datasynth_generators/audit/

sample_generator.rs

//! Audit sample generator per ISA 530.
//!
//! Generates `AuditSample` records with realistic item distributions for
//! workpapers that use statistical sampling.  Workpapers with `SamplingMethod::Judgmental`
//! and `population_size == 0` are treated as non-sampling procedures and receive
//! no sample (returns `None`).

use datasynth_core::utils::seeded_rng;
use rand::RngExt;
use rand_chacha::ChaCha8Rng;
use rust_decimal::Decimal;
use uuid::Uuid;

use datasynth_core::models::audit::{
    AuditSample, SampleItem, SampleItemResult, SamplingMethod, Workpaper,
};

/// Configuration for the sample generator (ISA 530).
#[derive(Debug, Clone)]
pub struct SampleGeneratorConfig {
    /// Number of items to include in each sample (min, max)
    pub items_per_sample: (u32, u32),
    /// Fraction of items that are correct (no misstatement)
    pub correct_ratio: f64,
    /// Fraction of items that have a misstatement
    pub misstatement_ratio: f64,
    /// Fraction of items that have a deviation / exception
    pub exception_ratio: f64,
    /// Only generate samples for workpapers with statistical sampling methods;
    /// when `false`, any workpaper with `population_size > 0` receives a sample.
    pub generate_for_non_sampling: bool,
}

impl Default for SampleGeneratorConfig {
    fn default() -> Self {
        Self {
            items_per_sample: (15, 60),
            correct_ratio: 0.90,
            misstatement_ratio: 0.07,
            exception_ratio: 0.03,
            generate_for_non_sampling: false,
        }
    }
}

/// Generator for `AuditSample` records per ISA 530.
pub struct SampleGenerator {
    /// Seeded random number generator
    rng: ChaCha8Rng,
    /// Configuration
    config: SampleGeneratorConfig,
    /// Counter for human-readable document references
    item_counter: u64,
}

impl SampleGenerator {
    /// Create a new generator with the given seed and default configuration.
    pub fn new(seed: u64) -> Self {
        Self {
            rng: seeded_rng(seed, 0),
            config: SampleGeneratorConfig::default(),
            item_counter: 0,
        }
    }

    /// Create a new generator with custom configuration.
    pub fn with_config(seed: u64, config: SampleGeneratorConfig) -> Self {
        Self {
            rng: seeded_rng(seed, 0),
            config,
            item_counter: 0,
        }
    }

    /// Generate an `AuditSample` for a workpaper, or `None` if sampling is not applicable.
    ///
    /// A sample is generated when:
    /// - The workpaper's `sampling_method` is one of the statistical methods
    ///   (`StatisticalRandom`, `MonetaryUnit`), **or**
    /// - `config.generate_for_non_sampling` is `true` and `workpaper.population_size > 0`.
    ///
    /// # Arguments
    /// * `workpaper`     — The workpaper to create the sample for.
    /// * `engagement_id` — The engagement UUID (must match `workpaper.engagement_id`).
    pub fn generate_sample(
        &mut self,
        workpaper: &Workpaper,
        engagement_id: Uuid,
    ) -> Option<AuditSample> {
        // Decide whether to generate a sample for this workpaper.
        let is_statistical = matches!(
            workpaper.sampling_method,
            SamplingMethod::StatisticalRandom | SamplingMethod::MonetaryUnit
        );
        let has_population = workpaper.population_size > 0;

        let should_generate =
            is_statistical || (self.config.generate_for_non_sampling && has_population);
        if !should_generate {
            return None;
        }

        let sample_count = self
            .rng
            .random_range(self.config.items_per_sample.0..=self.config.items_per_sample.1);

        // Population description derived from workpaper title.
        let pop_description = format!("{} — sampled population", workpaper.title);

        let mut sample = AuditSample::new(
            workpaper.workpaper_id,
            engagement_id,
            pop_description,
            workpaper.population_size.max(sample_count as u64),
            workpaper.sampling_method,
            sample_count,
        );

        // Set population value (rough estimate: average item ~$50k × population size).
        let pop_value_units: i64 = (workpaper.population_size as i64)
            .saturating_mul(50_000_i64)
            .max(100_000);
        sample.population_value = Some(Decimal::new(pop_value_units, 0));

        // Tolerable misstatement ≈ 5% of population value.
        sample.tolerable_misstatement = sample.population_value.map(|v| v / Decimal::from(20));

