profile-inspect 0.1.3

//! Intelligent recommendation engine for CPU profiles
//!
//! Generates actionable, prioritized recommendations with quantified impact estimates.

use crate::ir::{FrameCategory, ProfileIR};

use super::{CpuAnalysis, FunctionStats, PerformancePattern};

/// Priority level for recommendations
#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
pub enum Priority {
    /// Must fix - critical performance issue
    Critical,
    /// Should fix - significant impact
    High,
    /// Could fix - moderate improvement
    Medium,
    /// Nice to have - minor optimization
    Low,
}

impl std::fmt::Display for Priority {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            Self::Critical => write!(f, "CRITICAL"),
            Self::High => write!(f, "HIGH"),
            Self::Medium => write!(f, "MEDIUM"),
            Self::Low => write!(f, "LOW"),
        }
    }
}

/// Effort level to implement a fix
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Effort {
    /// Quick fix - minutes to implement
    QuickWin,
    /// Moderate - hours of work
    Moderate,
    /// Significant - days of refactoring
    Significant,
    /// Major - architectural changes
    Major,
}

impl std::fmt::Display for Effort {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            Self::QuickWin => write!(f, "Quick Win"),
            Self::Moderate => write!(f, "Moderate"),
            Self::Significant => write!(f, "Significant"),
            Self::Major => write!(f, "Major Refactor"),
        }
    }
}

/// Type of issue detected
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum IssueType {
    /// Algorithm complexity issue
    Algorithm,
    /// Memory/allocation pressure
    Memory,
    /// Dependency overhead
    Dependency,
    /// I/O or serialization
    Serialization,
    /// Repeated work that could be cached
    Caching,
    /// Startup/initialization overhead
    Startup,
    /// Recursion that could be optimized
    Recursion,
    /// General hotspot
    Hotspot,
}

impl std::fmt::Display for IssueType {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            Self::Algorithm => write!(f, "Algorithm"),
            Self::Memory => write!(f, "Memory"),
            Self::Dependency => write!(f, "Dependency"),
            Self::Serialization => write!(f, "Serialization"),
            Self::Caching => write!(f, "Caching"),
            Self::Startup => write!(f, "Startup"),
            Self::Recursion => write!(f, "Recursion"),
            Self::Hotspot => write!(f, "Hotspot"),
        }
    }
}

/// A single actionable recommendation
#[derive(Debug, Clone)]
pub struct Recommendation {
    /// Priority level
    pub priority: Priority,
    /// Effort to implement
    pub effort: Effort,
    /// Type of issue
    pub issue_type: IssueType,
    /// Short title (action-oriented)
    pub title: String,
    /// Function or component involved
    pub target: String,
    /// Location in code
    pub location: String,
    /// Current time spent (microseconds)
    pub current_time_us: u64,
    /// Estimated time savings (microseconds)
    pub estimated_savings_us: u64,
    /// Why this is a problem (root cause)
    pub root_cause: String,
    /// Specific actions to take
    pub actions: Vec<String>,
    /// Code patterns to look for
    pub code_patterns: Vec<String>,
    /// Evidence from the profile
    pub evidence: Vec<String>,
}

impl Recommendation {
    /// Calculate the ROI score (savings / effort)
    pub fn roi_score(&self) -> f64 {
        let effort_multiplier = match self.effort {
            Effort::QuickWin => 4.0,
            Effort::Moderate => 2.0,
            Effort::Significant => 1.0,
            Effort::Major => 0.5,
        };
        self.estimated_savings_us as f64 * effort_multiplier
    }

    /// Format estimated savings as percentage
    #[expect(clippy::cast_precision_loss)]
    pub fn savings_percent(&self, total_time: u64) -> f64 {
        if total_time == 0 {
            0.0
        } else {
            (self.estimated_savings_us as f64 / total_time as f64) * 100.0
        }
    }
}

/// Complete set of recommendations for a profile
#[derive(Debug, Clone)]
pub struct RecommendationReport {
    /// All recommendations, sorted by priority then ROI
    pub recommendations: Vec<Recommendation>,
    /// Summary insights
    pub insights: Vec<String>,
    /// Quick wins (high impact, low effort)
    pub quick_wins: Vec<usize>, // indices into recommendations
    /// Investigation items (need more analysis)
    pub investigations: Vec<String>,
}

