phrs 0.4.0

A tabular data shell tool compatible with ph
Documentation 
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

//! Rolling numeric window reductions. Windows align to the right by default,
//! and match pandas' min_periods/window and sample-standard-deviation defaults.
use super::statistics::{self, Statistic};
use crate::{cli::Invocation, error::PhError, io};
use polars::prelude::*;

// Sliding sum and mean run in O(n), independent of window size.
fn sliding_sum_or_mean(
    data: &[Option<f64>],
    window: usize,
    center: bool,
    min_periods: usize,
    statistic: Statistic,
) -> Vec<Option<f64>> {
    let mut left = 0;
    let mut right = 0;
    let mut sum = 0.0;
    let mut count = 0_usize;
    let mut output = Vec::with_capacity(data.len());
    for row in 0..data.len() {
        let (before, after) = if center {
            (window / 2, (window - 1) / 2)
        } else {
            (window - 1, 0)
        };
        let start = row.saturating_sub(before);
        let end = row.saturating_add(after).saturating_add(1).min(data.len());
        while right < end {
            if let Some(value) = data[right] {
                sum += value;
                count += 1;
            }
            right += 1;
        }
        while left < start {
            if let Some(value) = data[left] {
                sum -= value;
                count -= 1;
            }
            left += 1;
        }
        output.push(if count < min_periods {
            None
        } else if statistic == Statistic::Sum {
            Some(sum)
        } else if count == 0 {
            None
        } else {
            Some(sum / count as f64)
        });
    }
    output
}

pub fn run(mut df: DataFrame, inv: &Invocation) -> Result<Vec<u8>, PhError> {
    let window = inv.args[0]
        .parse::<usize>()
        .map_err(|_| PhError::new("rolling window must be a positive integer."))?;
    if window == 0 || window > isize::MAX as usize {
        return Err(PhError::new("rolling window must be a positive integer."));
    }
    let statistic = Statistic::parse(inv.option("how").unwrap_or("sum"))?;
    let min_periods = inv.option("min_periods").unwrap_or("");
    let min_periods = if min_periods.is_empty() {
        window
    } else {
        min_periods
            .parse::<usize>()
            .map_err(|_| PhError::new("--min_periods must be a non-negative integer."))?
    };
    if min_periods > window {
        return Err(PhError::new("--min_periods cannot exceed the window size."));
    }
    let center = statistics::boolean(inv.option("center").unwrap_or("False"), "center")?;
    let ddof = inv
        .option("ddof")
        .unwrap_or("1")
        .parse::<i64>()
        .map_err(|_| PhError::new("--ddof must be an integer."))?;

    // With named columns, replace only those columns and leave the others
    // (e.g. date labels) unchanged. With no names, pandas' rolling reduction
    // operates on numeric columns and omits nonnumeric ones.
    let explicit = inv.args.len() > 1;
    let names: Vec<String> = if explicit {
        inv.args[1..].to_vec()
    } else {
        df.columns()
            .iter()
            .filter(|col| statistics::is_numeric(col.dtype()))
            .map(|col| col.name().to_string())
            .collect()
    };
    if names.is_empty() {
        return Err(PhError::new("rolling needs at least one numeric column."));
    }
    for name in &names {
        let col = df
            .column(name.as_str())
            .map_err(|_| PhError::new(format!("Unknown column {name}.")))?;
        if !statistics::is_numeric(col.dtype()) {
            return Err(PhError::new(format!(
                "rolling requires a numeric column: {name}."
            )));
        }
        let converted = col.cast(&DataType::Float64)?;
        let data: Vec<Option<f64>> = converted
            .f64()?
            .iter()
            .map(|item| item.filter(|v| v.is_finite()))
            .collect();
        let result = if matches!(statistic, Statistic::Sum | Statistic::Mean) {
            sliding_sum_or_mean(&data, window, center, min_periods, statistic)
        } else {
            let mut result = Vec::with_capacity(data.len());
            for row in 0..data.len() {
                // For even centered windows, one more observation is on the left.
                let (before, after) = if center {
                    (window / 2, (window - 1) / 2)
                } else {
                    (window - 1, 0)
                };
                let start = row.saturating_sub(before);
                let end = row.saturating_add(after).saturating_add(1).min(data.len());
                let values: Vec<f64> = data[start..end].iter().filter_map(|&v| v).collect();
                result.push(if values.len() >= min_periods {
                    statistics::reduce_numeric(statistic, &values, ddof, 0)
                } else {
                    None
                });
            }
            result
        };
        df.replace(name, Series::new(name.as_str().into(), result).into())?;
    }
    if !explicit {
        return io::write_csv(df.select(names.iter().map(String::as_str))?);
    }
    io::write_csv(df)
}