s + d * d, c + 1)
});
if count == 0 {
return Ok(T::nan());
}
Ok(sum_sq / T::from(count).unwrap_or(one))
}
/// Compute the standard deviation of unmasked elements (population, ddof=0).
///
/// Returns `NaN` if no elements are unmasked.
///
/// # Errors
/// Returns an error only for internal failures.
pub fn std(&self) -> FerrayResult<T> {
Ok(self.var()?.sqrt())
}
}