        // Generate the individual items.
        for _ in 0..sample_count {
            self.item_counter += 1;
            let doc_ref = format!("DOC-{:06}", self.item_counter);

            // Book value: $1k – $500k
            let book_units: i64 = self.rng.random_range(1_000_i64..=500_000_i64);
            let book_value = Decimal::new(book_units, 0);

            let roll: f64 = self.rng.random();
            let misstatement_cutoff = self.config.misstatement_ratio;
            let exception_cutoff = misstatement_cutoff + self.config.exception_ratio;

            let mut item = SampleItem::new(&doc_ref, book_value);

            if roll < misstatement_cutoff {
                // Misstatement: audited value differs by 1–15% of book.
                let pct: f64 = self.rng.random_range(0.01..0.15);
                let diff_units = (book_units as f64 * pct).round() as i64;
                let diff = Decimal::new(diff_units.max(1), 0);
                // Randomly overstate or understate.
                let audited = if self.rng.random::<bool>() {
                    book_value + diff
                } else {
                    (book_value - diff).max(Decimal::ZERO)
                };
                let misstatement = book_value - audited;

                item.audited_value = Some(audited);
                item.misstatement = Some(misstatement);
                item.result = SampleItemResult::Misstatement;
            } else if roll < exception_cutoff {
                // Exception / deviation: audited value differs by 5–20%.
                let pct: f64 = self.rng.random_range(0.05..0.20);
                let diff_units = (book_units as f64 * pct).round() as i64;
                let diff = Decimal::new(diff_units.max(1), 0);
                let audited = (book_value - diff).max(Decimal::ZERO);
                let misstatement = book_value - audited;

                item.audited_value = Some(audited);
                item.misstatement = Some(misstatement);
                item.result = SampleItemResult::Exception;
            } else {
                // Correct: audited value equals book value.
                item.audited_value = Some(book_value);
                item.result = SampleItemResult::Correct;
            }

            sample.add_item(item);
        }

        // Compute projection and reach a conclusion.
        sample.conclude();

        // Upgrade to InsufficientEvidence → use actual projection.
        // (conclude() handles this already; no override needed.)

        Some(sample)
    }
}

// =============================================================================
// Tests
// =============================================================================

#[cfg(test)]
mod tests {
    use super::*;
    use datasynth_core::models::audit::{
        ProcedureType, SampleConclusion, Workpaper, WorkpaperScope, WorkpaperSection,
    };

    fn make_gen(seed: u64) -> SampleGenerator {
        SampleGenerator::new(seed)
    }

    /// Build a workpaper that will receive a sample (statistical method + population).
    fn sampling_workpaper(method: SamplingMethod) -> Workpaper {
        Workpaper::new(
            Uuid::new_v4(),
            "D-100",
            "Accounts Receivable Testing",
            WorkpaperSection::SubstantiveTesting,
        )
        .with_procedure("Test AR balances", ProcedureType::SubstantiveTest)
        .with_scope(WorkpaperScope::default(), 1_000, 50, method)
    }

    fn non_sampling_workpaper() -> Workpaper {
        Workpaper::new(
            Uuid::new_v4(),
            "C-100",
            "Controls Walk-through",
            WorkpaperSection::ControlTesting,
        )
        .with_scope(WorkpaperScope::default(), 0, 0, SamplingMethod::Judgmental)
    }

    // -------------------------------------------------------------------------

    /// A statistical-method workpaper produces a sample with items in range.
    #[test]
    fn test_generates_sample() {
        let wp = sampling_workpaper(SamplingMethod::StatisticalRandom);
        let eng_id = wp.engagement_id;
        let mut gen = make_gen(42);
        let sample = gen.generate_sample(&wp, eng_id).unwrap();

        let cfg = SampleGeneratorConfig::default();
        let min = cfg.items_per_sample.0 as usize;
        let max = cfg.items_per_sample.1 as usize;
        assert!(
            sample.items.len() >= min && sample.items.len() <= max,
            "expected {min}..={max} items, got {}",
            sample.items.len()
        );
        assert!(
            sample.conclusion.is_some(),
            "sample should have a conclusion"
        );
    }

    /// A non-sampling workpaper with population_size == 0 returns None.
    #[test]
    fn test_no_sample_for_non_sampling() {
        let wp = non_sampling_workpaper();
        let eng_id = wp.engagement_id;
        let mut gen = make_gen(99);
        let result = gen.generate_sample(&wp, eng_id);
        assert!(result.is_none(), "expected None for non-sampling workpaper");
    }

