struct_compression_analyzer/brute_force/

brute_force_custom.rs

use super::{
    find_optimal_coefficients_for_metrics_parallel, BruteForceComparisonMetrics, BruteForceConfig,
    OptimizationResult,
};
use crate::results::analysis_results::AnalysisResults;

/// Result of a brute force optimization on a custom comparison.
#[derive(Debug, Clone)]
pub struct CustomComparisonOptimizationResult {
    /// Optimal parameters for the baseline group
    pub baseline: OptimizationResult,
    /// Optimal parameters for each comparison group
    pub comparisons: Box<[OptimizationResult]>,
}

/// Finds the optimal values for `lz_match_multiplier` and `entropy_multiplier` for all custom
/// comparison results within a given [`AnalysisResults`] item.
///
/// # Arguments
///
/// * `individual_results` - Mutable reference to the [`AnalysisResults`] containing the data.
///   This is where we pull the data from, and where we will update the results.
/// * `config` - Configuration for the optimization process (optional, uses default if [`None`])
pub fn find_optimal_custom_result_coefficients(
    individual_results: &mut [AnalysisResults],
    config: Option<&BruteForceConfig>,
) -> Vec<(String, CustomComparisonOptimizationResult)> {
    // Use a reference to the default config if None is provided
    let default_config = BruteForceConfig::default();
    let config = config.unwrap_or(&default_config);

    let mut results: Vec<(String, CustomComparisonOptimizationResult)> = Vec::new();

    for (comparison_idx, comparison) in individual_results[0].custom_comparisons.iter().enumerate()
    {
        results.push((
            comparison.name.clone(),
            find_optimal_custom_result_coefficients_for_comparison(
                comparison_idx,
                config,
                individual_results,
            ),
        ));
    }

    results
}

/// This function finds the optimal coefficients for the baseline and each comparison group in a custom comparison
#[allow(clippy::needless_range_loop)]
fn find_optimal_custom_result_coefficients_for_comparison(
    comparison_idx: usize,
    config: &BruteForceConfig,
    original_results: &[AnalysisResults], // guaranteed non-empty
) -> CustomComparisonOptimizationResult {
    // Get the first result to determine how many comparison groups exist
    let first_result = &original_results[0].custom_comparisons[comparison_idx];
    let num_comparisons = first_result.group_metrics.len();

    // Extract baseline metrics and find optimal coefficients
    let baseline_metrics = extract_baseline_metrics(comparison_idx, original_results);
    let baseline_best = find_optimal_coefficients_for_metrics_parallel(&baseline_metrics, config);

    // Initialize comparison group optimization results
    let mut comparison_bests = Vec::with_capacity(num_comparisons);

    // Process each comparison group separately
    for group_idx in 0..num_comparisons {
        let group_metrics =
            extract_comparison_group_metrics(comparison_idx, group_idx, original_results);

        // Find optimal coefficients for this comparison group
        let group_best = find_optimal_coefficients_for_metrics_parallel(&group_metrics, config);
        comparison_bests.push(group_best);
    }

    CustomComparisonOptimizationResult {
        baseline: baseline_best,
        comparisons: comparison_bests.into_boxed_slice(),
    }
}

/// Extracts all the baseline metrics from each [`AnalysisResults`], at a given comparison index.
/// Returns a boxed slice of all metrics.
fn extract_baseline_metrics(
    comparison_idx: usize,
    original_results: &[AnalysisResults], // guaranteed non-empty
) -> Box<[BruteForceComparisonMetrics]> {
    original_results
        .iter()
        .map(|result| {
            result.custom_comparisons[comparison_idx]
                .baseline_metrics
                .into()
        })
        .collect()
}

/// Extracts all the metrics for a specific comparison group from each [`AnalysisResults`], at a given comparison index.
/// Returns a boxed slice of all metrics.
///
/// # Arguments
///
/// * `comparison_idx` - The index of the custom comparison in the custom_comparisons array
/// * `group_idx` - The index of the comparison group in the group_metrics array
/// * `original_results` - The original results to extract metrics from
fn extract_comparison_group_metrics(
    comparison_idx: usize,
    group_idx: usize,
    original_results: &[AnalysisResults], // guaranteed non-empty
) -> Box<[BruteForceComparisonMetrics]> {
    original_results
        .iter()
        .map(|result| result.custom_comparisons[comparison_idx].group_metrics[group_idx].into())
        .collect()
}

