opentelemetry_sdk/metrics/internal/
histogram.rs

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
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
use std::mem::replace;
use std::ops::DerefMut;
use std::sync::Mutex;

use crate::metrics::data::HistogramDataPoint;
use crate::metrics::data::{self, Aggregation};
use crate::metrics::Temporality;
use opentelemetry::KeyValue;

use super::aggregate::AggregateTimeInitiator;
use super::aggregate::AttributeSetFilter;
use super::ComputeAggregation;
use super::Measure;
use super::ValueMap;
use super::{Aggregator, Number};

impl<T> Aggregator for Mutex<Buckets<T>>
where
    T: Number,
{
    type InitConfig = usize;
    /// Value and bucket index
    type PreComputedValue = (T, usize);

    fn update(&self, (value, index): (T, usize)) {
        let mut buckets = self.lock().unwrap_or_else(|err| err.into_inner());

        buckets.total += value;
        buckets.count += 1;
        buckets.counts[index] += 1;
        if value < buckets.min {
            buckets.min = value;
        }
        if value > buckets.max {
            buckets.max = value
        }
    }

    fn create(count: &usize) -> Self {
        Mutex::new(Buckets::<T>::new(*count))
    }

    fn clone_and_reset(&self, count: &usize) -> Self {
        let mut current = self.lock().unwrap_or_else(|err| err.into_inner());
        Mutex::new(replace(current.deref_mut(), Buckets::new(*count)))
    }
}

#[derive(Default)]
struct Buckets<T> {
    counts: Vec<u64>,
    count: u64,
    total: T,
    min: T,
    max: T,
}

impl<T: Number> Buckets<T> {
    /// returns buckets with `n` bins.
    fn new(n: usize) -> Buckets<T> {
        Buckets {
            counts: vec![0; n],
            min: T::max(),
            max: T::min(),
            ..Default::default()
        }
    }
}

/// Summarizes a set of measurements as a histogram with explicitly defined
/// buckets.
pub(crate) struct Histogram<T: Number> {
    value_map: ValueMap<Mutex<Buckets<T>>>,
    init_time: AggregateTimeInitiator,
    temporality: Temporality,
    filter: AttributeSetFilter,
    bounds: Vec<f64>,
    record_min_max: bool,
    record_sum: bool,
}

impl<T: Number> Histogram<T> {
    #[allow(unused_mut)]
    pub(crate) fn new(
        temporality: Temporality,
        filter: AttributeSetFilter,
        mut bounds: Vec<f64>,
        record_min_max: bool,
        record_sum: bool,
    ) -> Self {
        #[cfg(feature = "spec_unstable_metrics_views")]
        {
            // TODO: When views are used, validate this upfront
            bounds.retain(|v| !v.is_nan());
            bounds.sort_by(|a, b| a.partial_cmp(b).expect("NaNs filtered out"));
        }

        let buckets_count = bounds.len() + 1;
        Histogram {
            value_map: ValueMap::new(buckets_count),
            init_time: AggregateTimeInitiator::default(),
            temporality,
            filter,
            bounds,
            record_min_max,
            record_sum,
        }
    }

    fn delta(&self, dest: Option<&mut dyn Aggregation>) -> (usize, Option<Box<dyn Aggregation>>) {
        let time = self.init_time.delta();

        let h = dest.and_then(|d| d.as_mut().downcast_mut::<data::Histogram<T>>());
        let mut new_agg = if h.is_none() {
            Some(data::Histogram {
                data_points: vec![],
                start_time: time.start,
                time: time.current,
                temporality: Temporality::Delta,
            })
        } else {
            None
        };
        let h = h.unwrap_or_else(|| new_agg.as_mut().expect("present if h is none"));
        h.temporality = Temporality::Delta;
        h.start_time = time.start;
        h.time = time.current;

        self.value_map
            .collect_and_reset(&mut h.data_points, |attributes, aggr| {
                let b = aggr.into_inner().unwrap_or_else(|err| err.into_inner());
                HistogramDataPoint {
                    attributes,
                    count: b.count,
                    bounds: self.bounds.clone(),
                    bucket_counts: b.counts,
                    sum: if self.record_sum {
                        b.total
                    } else {
                        T::default()
                    },
                    min: if self.record_min_max {
                        Some(b.min)
                    } else {
                        None
                    },
                    max: if self.record_min_max {
                        Some(b.max)
                    } else {
                        None
                    },
                    exemplars: vec![],
                }
            });

