do-memory-core 0.1.31

//! Tests for heuristic extraction

use crate::episode::{Episode, ExecutionStep};
use crate::patterns::extractors::heuristic::extraction::is_decision_action;
use crate::patterns::extractors::heuristic::{HeuristicExtractor, HeuristicExtractorConfig};
use crate::types::{ComplexityLevel, ExecutionResult, TaskContext, TaskOutcome, TaskType};

fn create_test_episode() -> Episode {
    let context = TaskContext {
        language: Some("rust".to_string()),
        framework: Some("tokio".to_string()),
        complexity: ComplexityLevel::Moderate,
        domain: "testing".to_string(),
        tags: vec!["async".to_string()],
    };

    Episode::new("Test task".to_string(), context, TaskType::Testing)
}

#[tokio::test]
async fn test_extract_from_complete_successful_episode() {
    // Use low threshold to ensure extraction happens
    let extractor = HeuristicExtractor::with_thresholds(0.5, 2);
    let mut episode = create_test_episode();

    // Add identical decision points so they group together
    for i in 0..3 {
        let mut step = ExecutionStep::new(
            i * 2 + 1,
            "validator".to_string(),
            "Check if input is valid".to_string(), // Same text for all
        );
        step.result = Some(ExecutionResult::Success {
            output: "Valid".to_string(),
        });
        episode.add_step(step);

        // Add follow-up action
        let mut action_step = ExecutionStep::new(
            i * 2 + 2,
            "processor".to_string(),
            "Process the data".to_string(),
        );
        action_step.result = Some(ExecutionResult::Success {
            output: "Processed".to_string(),
        });
        episode.add_step(action_step);
    }

    episode.complete(TaskOutcome::Success {
        verdict: "All validations passed".to_string(),
        artifacts: vec![],
    });

    let heuristics = extractor.extract(&episode).await.unwrap();

    eprintln!("DEBUG: Extracted {} heuristics", heuristics.len());
    for (i, h) in heuristics.iter().enumerate() {
        eprintln!(
            "  Heuristic {}: condition='{}', action='{}', confidence={}, sample_size={}",
            i, h.condition, h.action, h.confidence, h.evidence.sample_size
        );
    }

    // Should extract heuristics (3 identical decisions should group)
    assert!(
        !heuristics.is_empty(),
        "Should extract heuristics. Check debug output above."
    );

    // Check heuristic properties
    for heuristic in &heuristics {
        assert!(heuristic.confidence >= extractor.config.min_confidence);
        assert!(heuristic.evidence.sample_size >= extractor.config.min_sample_size);
        assert!(!heuristic.condition.is_empty());
        assert!(!heuristic.action.is_empty());
    }
}

#[tokio::test]
async fn test_no_extraction_from_incomplete_episode() {
    let extractor = HeuristicExtractor::new();
    let mut episode = create_test_episode();

    let mut step = ExecutionStep::new(
        1,
        "validator".to_string(),
        "Check if input is valid".to_string(),
    );
    step.result = Some(ExecutionResult::Success {
        output: "Valid".to_string(),
    });
    episode.add_step(step);

    // Don't complete the episode

    let heuristics = extractor.extract(&episode).await.unwrap();
    assert!(
        heuristics.is_empty(),
        "Should not extract from incomplete episode"
    );
}

#[tokio::test]
async fn test_no_extraction_from_failed_episode() {
    let extractor = HeuristicExtractor::new();
    let mut episode = create_test_episode();

    let mut step = ExecutionStep::new(
        1,
        "validator".to_string(),
        "Check if input is valid".to_string(),
    );
    step.result = Some(ExecutionResult::Success {
        output: "Valid".to_string(),
    });
    episode.add_step(step);

    episode.complete(TaskOutcome::Failure {
        reason: "Validation failed".to_string(),
        error_details: None,
    });

    let heuristics = extractor.extract(&episode).await.unwrap();
    assert!(
        heuristics.is_empty(),
        "Should not extract from failed episode"
    );
}

#[tokio::test]
async fn test_confidence_calculation() {
    let extractor = HeuristicExtractor::with_thresholds(0.0, 1);
    let mut episode = create_test_episode();