    /// With a large count, the item result distribution should roughly match the config.
    #[test]
    fn test_item_distribution() {
        let wp = sampling_workpaper(SamplingMethod::MonetaryUnit);
        let eng_id = wp.engagement_id;
        let config = SampleGeneratorConfig {
            items_per_sample: (300, 300),
            correct_ratio: 0.90,
            misstatement_ratio: 0.07,
            exception_ratio: 0.03,
            generate_for_non_sampling: false,
        };
        let mut gen = SampleGenerator::with_config(77, config);
        let sample = gen.generate_sample(&wp, eng_id).unwrap();

        let total = sample.items.len() as f64;
        let correct_count = sample
            .items
            .iter()
            .filter(|i| i.result == SampleItemResult::Correct)
            .count() as f64;

        // Correct ratio should be within ±15% of 90%.
        let ratio = correct_count / total;
        assert!(
            (0.75..=1.00).contains(&ratio),
            "correct ratio {ratio:.2} outside expected 75–100%"
        );
    }

    /// Same seed produces identical output.
    #[test]
    fn test_deterministic() {
        let wp = sampling_workpaper(SamplingMethod::StatisticalRandom);
        let eng_id = wp.engagement_id;

        let sample_a = SampleGenerator::new(1234)
            .generate_sample(&wp, eng_id)
            .unwrap();
        let sample_b = SampleGenerator::new(1234)
            .generate_sample(&wp, eng_id)
            .unwrap();

        assert_eq!(sample_a.items.len(), sample_b.items.len());
        for (a, b) in sample_a.items.iter().zip(sample_b.items.iter()) {
            assert_eq!(a.document_ref, b.document_ref);
            assert_eq!(a.book_value, b.book_value);
            assert_eq!(a.result, b.result);
        }
        assert_eq!(sample_a.conclusion, sample_b.conclusion);
    }

    /// `generate_for_non_sampling = true` causes a Judgmental-method workpaper
    /// with population_size > 0 to receive a sample.
    #[test]
    fn test_generate_for_non_sampling_flag() {
        let mut wp = non_sampling_workpaper();
        wp.population_size = 500; // non-zero population
        let eng_id = wp.engagement_id;

        let config = SampleGeneratorConfig {
            generate_for_non_sampling: true,
            ..Default::default()
        };
        let mut gen = SampleGenerator::with_config(55, config);
        let result = gen.generate_sample(&wp, eng_id);
        assert!(
            result.is_some(),
            "expected Some when generate_for_non_sampling = true"
        );
    }

    /// Misstatement items should have a non-zero misstatement amount.
    #[test]
    fn test_misstatement_items_have_amounts() {
        let wp = sampling_workpaper(SamplingMethod::StatisticalRandom);
        let eng_id = wp.engagement_id;
        let config = SampleGeneratorConfig {
            items_per_sample: (200, 200),
            misstatement_ratio: 0.50, // inflate so we always get some
            exception_ratio: 0.10,
            correct_ratio: 0.40,
            generate_for_non_sampling: false,
        };
        let mut gen = SampleGenerator::with_config(33, config);
        let sample = gen.generate_sample(&wp, eng_id).unwrap();

        let mist_items: Vec<_> = sample
            .items
            .iter()
            .filter(|i| i.result == SampleItemResult::Misstatement)
            .collect();

        assert!(!mist_items.is_empty(), "expected some misstatement items");
        for item in mist_items {
            assert!(
                item.misstatement.is_some(),
                "misstatement item should have a misstatement amount"
            );
            // misstatement amount should be non-zero
            assert_ne!(
                item.misstatement.unwrap(),
                Decimal::ZERO,
                "misstatement amount should not be zero"
            );
        }
    }

    /// The sample conclusion should be a valid `SampleConclusion` variant.
    #[test]
    fn test_conclusion_is_set() {
        let wp = sampling_workpaper(SamplingMethod::MonetaryUnit);
        let eng_id = wp.engagement_id;
        let mut gen = make_gen(12);
        let sample = gen.generate_sample(&wp, eng_id).unwrap();

        let conclusion = sample.conclusion.unwrap();
        let valid = matches!(
            conclusion,
            SampleConclusion::ProjectedBelowTolerable
                | SampleConclusion::ProjectedExceedsTolerable
                | SampleConclusion::InsufficientEvidence
        );
        assert!(valid, "unexpected SampleConclusion variant");
    }
}