/// Print optimization results in a user-friendly format.
///
/// # Arguments
///
/// * `writer` - The writer to print results to
/// * `results` - Vector of (comparison name, CustomComparisonOptimizationResult) tuples
pub fn print_optimization_results<W: std::io::Write>(
    writer: &mut W,
    results: &[(String, CustomComparisonOptimizationResult)],
) -> std::io::Result<()> {
    writeln!(
        writer,
        "\n=== Custom Comparison Parameter Optimization Results ==="
    )?;
    writeln!(
        writer,
        "Comparison Name | Group | LZ Multiplier | Entropy Multiplier |"
    )?;
    writeln!(
        writer,
        "----------------|-------|---------------|--------------------|"
    )?;

    for (name, result) in results {
        writeln!(
            writer,
            "{:<16}|{:<7}|{:<15.3}|{:<20.3}|",
            name, "BASE", result.baseline.lz_match_multiplier, result.baseline.entropy_multiplier
        )?;

        for (i, comparison) in result.comparisons.iter().enumerate() {
            writeln!(
                writer,
                "{:<16}|{:<7}|{:<15.3}|{:<20.3}|",
                "", i, comparison.lz_match_multiplier, comparison.entropy_multiplier
            )?;
        }
    }

    Ok(())
}

#[cfg(test)]
mod tests {
    use ahash::AHashMap;

    use super::*;
    use crate::{
        brute_force::calculate_error_for_bruteforce_metrics,
        comparison::{
            compare_groups::GroupComparisonResult, GroupComparisonMetrics, GroupDifference,
        },
        schema::Metadata,
    };

    // Helper function to create a mock GroupComparisonResult for testing
    #[allow(clippy::too_many_arguments)]
    fn create_mock_group_comparison_result(
        name: &str,
        baseline_lz_matches: u64,
        baseline_entropy: f64,
        baseline_zstd_size: u64,
        baseline_original_size: u64,
        comparison_group_count: usize,
        comparison_lz_matches: u64,
        comparison_entropy: f64,
        comparison_zstd_size: u64,
        comparison_original_size: u64,
    ) -> GroupComparisonResult {
        let baseline_metrics = GroupComparisonMetrics {
            lz_matches: baseline_lz_matches,
            entropy: baseline_entropy,
            estimated_size: 0, // Not used in testing
            zstd_size: baseline_zstd_size,
            original_size: baseline_original_size,
        };

        let mut group_names = Vec::with_capacity(comparison_group_count);
        let mut group_metrics = Vec::with_capacity(comparison_group_count);
        let mut differences = Vec::with_capacity(comparison_group_count);

        for i in 0..comparison_group_count {
            group_names.push(format!("group_{}", i));

            let metrics = GroupComparisonMetrics {
                lz_matches: comparison_lz_matches,
                entropy: comparison_entropy,
                estimated_size: 0, // Not used in testing
                zstd_size: comparison_zstd_size,
                original_size: comparison_original_size,
            };

            group_metrics.push(metrics);
            differences.push(GroupDifference::from_metrics(&baseline_metrics, &metrics));
        }

        GroupComparisonResult {
            name: name.to_string(),
            description: "Test comparison".to_string(),
            baseline_metrics,
            group_names,
            group_metrics,
            differences,
        }
    }

    // Helper function to create a mock AnalysisResults with a custom comparison
    #[allow(clippy::too_many_arguments)]
    fn create_mock_analysis_results_with_custom(
        comparison_name: &str,
        baseline_lz_matches: u64,
        baseline_entropy: f64,
        baseline_zstd_size: u64,
        baseline_original_size: u64,
        comparison_0_group_count: usize,
        comparison_0_lz_matches: u64,
        comparison_0_entropy: f64,
        comparison_0_zstd_size: u64,
        comparison_0_original_size: u64,
    ) -> AnalysisResults {
        let custom_comparison = create_mock_group_comparison_result(
            comparison_name,
            baseline_lz_matches,
            baseline_entropy,
            baseline_zstd_size,
            baseline_original_size,
            comparison_0_group_count,
            comparison_0_lz_matches,
            comparison_0_entropy,
            comparison_0_zstd_size,
            comparison_0_original_size,
        );

