etl-unit 0.1.0

Semantic data model for ETL units — qualities and measurements over subjects and time. Built on Polars.
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

use polars::{
    prelude::{
        DataFrame, DataType, IntoLazy, RollingOptionsFixedWindow, SortMultipleOptions, col, lit,
    },
    series::ops::NullBehavior,
};

use crate::{EtlResult, EtlSchema, TimeExpr, TimeUnit};

/// Compute an over_time derivation.
pub(crate) fn compute_over_time(
    df: DataFrame,
    output_name: &str,
    expr: &TimeExpr,
    schema: &EtlSchema,
) -> EtlResult<DataFrame> {
    let subject_col = schema.subject.as_str();
    let result = match expr {
        TimeExpr::Derivative { input, time_unit } => {
            compute_derivative(&df, output_name, input, schema, time_unit)?
        }
        TimeExpr::RollingMean { input, window } => df
            .lazy()
            .with_column(
                col(input)
                    .rolling_mean(RollingOptionsFixedWindow {
                        window_size: *window,
                        min_periods: 1,
                        ..Default::default()
                    })
                    .over([col(subject_col)])
                    .alias(output_name),
            )
            .collect()?,
        TimeExpr::RollingSum { input, window } => df
            .lazy()
            .with_column(
                col(input)
                    .rolling_sum(RollingOptionsFixedWindow {
                        window_size: *window,
                        min_periods: 1,
                        ..Default::default()
                    })
                    .over([col(subject_col)])
                    .alias(output_name),
            )
            .collect()?,
        TimeExpr::Lag { input, periods } => df
            .lazy()
            .with_column(
                col(input)
                    .shift(lit(*periods as i64))
                    .over([col(subject_col)])
                    .alias(output_name),
            )
            .collect()?,
        TimeExpr::Lead { input, periods } => df
            .lazy()
            .with_column(
                col(input)
                    .shift(lit(-(*periods as i64)))
                    .over([col(subject_col)])
                    .alias(output_name),
            )
            .collect()?,
        TimeExpr::CumSum { input } => df
            .lazy()
            .with_column(
                col(input)
                    .cum_sum(false)
                    .over([col(subject_col)])
                    .alias(output_name),
            )
            .collect()?,
        TimeExpr::Diff { input, periods } => df
            .lazy()
            .with_column(
                col(input)
                    .diff(lit(*periods as i64), NullBehavior::Ignore)
                    .over([col(subject_col)])
                    .alias(output_name),
            )
            .collect()?,
    };

    Ok(result)
}

/// Compute derivative (rate of change) per subject over time.
pub(crate) fn compute_derivative(
    df: &DataFrame,
    output_name: &str,
    input: &str,
    schema: &EtlSchema,
    time_unit: &TimeUnit,
) -> EtlResult<DataFrame> {
    let subject_col = schema.subject.as_str();
    let time_col = schema.time.as_str();
    let time_divisor = time_unit.from_microseconds();
    let original_cols: Vec<String> = df
        .get_column_names()
        .iter()
        .map(|s| s.to_string())
        .collect();

    let result = df
        .clone()
        .lazy()
        .sort([subject_col, time_col], SortMultipleOptions::default())
        .with_columns([
            col(input)
                .diff(lit(1), NullBehavior::Ignore)
                .over([col(subject_col)])
                .alias("_dv"),
            col(time_col)
                .diff(lit(1), NullBehavior::Ignore)
                .over([col(subject_col)])
                .dt()
                .total_microseconds()
                .alias("_dt_us"),
        ])
        .with_column(
            (col("_dv") / (col("_dt_us").cast(DataType::Float64) / lit(time_divisor)))
                .alias(output_name),
        )
        .select(
            original_cols
                .iter()
                .map(|c| col(c.as_str()))
                .chain(std::iter::once(col(output_name)))
                .collect::<Vec<_>>(),
        )
        .collect()?;

    Ok(result)
}