/// Generates recommendations from CPU analysis
pub struct RecommendationEngine;

impl RecommendationEngine {
    /// Check if a category should be included based on filters
    fn should_include_category(category: FrameCategory, analysis: &CpuAnalysis) -> bool {
        let filters = &analysis.metadata.filter_categories;
        // If no category filter, include all
        // If category filter is set, only include matching categories
        filters.is_empty() || filters.contains(&category)
    }

    /// Generate recommendations from analysis
    #[expect(clippy::cast_precision_loss)]
    pub fn analyze(profile: &ProfileIR, analysis: &CpuAnalysis) -> RecommendationReport {
        let mut recommendations = Vec::new();
        let mut insights = Vec::new();
        let mut investigations = Vec::new();

        let total_time = analysis.total_time;
        let filters = &analysis.metadata.filter_categories;

        // 1. Analyze top hotspots (always, but functions are already filtered)
        Self::analyze_hotspots(analysis, total_time, &mut recommendations, &mut insights);

        // 2. Analyze GC pressure (skip if filtering to specific categories that don't include V8Internal)
        let should_analyze_gc = filters.is_empty()
            || filters.contains(&FrameCategory::V8Internal)
            || filters.contains(&FrameCategory::App)  // GC affects app code too
            || filters.contains(&FrameCategory::Deps);
        if should_analyze_gc {
            Self::analyze_gc(analysis, total_time, &mut recommendations, &mut insights);
        }

        // 3. Analyze dependencies (skip if filtering to non-Deps categories)
        if Self::should_include_category(FrameCategory::Deps, analysis) {
            Self::analyze_dependencies(analysis, total_time, &mut recommendations, &mut insights);
        }

        // 4. Analyze recursion (uses filtered functions)
        Self::analyze_recursion(analysis, total_time, &mut recommendations);

        // 5. Analyze startup vs steady state
        Self::analyze_phases(analysis, total_time, &mut recommendations, &mut insights);

        // 6. Detect patterns from function names (uses filtered functions)
        Self::detect_patterns(profile, analysis, total_time, &mut recommendations);

        // 7. Generate investigation items (respect filters)
        Self::generate_investigations(analysis, &mut investigations);

        // Filter out recommendations with negligible impact or bad ROI
        // - Must save at least 0.5% of total time to be worth showing
        // - Don't recommend "Significant" effort for <2% savings
        let min_savings_threshold = total_time / 200; // 0.5%
        recommendations.retain(|r| {
            let savings_pct = r.savings_percent(total_time);
            // Keep if savings are meaningful
            if r.estimated_savings_us < min_savings_threshold {
                return false;
            }
            // Don't show "Significant" or "Major" effort for tiny gains
            if matches!(r.effort, Effort::Significant | Effort::Major) && savings_pct < 2.0 {
                return false;
            }
            true
        });

        // Sort recommendations by priority, then by ROI
        recommendations.sort_by(|a, b| {
            a.priority.cmp(&b.priority).then_with(|| {
                b.roi_score()
                    .partial_cmp(&a.roi_score())
                    .unwrap_or(std::cmp::Ordering::Equal)
            })
        });

        // Identify quick wins (meaningful impact + low effort)
        let quick_wins: Vec<usize> = recommendations
            .iter()
            .enumerate()
            .filter(|(_, r)| r.effort == Effort::QuickWin && r.savings_percent(total_time) >= 1.0)
            .map(|(i, _)| i)
            .collect();

        RecommendationReport {
            recommendations,
            insights,
            quick_wins,
            investigations,
        }
    }

    #[expect(clippy::cast_precision_loss)]
    fn analyze_hotspots(
        analysis: &CpuAnalysis,
        total_time: u64,
        recommendations: &mut Vec<Recommendation>,
        insights: &mut Vec<String>,
    ) {
        for func in analysis.functions.iter().take(10) {
            let self_pct = func.self_percent(total_time);
            let total_pct = func.total_percent(total_time);