        AnalysisResults {
            schema_metadata: Metadata {
                name: "Test Schema".to_string(),
                description: "Test Schema Description".to_string(),
            },
            file_entropy: 0.0,
            file_lz_matches: 0,
            zstd_file_size: 0,
            original_size: 0,
            per_field: AHashMap::new(),
            split_comparisons: Vec::new(),
            custom_comparisons: vec![custom_comparison],
        }
    }

    #[test]
    fn can_find_optimal_custom_result_coefficients() {
        // Create a mock analysis results with a custom comparison
        let analysis_results1 = create_mock_analysis_results_with_custom(
            "test_comparison",
            100, // baseline
            1.0,
            110,
            1000,
            2, // comparison_group_count
            210,
            1.6,
            230,
            1000,
        );

        let config = BruteForceConfig::default();

        // Find optimal coefficients
        let mut original_results = vec![analysis_results1];
        let optimal_results =
            find_optimal_custom_result_coefficients(&mut original_results, Some(&config));

        // Verify results
        assert_eq!(optimal_results.len(), 1);
        assert_eq!(optimal_results[0].0, "test_comparison");

        // Verify there are results for baseline and 2 comparison groups
        // Check baseline values are within config range
        assert!(optimal_results[0].1.baseline.lz_match_multiplier >= config.min_lz_multiplier);
        assert!(optimal_results[0].1.baseline.lz_match_multiplier <= config.max_lz_multiplier);
        assert!(optimal_results[0].1.baseline.entropy_multiplier >= config.min_entropy_multiplier);
        assert!(optimal_results[0].1.baseline.entropy_multiplier <= config.max_entropy_multiplier);

        // Check we have the right number of comparison groups
        assert_eq!(optimal_results[0].1.comparisons.len(), 2);

        // Check first comparison group values are within config range
        let comparisons = &optimal_results[0].1.comparisons;
        assert!(comparisons[0].lz_match_multiplier >= config.min_lz_multiplier);
        assert!(comparisons[0].lz_match_multiplier <= config.max_lz_multiplier);
        assert!(comparisons[0].entropy_multiplier >= config.min_entropy_multiplier);
        assert!(comparisons[0].entropy_multiplier <= config.max_entropy_multiplier);

        // Check second comparison group values are within config range
        assert!(comparisons[1].lz_match_multiplier >= config.min_lz_multiplier);
        assert!(comparisons[1].lz_match_multiplier <= config.max_lz_multiplier);
        assert!(comparisons[1].entropy_multiplier >= config.min_entropy_multiplier);
        assert!(comparisons[1].entropy_multiplier <= config.max_entropy_multiplier);

        // Calculate baseline error using the optimal parameters
        let baseline_metrics = extract_baseline_metrics(0, &original_results);
        let baseline_error = calculate_error_for_bruteforce_metrics(
            &baseline_metrics,
            optimal_results[0].1.baseline.lz_match_multiplier,
            optimal_results[0].1.baseline.entropy_multiplier,
        );

        // Assert the error is below a reasonable threshold
        assert!(
            baseline_error < 5.0,
            "Baseline error {} should be less than 5.0",
            baseline_error
        );

        // Check errors for each comparison group
        for i in 0..2 {
            let group_metrics = extract_comparison_group_metrics(0, i, &original_results);
            let group_error = calculate_error_for_bruteforce_metrics(
                &group_metrics,
                optimal_results[0].1.comparisons[i].lz_match_multiplier,
                optimal_results[0].1.comparisons[i].entropy_multiplier,
            );

            assert!(
                group_error < 5.0,
                "Comparison group {} error {} should be less than 5.0",
                i,
                group_error
            );
        }
    }

    #[test]
    fn handles_empty_custom_results() {
        let analysis_results = AnalysisResults::default();

        // Find optimal coefficients
        let mut original_results = vec![analysis_results];
        let optimal_results = find_optimal_custom_result_coefficients(&mut original_results, None);

        // Verify results are empty
        assert!(optimal_results.is_empty());
    }
}