    // Add 4 identical decision points so they group together
    for i in 0..4 {
        let mut step = ExecutionStep::new(
            i * 2 + 1,
            "validator".to_string(),
            "Verify input validity".to_string(), // Same text for all
        );
        step.result = Some(ExecutionResult::Success {
            output: "Valid".to_string(),
        });
        episode.add_step(step);

        // Add follow-up action (same for all)
        let mut action = ExecutionStep::new(
            i * 2 + 2,
            "processor".to_string(),
            "Process data".to_string(),
        );
        action.result = Some(ExecutionResult::Success {
            output: "Done".to_string(),
        });
        episode.add_step(action);
    }

    episode.complete(TaskOutcome::Success {
        verdict: "Success".to_string(),
        artifacts: vec![],
    });

    let heuristics = extractor.extract(&episode).await.unwrap();

    assert!(
        !heuristics.is_empty(),
        "Should extract at least one heuristic"
    );

    // For a success with 4 samples: confidence = 1.0 × √4 = 2.0
    // Actually looking at the code, confidence can exceed 1.0
    let heuristic = &heuristics[0];
    let expected_confidence = 1.0 * (4.0_f32).sqrt(); // = 2.0
    assert!(
        (heuristic.confidence - expected_confidence).abs() < 0.1,
        "Confidence should be ~{}, got {}",
        expected_confidence,
        heuristic.confidence
    );
}

#[tokio::test]
async fn test_filtering_by_confidence_threshold() {
    // Set high confidence threshold
    let extractor = HeuristicExtractor::with_thresholds(2.0, 1);
    let mut episode = create_test_episode();

    // Add only 1 decision point (confidence = 1.0 × √1 = 1.0, below threshold of 2.0)
    let mut step = ExecutionStep::new(1, "validator".to_string(), "Check validity".to_string());
    step.result = Some(ExecutionResult::Success {
        output: "Valid".to_string(),
    });
    episode.add_step(step);

    let mut action = ExecutionStep::new(2, "processor".to_string(), "Process".to_string());
    action.result = Some(ExecutionResult::Success {
        output: "Done".to_string(),
    });
    episode.add_step(action);

    episode.complete(TaskOutcome::Success {
        verdict: "Success".to_string(),
        artifacts: vec![],
    });

    let heuristics = extractor.extract(&episode).await.unwrap();

    // Should be filtered out due to low confidence
    assert!(
        heuristics.is_empty(),
        "Should filter out low confidence heuristics"
    );
}

#[tokio::test]
async fn test_filtering_by_sample_size() {
    // Set minimum sample size to 3
    let extractor = HeuristicExtractor::with_thresholds(0.0, 3);
    let mut episode = create_test_episode();

    // Add only 2 similar decision points (below min sample size of 3)
    for i in 0..2 {
        let mut step =
            ExecutionStep::new(i + 1, "validator".to_string(), "Check validity".to_string());
        step.result = Some(ExecutionResult::Success {
            output: "Valid".to_string(),
        });
        episode.add_step(step);

        let mut action = ExecutionStep::new(i + 3, "processor".to_string(), "Process".to_string());
        action.result = Some(ExecutionResult::Success {
            output: "Done".to_string(),
        });
        episode.add_step(action);
    }

    episode.complete(TaskOutcome::Success {
        verdict: "Success".to_string(),
        artifacts: vec![],
    });

    let heuristics = extractor.extract(&episode).await.unwrap();

    // Should be filtered out due to small sample size
    assert!(
        heuristics.is_empty(),
        "Should filter out heuristics with insufficient samples"
    );
}

#[tokio::test]
async fn test_decision_action_detection() {
    assert!(is_decision_action("Check if valid"));
    assert!(is_decision_action("Verify the input"));
    assert!(is_decision_action("Validate parameters"));
    assert!(is_decision_action("When ready"));
    assert!(is_decision_action("Ensure safety"));
    assert!(is_decision_action("Decide on approach"));
    assert!(is_decision_action("Determine the path"));

    assert!(!is_decision_action("Read file"));
    assert!(!is_decision_action("Write data"));
    assert!(!is_decision_action("Process input"));
}