            // Only recommend functions with meaningful impact
            // Require at least 3% self time OR 10% inclusive time
            if self_pct < 3.0 && total_pct < 10.0 {
                continue;
            }

            let pattern = func.performance_pattern(analysis.total_samples);
            let (priority, issue_type, root_cause, actions, effort, savings_ratio) =
                Self::classify_hotspot(func, &pattern, self_pct, total_pct);

            let estimated_savings = (func.self_time as f64 * savings_ratio) as u64;

            let mut evidence = vec![
                format!("{:.1}% of total CPU time", self_pct),
                format!("{} samples", func.self_samples),
            ];

            if func.is_recursive() {
                evidence.push(format!(
                    "Recursive (max depth: {})",
                    func.max_recursion_depth
                ));
            }

            let code_patterns = Self::suggest_code_patterns(&func.name, &func.location, issue_type);

            recommendations.push(Recommendation {
                priority,
                effort,
                issue_type,
                title: Self::generate_title(func, issue_type),
                target: func.name.clone(),
                location: func.location.clone(),
                current_time_us: func.self_time,
                estimated_savings_us: estimated_savings,
                root_cause,
                actions,
                code_patterns,
                evidence,
            });
        }

        // Generate insight about hotspot concentration
        let top3_pct: f64 = analysis
            .functions
            .iter()
            .take(3)
            .map(|f| f.self_percent(total_time))
            .sum();

        if top3_pct > 50.0 {
            insights.push(format!(
                "Top 3 functions account for {:.0}% of CPU time — focused optimization will have high impact",
                top3_pct
            ));
        } else if top3_pct < 20.0 {
            insights.push(
                "CPU time is well-distributed — consider architectural improvements over micro-optimizations".to_string()
            );
        }
    }

    fn classify_hotspot(
        func: &FunctionStats,
        pattern: &PerformancePattern,
        self_pct: f64,
        total_pct: f64,
    ) -> (Priority, IssueType, String, Vec<String>, Effort, f64) {
        match pattern {
            PerformancePattern::CriticalPath => {
                let priority = if self_pct >= 20.0 {
                    Priority::Critical
                } else {
                    Priority::High
                };
                (
                    priority,
                    IssueType::Algorithm,
                    "Called frequently AND expensive per call — this is on the critical path"
                        .to_string(),
                    vec![
                        "Profile this function in isolation to find the slow code path".to_string(),
                        "Consider algorithmic improvements (caching, better data structures)"
                            .to_string(),
                        "Look for unnecessary work that can be skipped".to_string(),
                        "Consider breaking into smaller functions to isolate the bottleneck"
                            .to_string(),
                    ],
                    Effort::Moderate,
                    0.5, // Expect 50% improvement with optimization
                )
            }
            PerformancePattern::ExpensiveOperation => {
                // Only worth significant effort if impact is meaningful
                let (priority, effort) = if self_pct >= 15.0 {
                    (Priority::High, Effort::Moderate)
                } else if self_pct >= 5.0 {
                    (Priority::Medium, Effort::Moderate)
                } else {
                    // Small impact - only worth quick investigation
                    (Priority::Low, Effort::QuickWin)
                };
                (
                    priority,
                    IssueType::Algorithm,
                    "Each call is expensive but infrequent — optimize the operation itself"
                        .to_string(),
                    vec![
                        "Check for O(n²) or worse algorithms".to_string(),
                        "Look for synchronous I/O or blocking operations".to_string(),
                        "Consider lazy evaluation or streaming".to_string(),
                        "Profile memory allocations in this function".to_string(),
                    ],
                    effort,
                    0.4,
                )
            }
            PerformancePattern::FrequentlyCalled => {
                let actions = if func.category == FrameCategory::Deps {
                    vec![
                        "Reduce call frequency by batching operations".to_string(),
                        "Cache results if the function is pure".to_string(),
                        "Consider inlining or replacing with native code".to_string(),
                    ]
                } else {
                    vec![
                        "Memoize results if inputs repeat".to_string(),
                        "Move invariant computations outside loops".to_string(),
                        "Consider batching multiple calls".to_string(),
                        "Check if calls can be eliminated entirely".to_string(),
                    ]
                };
                // Require meaningful impact for higher priority
                let priority = if self_pct >= 10.0 {
                    Priority::High
                } else if self_pct >= 3.0 {
                    Priority::Medium
                } else {
                    Priority::Low
                };
                (
                    priority,
                    IssueType::Caching,
                    "Called very frequently — consider caching or batching".to_string(),
                    actions,
                    Effort::QuickWin,
                    0.3,
                )
            }
            PerformancePattern::Normal => {
                if total_pct >= 20.0 {
                    (
                        Priority::Medium,
                        IssueType::Hotspot,
                        "High inclusive time suggests expensive callees".to_string(),
                        vec![
                            "Examine what this function calls".to_string(),
                            "The bottleneck may be in a callee, not this function".to_string(),
                            "Check the caller/callee analysis for this function".to_string(),
                        ],
                        Effort::Moderate,
                        0.2,
                    )
                } else {
                    (
                        Priority::Low,
                        IssueType::Hotspot,
                        "Minor hotspot".to_string(),
                        vec![
                            "Low priority — optimize only if other issues are resolved".to_string(),
                        ],
                        Effort::Moderate,
                        0.1,
                    )
                }
            }
        }
    }

