opendp 0.16.0

A library of differential privacy algorithms for the statistical analysis of sensitive private data.
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

use std::sync::Arc;

use polars::chunked_array::cast::CastOptions;
use polars::prelude::*;
use polars_plan::dsl::Expr;
use std::collections::HashMap;
use std::sync::LazyLock;

use super::StableExpr;
use crate::core::{Function, MetricSpace, StabilityMap, Transformation};
use crate::domains::{AtomDomain, ExprDomain, WildExprDomain};
use crate::error::*;
use crate::metrics::{L01InfDistance, LpDistance};
use crate::transformations::traits::UnboundedMetric;

#[cfg(test)]
mod test;

// Constant hashmap of types to their allowed cast types.
// This is necessary to prevent downcasting.
static ALLOWED_TRANSFORMATIONS: LazyLock<HashMap<DataType, Vec<DataType>>> = LazyLock::new(|| {
    let mut m = HashMap::new();
    m.insert(
        DataType::Int8,
        vec![
            DataType::Int8,
            DataType::Int16,
            DataType::Int32,
            DataType::Int64,
        ],
    );
    m.insert(
        DataType::Int16,
        vec![DataType::Int16, DataType::Int32, DataType::Int64],
    );
    m.insert(DataType::Int32, vec![DataType::Int32, DataType::Int64]);
    m.insert(DataType::Int64, vec![DataType::Int64]);
    m.insert(DataType::UInt32, vec![DataType::Int64]);
    m
});

/// Make a Transformation that casts an aggregation output to int 64.
/// Casting aggregations to i64 before noise is added can enable negative values.
///
/// # Arguments
/// * `input_domain` - Domain of input data
/// * `input_metric` - Metric on input domain
/// * `expr` - The input measurement to be cast
///
/// # Generics
/// * `TIA` - Atomic Input Type to cast from
/// * `TOA` - Atomic Output Type to cast into
pub fn make_cast_aggregation<MI, const P: usize>(
    input_domain: WildExprDomain,
    input_metric: L01InfDistance<MI>,
    expr: Expr,
) -> Fallible<Transformation<WildExprDomain, L01InfDistance<MI>, ExprDomain, LpDistance<P, f64>>>
where
    MI: 'static + UnboundedMetric,
    (WildExprDomain, L01InfDistance<MI>): MetricSpace,
    (ExprDomain, LpDistance<P, f64>): MetricSpace,
    Expr: StableExpr<L01InfDistance<MI>, LpDistance<P, f64>>,
{
    let Expr::Cast {
        expr: input,
        dtype: to_type,
        mut options,
    } = expr
    else {
        return fallible!(MakeTransformation, "expected cast expression");
    };

    let to_type_dtype = to_type
        .as_literal()
        .ok_or_else(|| {
            err!(
                MakeTransformation,
                "make_cast_aggregation only supports literal dtype"
            )
        })?
        .clone();

    if matches!(options, CastOptions::Strict) {
        options = CastOptions::NonStrict;
    }

    let t_prior = input
        .as_ref()
        .clone()
        .make_stable(input_domain, input_metric)?;

    let (middle_domain, middle_metric) = t_prior.output_space();

    let mut output_domain = middle_domain.clone();
    let active_column = &mut output_domain.column;

    if ALLOWED_TRANSFORMATIONS.contains_key(&to_type_dtype) {
        match to_type_dtype {
            DataType::Int8 => active_column.set_element_domain(AtomDomain::<i8>::default()),
            DataType::Int16 => active_column.set_element_domain(AtomDomain::<i16>::default()),
            DataType::Int32 => active_column.set_element_domain(AtomDomain::<i32>::default()),
            DataType::Int64 => active_column.set_element_domain(AtomDomain::<i64>::default()),
            _ => return fallible!(MakeTransformation, "Unsupported integer target type."),
        }
    } else {
        return fallible!(
            MakeTransformation,
            "make_cast_aggregation cannot cast from {} to {}.",
            active_column.dtype(),
            to_type_dtype
        );
    }

    if !ALLOWED_TRANSFORMATIONS
        .get(&middle_domain.column.dtype())
        .is_some_and(|targets| targets.contains(&to_type_dtype))
    {
        return fallible!(
            MakeTransformation,
            "cannot downcast from {} to {}",
            middle_domain.column.dtype(),
            to_type_dtype
        );
    }

    t_prior
        >> Transformation::new(
            middle_domain.clone(),
            middle_metric.clone(),
            output_domain,
            middle_metric,
            Function::then_expr(move |expr| Expr::Cast {
                expr: Arc::new(expr),
                dtype: to_type.clone(),
                options,
            }),
            StabilityMap::new(Clone::clone),
        )?
}