opendp 0.14.2-dev.20260401.2

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

use crate::core::{Function, MetricSpace, Transformation};
use crate::domains::{AtomDomain, Context, ExprDomain, Margin, SeriesDomain, WildExprDomain};
use crate::error::*;
use crate::metrics::{L01InfDistance, LpDistance};
use crate::transformations::traits::UnboundedMetric;
use polars::prelude::len;
use polars_plan::dsl::Expr;
use polars_plan::plans::typed_lit;

use super::expr_count::counting_query_stability_map;

#[cfg(test)]
mod test;

/// Polars operator to get the length of a LazyFrame
///
/// | input_metric                              |
/// | ----------------------------------------- |
/// | `PartitionDistance<SymmetricDistance>`    |
/// | `PartitionDistance<InsertDeleteDistance>` |
///
/// # Arguments
/// * `input_domain` - ExprDomain
/// * `input_metric` - valid selections shown in table above
/// * `expr` - a length expression
pub fn make_expr_len<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,
{
    let Expr::Len = expr else {
        return fallible!(MakeTransformation, "expected len expression");
    };

    let old_margin = input_domain.context.aggregation("len")?;
    let margin = Margin {
        by: old_margin.by,
        max_length: Some(1),
        max_groups: Some(1),
        invariant: old_margin.invariant,
    };

    // build output domain
    let output_domain = ExprDomain {
        column: SeriesDomain::new("len", AtomDomain::<u32>::default()),
        context: Context::Aggregation {
            margin: margin.clone(),
        },
    };

    Transformation::new(
        input_domain,
        input_metric,
        output_domain,
        LpDistance::default(),
        Function::from_expr(len()).fill_with(typed_lit(0u32)),
        counting_query_stability_map(margin.invariant),
    )
}