    fn generate_title(func: &FunctionStats, issue_type: IssueType) -> String {
        match issue_type {
            IssueType::Algorithm => format!("Optimize algorithm in `{}`", func.name),
            IssueType::Caching => format!("Add caching/memoization to `{}`", func.name),
            IssueType::Memory => format!("Reduce allocations in `{}`", func.name),
            IssueType::Dependency => format!("Optimize or replace `{}`", func.name),
            IssueType::Serialization => format!("Optimize serialization in `{}`", func.name),
            IssueType::Startup => format!("Defer or lazy-load `{}`", func.name),
            IssueType::Recursion => format!("Convert `{}` to iterative", func.name),
            IssueType::Hotspot => format!("Investigate `{}`", func.name),
        }
    }

    fn suggest_code_patterns(name: &str, location: &str, issue_type: IssueType) -> Vec<String> {
        let mut patterns = Vec::new();
        let name_lower = name.to_lowercase();
        let loc_lower = location.to_lowercase();

        // JSON patterns
        if name_lower.contains("json")
            || name_lower.contains("parse")
            || name_lower.contains("stringify")
        {
            patterns.push("JSON.parse() / JSON.stringify() calls".to_string());
            patterns.push("Consider streaming JSON parsing for large payloads".to_string());
        }

        // Regex patterns
        if name_lower.contains("regex")
            || name_lower.contains("regexp")
            || name_lower.contains("match")
        {
            patterns.push("Regular expression operations".to_string());
            patterns.push("Compile regex once and reuse, avoid in loops".to_string());
        }

        // Sort/search patterns
        if name_lower.contains("sort")
            || name_lower.contains("find")
            || name_lower.contains("search")
        {
            patterns.push("Sorting or searching operations".to_string());
            patterns.push("Check if data structure supports faster lookups (Map/Set)".to_string());
        }

        // Loop patterns
        if name_lower.contains("each")
            || name_lower.contains("map")
            || name_lower.contains("filter")
        {
            patterns.push("Array iteration methods".to_string());
            patterns
                .push("Consider early termination, or use for loop for performance".to_string());
        }

        // Transform patterns
        if name_lower.contains("transform")
            || name_lower.contains("convert")
            || name_lower.contains("compile")
        {
            patterns.push("Data transformation/compilation".to_string());
            patterns.push("Cache transformation results if inputs repeat".to_string());
        }

        // Path patterns
        if loc_lower.contains("lodash") {
            patterns.push("Consider native alternatives to lodash functions".to_string());
        }
        if loc_lower.contains("moment") {
            patterns.push("Consider lighter date libraries (date-fns, dayjs)".to_string());
        }

        // General patterns based on issue type
        match issue_type {
            IssueType::Caching => {
                patterns.push("Look for repeated calls with same arguments".to_string());
                patterns.push("Check if results can be memoized with a Map/WeakMap".to_string());
            }
            IssueType::Memory => {
                patterns.push("Look for object/array creation in loops".to_string());
                patterns.push("Consider object pooling or reuse".to_string());
            }
            _ => {}
        }

        patterns
    }

    #[expect(clippy::cast_precision_loss)]
    fn analyze_gc(
        analysis: &CpuAnalysis,
        total_time: u64,
        recommendations: &mut Vec<Recommendation>,
        insights: &mut Vec<String>,
    ) {
        let Some(gc) = &analysis.gc_analysis else {
            return;
        };

        let gc_pct = if total_time > 0 {
            (gc.total_time as f64 / total_time as f64) * 100.0
        } else {
            0.0
        };

        if gc_pct < 5.0 {
            return;
        }

        let priority = if gc_pct >= 15.0 {
            Priority::Critical
        } else if gc_pct >= 10.0 {
            Priority::High
        } else {
            Priority::Medium
        };

