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

use crate::error::EvaluatorError;
use crate::evaluator::*;
use crate::float_trait::Float;
use crate::time_series::TimeSeries;

use std::marker::PhantomData;

macro_const! {
    const DOC: &str = r#"
The engine which extracts features one by one

- Depends on: as reuired by feature evaluators
- Minimum number of observations: as required by feature evaluators
- Number of features: total for all feature evaluators
"#;
}

#[doc = DOC!()]
#[derive(Clone, Debug, Serialize, Deserialize)]
#[serde(
    into = "FeatureExtractorParameters<F>",
    from = "FeatureExtractorParameters<F>",
    bound = "T: Float, F: FeatureEvaluator<T>"
)]
pub struct FeatureExtractor<T, F> {
    features: Vec<F>,
    info: Box<EvaluatorInfo>,
    phantom: PhantomData<T>,
}

impl<T, F> FeatureExtractor<T, F>
where
    T: Float,
    F: FeatureEvaluator<T>,
{
    pub fn new(features: Vec<F>) -> Self {
        let info = EvaluatorInfo {
            size: features.iter().map(|x| x.size_hint()).sum(),
            min_ts_length: features
                .iter()
                .map(|x| x.min_ts_length())
                .max()
                .unwrap_or(0),
            t_required: features.iter().any(|x| x.is_t_required()),
            m_required: features.iter().any(|x| x.is_m_required()),
            w_required: features.iter().any(|x| x.is_w_required()),
            sorting_required: features.iter().any(|x| x.is_sorting_required()),
        }
        .into();
        Self {
            info,
            features,
            phantom: PhantomData,
        }
    }

    pub fn get_features(&self) -> &Vec<F> {
        &self.features
    }

    pub fn into_vec(self) -> Vec<F> {
        self.features
    }

    pub fn add_feature(&mut self, feature: F) {
        self.features.push(feature);
    }
}

impl<T, F> FeatureExtractor<T, F> {
    pub fn doc() -> &'static str {
        DOC
    }
}

impl<T, F> FeatureEvaluator<T> for FeatureExtractor<T, F>
where
    T: Float,
    F: FeatureEvaluator<T>,
{
    fn eval(&self, ts: &mut TimeSeries<T>) -> Result<Vec<T>, EvaluatorError> {
        let mut vec = Vec::with_capacity(self.size_hint());
        for x in self.features.iter() {
            vec.extend(x.eval(ts)?);
        }
        Ok(vec)
    }

    fn eval_or_fill(&self, ts: &mut TimeSeries<T>, fill_value: T) -> Vec<T> {
        self.features
            .iter()
            .flat_map(|x| x.eval_or_fill(ts, fill_value))
            .collect()
    }

    fn get_info(&self) -> &EvaluatorInfo {
        &self.info
    }

    /// Get feature names
    fn get_names(&self) -> Vec<&str> {
        self.features.iter().flat_map(|x| x.get_names()).collect()
    }

    /// Get feature descriptions
    fn get_descriptions(&self) -> Vec<&str> {
        self.features
            .iter()
            .flat_map(|x| x.get_descriptions())
            .collect()
    }
}

#[cfg(test)]
impl<T, F> Default for FeatureExtractor<T, F>
where
    T: Float,
    F: FeatureEvaluator<T>,
{
    fn default() -> Self {
        Self::new(vec![])
    }
}

#[derive(Serialize, Deserialize, JsonSchema)]
#[serde(rename = "FeatureExtractor")]
struct FeatureExtractorParameters<F> {
    features: Vec<F>,
}

impl<T, F> From<FeatureExtractor<T, F>> for FeatureExtractorParameters<F> {
    fn from(f: FeatureExtractor<T, F>) -> Self {
        Self {
            features: f.features,
        }
    }
}

impl<T, F> From<FeatureExtractorParameters<F>> for FeatureExtractor<T, F>
where
    T: Float,
    F: FeatureEvaluator<T>,
{
    fn from(p: FeatureExtractorParameters<F>) -> Self {
        Self::new(p.features)
    }
}

impl<T, F> JsonSchema for FeatureExtractor<T, F>
where
    F: JsonSchema,
{
    json_schema!(FeatureExtractorParameters<F>, true);
}

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

    use serde_test::{assert_ser_tokens, Token};

    serialization_name_test!(FeatureExtractor<f64, Feature<f64>>);

    serde_json_test!(
        feature_extractor_ser_json_de,
        FeatureExtractor<f64, Feature<f64>>,
        FeatureExtractor::new(vec![crate::Amplitude{}.into(), crate::BeyondNStd::new(2.0).into()]),
    );

    check_doc_static_method!(feature_extractor_doc_static_method, FeatureExtractor<f64, Feature<f64>>);

    #[test]
    fn serialization_empty() {
        let fe: FeatureExtractor<f64, Feature<_>> = FeatureExtractor::new(vec![]);
        assert_ser_tokens(
            &fe,
            &[
                //
                Token::Struct {
                    len: 1,
                    name: "FeatureExtractor",
                },
                //
                Token::String("features"),
                Token::Seq { len: Some(0) },
                Token::SeqEnd,
                //
                Token::StructEnd,
            ],
        )
    }
}