        (h.data_points.len(), new_agg.map(|a| Box::new(a) as Box<_>))
    }

    fn cumulative(
        &self,
        dest: Option<&mut dyn Aggregation>,
    ) -> (usize, Option<Box<dyn Aggregation>>) {
        let time = self.init_time.cumulative();
        let h = dest.and_then(|d| d.as_mut().downcast_mut::<data::Histogram<T>>());
        let mut new_agg = if h.is_none() {
            Some(data::Histogram {
                data_points: vec![],
                start_time: time.start,
                time: time.current,
                temporality: Temporality::Cumulative,
            })
        } else {
            None
        };
        let h = h.unwrap_or_else(|| new_agg.as_mut().expect("present if h is none"));
        h.temporality = Temporality::Cumulative;
        h.start_time = time.start;
        h.time = time.current;

        self.value_map
            .collect_readonly(&mut h.data_points, |attributes, aggr| {
                let b = aggr.lock().unwrap_or_else(|err| err.into_inner());
                HistogramDataPoint {
                    attributes,
                    count: b.count,
                    bounds: self.bounds.clone(),
                    bucket_counts: b.counts.clone(),
                    sum: if self.record_sum {
                        b.total
                    } else {
                        T::default()
                    },
                    min: if self.record_min_max {
                        Some(b.min)
                    } else {
                        None
                    },
                    max: if self.record_min_max {
                        Some(b.max)
                    } else {
                        None
                    },
                    exemplars: vec![],
                }
            });

        (h.data_points.len(), new_agg.map(|a| Box::new(a) as Box<_>))
    }
}

impl<T> Measure<T> for Histogram<T>
where
    T: Number,
{
    fn call(&self, measurement: T, attrs: &[KeyValue]) {
        let f = measurement.into_float();
        // This search will return an index in the range `[0, bounds.len()]`, where
        // it will return `bounds.len()` if value is greater than the last element
        // of `bounds`. This aligns with the buckets in that the length of buckets
        // is `bounds.len()+1`, with the last bucket representing:
        // `(bounds[bounds.len()-1], +∞)`.
        let index = self.bounds.partition_point(|&x| x < f);

        self.filter.apply(attrs, |filtered| {
            self.value_map.measure((measurement, index), filtered);
        })
    }
}

impl<T> ComputeAggregation for Histogram<T>
where
    T: Number,
{
    fn call(&self, dest: Option<&mut dyn Aggregation>) -> (usize, Option<Box<dyn Aggregation>>) {
        match self.temporality {
            Temporality::Delta => self.delta(dest),
            _ => self.cumulative(dest),
        }
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn check_buckets_are_selected_correctly() {
        let hist = Histogram::<i64>::new(
            Temporality::Cumulative,
            AttributeSetFilter::new(None),
            vec![1.0, 3.0, 6.0],
            false,
            false,
        );
        for v in 1..11 {
            Measure::call(&hist, v, &[]);
        }
        let (count, dp) = ComputeAggregation::call(&hist, None);
        let dp = dp.unwrap();
        let dp = dp.as_any().downcast_ref::<data::Histogram<i64>>().unwrap();
        assert_eq!(count, 1);
        assert_eq!(dp.data_points[0].count, 10);
        assert_eq!(dp.data_points[0].bucket_counts.len(), 4);
        assert_eq!(dp.data_points[0].bucket_counts[0], 1); // 1
        assert_eq!(dp.data_points[0].bucket_counts[1], 2); // 2, 3
        assert_eq!(dp.data_points[0].bucket_counts[2], 3); // 4, 5, 6
        assert_eq!(dp.data_points[0].bucket_counts[3], 4); // 7, 8, 9, 10
    }
}