#[tokio::test]
async fn test_debug_extraction() {
    let extractor = HeuristicExtractor::with_thresholds(0.0, 1);
    let mut episode = create_test_episode();

    // Add a simple decision point
    let mut step = ExecutionStep::new(
        1,
        "validator".to_string(),
        "Check if input is valid".to_string(),
    );
    step.result = Some(ExecutionResult::Success {
        output: "Valid".to_string(),
    });
    episode.add_step(step);

    // Add follow-up action
    let mut action = ExecutionStep::new(2, "processor".to_string(), "Process data".to_string());
    action.result = Some(ExecutionResult::Success {
        output: "Done".to_string(),
    });
    episode.add_step(action);

    episode.complete(TaskOutcome::Success {
        verdict: "Success".to_string(),
        artifacts: vec![],
    });

    let heuristics = extractor.extract(&episode).await.unwrap();

    eprintln!("Extracted {} heuristics", heuristics.len());
    for (i, h) in heuristics.iter().enumerate() {
        eprintln!(
            "Heuristic {}: condition='{}', action='{}', confidence={}, sample_size={}",
            i, h.condition, h.action, h.confidence, h.evidence.sample_size
        );
    }

    // With min_sample_size=1, should extract at least one
    assert!(
        !heuristics.is_empty(),
        "Should extract heuristic with min_sample_size=1"
    );
}

#[tokio::test]
async fn test_partial_success_lower_confidence() {
    let extractor = HeuristicExtractor::with_thresholds(0.0, 1);
    let mut episode = create_test_episode();

    // Add 4 similar decision points
    for i in 0..4 {
        let mut step =
            ExecutionStep::new(i + 1, "validator".to_string(), "Check input".to_string());
        step.result = Some(ExecutionResult::Success {
            output: "Valid".to_string(),
        });
        episode.add_step(step);

        let mut action = ExecutionStep::new(i + 5, "processor".to_string(), "Process".to_string());
        action.result = Some(ExecutionResult::Success {
            output: "Done".to_string(),
        });
        episode.add_step(action);
    }

    episode.complete(TaskOutcome::PartialSuccess {
        verdict: "Partial success".to_string(),
        completed: vec!["some".to_string()],
        failed: vec!["others".to_string()],
    });

    let heuristics = extractor.extract(&episode).await.unwrap();

    assert!(!heuristics.is_empty());

    // For partial success with 4 samples: confidence = 0.5 × √4 = 1.0
    let heuristic = &heuristics[0];
    let expected_confidence = 0.5 * (4.0_f32).sqrt();
    assert!(
        (heuristic.confidence - expected_confidence).abs() < 0.1,
        "Confidence should be ~{}, got {}",
        expected_confidence,
        heuristic.confidence
    );
}

/// Test that default config has expected values
#[test]
fn test_default_config() {
    let config = HeuristicExtractorConfig::default();
    assert_eq!(config.min_confidence, 0.7);
    assert_eq!(config.min_sample_size, 2);
}

/// Test that config can be created with custom values
#[test]
fn test_config_with_custom_values() {
    let config = HeuristicExtractorConfig {
        min_confidence: 0.5,
        min_sample_size: 5,
    };
    assert_eq!(config.min_confidence, 0.5);
    assert_eq!(config.min_sample_size, 5);
}

/// Test extractor can be created with custom config
#[test]
fn test_extractor_with_config() {
    let config = HeuristicExtractorConfig {
        min_confidence: 0.3,
        min_sample_size: 10,
    };
    let extractor = HeuristicExtractor::with_config(config);
    assert_eq!(extractor.config.min_confidence, 0.3);
    assert_eq!(extractor.config.min_sample_size, 10);
}

/// Test extractor can be created with thresholds
#[test]
fn test_extractor_with_thresholds() {
    let extractor = HeuristicExtractor::with_thresholds(0.8, 5);
    assert_eq!(extractor.config.min_confidence, 0.8);
    assert_eq!(extractor.config.min_sample_size, 5);
}