        // Target 2% GC as "normal"
        let target_gc_time = (total_time as f64 * 0.02) as u64;
        let potential_savings = gc.total_time.saturating_sub(target_gc_time);

        insights.push(format!(
            "GC overhead is {:.1}% — reducing allocations could save {:.1}ms",
            gc_pct,
            potential_savings as f64 / 1000.0
        ));

        // Create recommendation for top allocation hotspot
        if let Some(hotspot) = gc.allocation_hotspots.first() {
            let hotspot_savings =
                (potential_savings as f64 * hotspot.gc_correlation / 100.0) as u64;

            recommendations.push(Recommendation {
                priority,
                effort: Effort::Moderate,
                issue_type: IssueType::Memory,
                title: format!("Reduce allocations in `{}`", hotspot.name),
                target: hotspot.name.clone(),
                location: hotspot.location.clone(),
                current_time_us: gc.total_time,
                estimated_savings_us: hotspot_savings,
                root_cause: format!(
                    "This function appears in {:.0}% of GC samples, indicating heavy allocation",
                    hotspot.gc_correlation
                ),
                actions: vec![
                    "Reuse objects instead of creating new ones".to_string(),
                    "Use object pools for frequently created objects".to_string(),
                    "Avoid creating closures in loops".to_string(),
                    "Pre-allocate arrays with known size".to_string(),
                    "Use TypedArrays for numeric data".to_string(),
                ],
                code_patterns: vec![
                    "new Object() / {} literals in loops".to_string(),
                    "Array.push() in tight loops (pre-allocate instead)".to_string(),
                    "String concatenation (use array.join or template literals)".to_string(),
                    "Spread operator creating copies".to_string(),
                ],
                evidence: vec![
                    format!("{:.1}% GC overhead", gc_pct),
                    format!("{} GC events", gc.sample_count),
                    format!("~{:.0}μs average GC pause", gc.avg_pause_us),
                ],
            });
        }
    }

    #[expect(clippy::cast_precision_loss)]
    fn analyze_dependencies(
        analysis: &CpuAnalysis,
        total_time: u64,
        recommendations: &mut Vec<Recommendation>,
        insights: &mut Vec<String>,
    ) {
        let deps_pct = analysis.category_breakdown.percent(FrameCategory::Deps);

        if deps_pct >= 40.0 {
            insights.push(format!(
                "Dependencies consume {:.0}% of CPU — review if all are necessary",
                deps_pct
            ));
        }

        for pkg in analysis.package_stats.iter().take(3) {
            let pkg_pct = if total_time > 0 {
                (pkg.time as f64 / total_time as f64) * 100.0
            } else {
                0.0
            };

            if pkg_pct < 3.0 {
                continue;
            }

            let (actions, effort, savings_ratio) = Self::get_package_advice(&pkg.package);

            recommendations.push(Recommendation {
                priority: if pkg_pct >= 15.0 {
                    Priority::High
                } else {
                    Priority::Medium
                },
                effort,
                issue_type: IssueType::Dependency,
                title: format!("Optimize `{}` usage", pkg.package),
                target: pkg.package.clone(),
                location: pkg.top_function_location.clone(),
                current_time_us: pkg.time,
                estimated_savings_us: (pkg.time as f64 * savings_ratio) as u64,
                root_cause: format!(
                    "Package `{}` consumes {:.1}% of CPU time",
                    pkg.package, pkg_pct
                ),
                actions,
                code_patterns: vec![
                    format!("import {{ ... }} from '{}'", pkg.package),
                    format!("require('{}')", pkg.package),
                ],
                evidence: vec![
                    format!("{:.1}% of total CPU time", pkg_pct),
                    format!("{:.1}% of dependency time", pkg.percent_of_deps),
                    format!("Hottest function: {}", pkg.top_function),
                ],
            });
        }
    }

    fn get_package_advice(package: &str) -> (Vec<String>, Effort, f64) {
        let pkg_lower = package.to_lowercase();

