tetration 0.1.9

Chunked tensor .tet format, tet CLI, and optional C ABI (tetration-ffi)
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
141
142
143
144

//! CF conventions (`scale_factor`, `add_offset`, `_FillValue`) applied at import.

use crate::utils::dtype::ElementDtype;

use super::ConvertError;

/// CF numeric packing decoded during tile read.
#[derive(Debug, Clone, Copy, PartialEq)]
pub(crate) struct CfTransform {
    pub scale: f64,
    pub offset: f64,
    pub fill_value: Option<f64>,
}

impl CfTransform {
    pub(crate) fn apply_tile_le(
        &self,
        dtype: ElementDtype,
        buf: &mut [u8],
    ) -> Result<(), ConvertError> {
        match dtype {
            ElementDtype::F32 => {
                apply_cf_f32(self, buf);
                Ok(())
            }
            ElementDtype::F64 => {
                apply_cf_f64(self, buf);
                Ok(())
            }
            _ => Err(ConvertError::UnsupportedDtype {
                name: "cf-packed variable".into(),
                detail: "CF decode supports f32/f64 storage only".into(),
            }),
        }
    }
}

fn apply_cf_f32(cf: &CfTransform, buf: &mut [u8]) {
    for chunk in buf.chunks_exact_mut(4) {
        let stored = f32::from_le_bytes(chunk.try_into().expect("f32 tile"));
        let decoded = decode_cf_f32(stored, cf);
        chunk.copy_from_slice(&decoded.to_le_bytes());
    }
}

fn apply_cf_f64(cf: &CfTransform, buf: &mut [u8]) {
    for chunk in buf.chunks_exact_mut(8) {
        let stored = f64::from_le_bytes(chunk.try_into().expect("f64 tile"));
        let decoded = decode_cf_f64(stored, cf);
        chunk.copy_from_slice(&decoded.to_le_bytes());
    }
}

fn decode_cf_f32(stored: f32, cf: &CfTransform) -> f32 {
    if cf_fill_matches_f32(stored, cf) {
        return f32::NAN;
    }
    (f64::from(stored) * cf.scale + cf.offset) as f32
}

fn decode_cf_f64(stored: f64, cf: &CfTransform) -> f64 {
    if cf_fill_matches_f64(stored, cf) {
        return f64::NAN;
    }
    stored * cf.scale + cf.offset
}

fn cf_fill_matches_f32(stored: f32, cf: &CfTransform) -> bool {
    cf.fill_value.is_some_and(|fill| {
        let fill_f32 = fill as f32;
        stored.to_bits() == fill_f32.to_bits()
            || (stored - fill_f32).abs() <= cf_fill_tolerance(fill) as f32
    })
}

fn cf_fill_matches_f64(stored: f64, cf: &CfTransform) -> bool {
    cf.fill_value.is_some_and(|fill| {
        if stored.is_nan() && fill.is_nan() {
            return true;
        }
        (stored - fill).abs() <= cf_fill_tolerance(fill)
    })
}

fn cf_fill_tolerance(fill: f64) -> f64 {
    (fill.abs() * 1.0e-6).max(1.0e-6)
}

#[cfg(feature = "tetration-hdf5")]
pub(crate) fn cf_from_hdf5(ds: &hdf5_metno::Dataset) -> Option<CfTransform> {
    let scale = hdf5_attr_f64(ds, "scale_factor");
    let offset = hdf5_attr_f64(ds, "add_offset");
    if scale.is_none() && offset.is_none() {
        return None;
    }
    Some(CfTransform {
        scale: scale.unwrap_or(1.0),
        offset: offset.unwrap_or(0.0),
        fill_value: hdf5_attr_f64(ds, "_FillValue"),
    })
}

#[cfg(feature = "tetration-hdf5")]
fn hdf5_attr_f64(ds: &hdf5_metno::Dataset, name: &str) -> Option<f64> {
    let attr = ds.attr(name).ok()?;
    let reader = attr.as_reader();
    reader
        .read_scalar::<f64>()
        .ok()
        .or_else(|| reader.read_scalar::<f32>().ok().map(f64::from))
        .or_else(|| reader.read_scalar::<i32>().ok().map(f64::from))
}

#[cfg(feature = "tetration-netcdf")]
pub(crate) fn cf_from_netcdf(var: &netcdf::Variable<'_>) -> Option<CfTransform> {
    let scale = nc_attr_f64(var, "scale_factor");
    let offset = nc_attr_f64(var, "add_offset");
    if scale.is_none() && offset.is_none() {
        return None;
    }
    let fill = nc_attr_f64(var, "_FillValue").or_else(|| nc_fill_value_f64(var));
    Some(CfTransform {
        scale: scale.unwrap_or(1.0),
        offset: offset.unwrap_or(0.0),
        fill_value: fill,
    })
}

#[cfg(feature = "tetration-netcdf")]
fn nc_attr_f64(var: &netcdf::Variable<'_>, name: &str) -> Option<f64> {
    match var.attribute_value(name) {
        Some(Ok(value)) => value.try_into().ok(),
        _ => None,
    }
}

#[cfg(feature = "tetration-netcdf")]
fn nc_fill_value_f64(var: &netcdf::Variable<'_>) -> Option<f64> {
    var.fill_value::<f32>()
        .ok()
        .flatten()
        .map(f64::from)
        .or_else(|| var.fill_value::<f64>().ok().flatten())
}