        // Known heavy packages with specific advice
        if pkg_lower.contains("lodash") {
            return (
                vec![
                    "Import only needed functions: `import map from 'lodash/map'`".to_string(),
                    "Consider native alternatives (Array.map, Object.keys, etc.)".to_string(),
                    "Use lodash-es for better tree-shaking".to_string(),
                ],
                Effort::QuickWin,
                0.3,
            );
        }

        if pkg_lower.contains("moment") {
            return (
                vec![
                    "Replace with date-fns or dayjs (10-20x smaller)".to_string(),
                    "Use native Intl.DateTimeFormat for formatting".to_string(),
                    "Avoid parsing strings repeatedly".to_string(),
                ],
                Effort::Moderate,
                0.5,
            );
        }

        if pkg_lower.contains("axios") {
            return (
                vec![
                    "Consider native fetch() API".to_string(),
                    "Reuse axios instances".to_string(),
                    "Check if interceptors add overhead".to_string(),
                ],
                Effort::Moderate,
                0.2,
            );
        }

        if pkg_lower.contains("babel") || pkg_lower.contains("typescript") {
            return (
                vec![
                    "This is build-time overhead — ensure not running in production".to_string(),
                    "Pre-compile code instead of runtime transpilation".to_string(),
                    "Check for accidental ts-node or @babel/register in prod".to_string(),
                ],
                Effort::QuickWin,
                0.8,
            );
        }

        if pkg_lower.contains("webpack")
            || pkg_lower.contains("esbuild")
            || pkg_lower.contains("vite")
        {
            return (
                vec![
                    "Build tools should not run in production".to_string(),
                    "Check for dev dependencies imported at runtime".to_string(),
                ],
                Effort::QuickWin,
                0.9,
            );
        }

        if pkg_lower.contains("ajv") || pkg_lower.contains("joi") || pkg_lower.contains("yup") {
            return (
                vec![
                    "Compile schemas once, reuse validators".to_string(),
                    "Consider lighter validation for hot paths".to_string(),
                    "Skip validation in trusted internal calls".to_string(),
                ],
                Effort::QuickWin,
                0.4,
            );
        }

        // Generic advice
        (
            vec![
                "Check if this package is necessary".to_string(),
                "Look for lighter alternatives".to_string(),
                "Consider lazy-loading if not needed at startup".to_string(),
            ],
            Effort::Moderate,
            0.3,
        )
    }

    #[expect(clippy::cast_precision_loss)]
    fn analyze_recursion(
        analysis: &CpuAnalysis,
        total_time: u64,
        recommendations: &mut Vec<Recommendation>,
    ) {
        for rec_func in &analysis.recursive_functions {
            let rec_pct = if total_time > 0 {
                (rec_func.recursive_time as f64 / total_time as f64) * 100.0
            } else {
                0.0
            };

            if rec_pct < 3.0 || rec_func.max_depth < 5 {
                continue;
            }

            recommendations.push(Recommendation {
                priority: if rec_pct >= 10.0 { Priority::High } else { Priority::Medium },
                effort: Effort::Moderate,
                issue_type: IssueType::Recursion,
                title: format!("Convert `{}` to iterative", rec_func.name),
                target: rec_func.name.clone(),
                location: rec_func.location.clone(),
                current_time_us: rec_func.recursive_time,
                estimated_savings_us: rec_func.recursive_time / 3, // ~30% savings from removing call overhead
                root_cause: format!(
                    "Deep recursion (max depth {}) causes stack overhead and potential stack overflow risk",
                    rec_func.max_depth
                ),
                actions: vec![
                    "Convert to iterative algorithm with explicit stack".to_string(),
                    "Consider tail-call optimization if applicable".to_string(),
                    "Add memoization to avoid redundant recursive calls".to_string(),
                    "Limit recursion depth with an iterative fallback".to_string(),
                ],
                code_patterns: vec![
                    "function f() { ... f() ... }".to_string(),
                    "Look for tree/graph traversal".to_string(),
                ],
                evidence: vec![
                    format!("Max recursion depth: {}", rec_func.max_depth),
                    format!("{} recursive samples", rec_func.recursive_samples),
                    format!("{:.1}% of CPU time", rec_pct),
                ],
            });
        }
    }

    #[expect(clippy::cast_precision_loss)]
    fn analyze_phases(
        analysis: &CpuAnalysis,
        total_time: u64,
        recommendations: &mut Vec<Recommendation>,
        insights: &mut Vec<String>,
    ) {
        let Some(phases) = &analysis.phase_analysis else {
            return;
        };

        let startup_time = phases.startup.end_us - phases.startup.start_us;
        let startup_pct = if total_time > 0 {
            (startup_time as f64 / total_time as f64) * 100.0
        } else {
            0.0
        };

        if startup_pct >= 30.0 && startup_time > 500_000 {
            // >500ms startup
            insights.push(format!(
                "Startup phase takes {:.0}ms ({:.0}% of profile) — consider lazy loading",
                startup_time as f64 / 1000.0,
                startup_pct
            ));

            if let Some(top_startup_func) = phases.startup.top_functions.first() {
                if top_startup_func.percent >= 20.0 {
                    recommendations.push(Recommendation {
                        priority: Priority::High,
                        effort: Effort::Moderate,
                        issue_type: IssueType::Startup,
                        title: format!("Defer `{}` initialization", top_startup_func.name),
                        target: top_startup_func.name.clone(),
                        location: top_startup_func.location.clone(),
                        current_time_us: top_startup_func.self_time,
                        estimated_savings_us: top_startup_func.self_time * 8 / 10, // 80% if fully deferred
                        root_cause: format!(
                            "This function takes {:.0}% of startup time",
                            top_startup_func.percent
                        ),
                        actions: vec![
                            "Lazy-load this module on first use".to_string(),
                            "Move initialization to background/idle time".to_string(),
                            "Consider code-splitting this functionality".to_string(),
                            "Defer non-critical initialization".to_string(),
                        ],
                        code_patterns: vec![
                            "Top-level await or sync initialization".to_string(),
                            "Large imports at module load time".to_string(),
                        ],
                        evidence: vec![
                            format!("{:.1}% of startup time", top_startup_func.percent),
                            format!("Startup phase: {:.0}ms", startup_time as f64 / 1000.0),
                        ],
                    });
                }
            }
        }
    }

    #[expect(clippy::cast_precision_loss)]
    fn detect_patterns(
        profile: &ProfileIR,
        analysis: &CpuAnalysis,
        total_time: u64,
        recommendations: &mut Vec<Recommendation>,
    ) {
        // Detect JSON serialization hotspots
        let json_funcs: Vec<_> = analysis
            .functions
            .iter()
            .filter(|f| {
                let name_lower = f.name.to_lowercase();
                name_lower.contains("json")
                    || name_lower.contains("parse")
                    || name_lower.contains("stringify")
                    || name_lower.contains("serialize")
            })
            .collect();

        let json_time: u64 = json_funcs.iter().map(|f| f.self_time).sum();
        let json_pct = if total_time > 0 {
            (json_time as f64 / total_time as f64) * 100.0
        } else {
            0.0
        };

        if json_pct >= 5.0 {
            recommendations.push(Recommendation {
                priority: if json_pct >= 15.0 {
                    Priority::High
                } else {
                    Priority::Medium
                },
                effort: Effort::Moderate,
                issue_type: IssueType::Serialization,
                title: "Optimize JSON serialization".to_string(),
                target: "JSON operations".to_string(),
                location: json_funcs
                    .first()
                    .map_or("(multiple)".to_string(), |f| f.location.clone()),
                current_time_us: json_time,
                estimated_savings_us: json_time / 2,
                root_cause: format!(
                    "JSON parsing/serialization consumes {:.1}% of CPU time",
                    json_pct
                ),
                actions: vec![
                    "Use streaming JSON parsing for large payloads".to_string(),
                    "Consider binary formats (MessagePack, Protocol Buffers)".to_string(),
                    "Cache parsed results when possible".to_string(),
                    "Avoid stringify/parse roundtrips for cloning (use structuredClone)"
                        .to_string(),
                ],
                code_patterns: vec![
                    "JSON.parse(JSON.stringify(obj)) for cloning".to_string(),
                    "Repeated parsing of same data".to_string(),
                    "Large object serialization".to_string(),
                ],
                evidence: json_funcs
                    .iter()
                    .take(3)
                    .map(|f| format!("`{}` - {:.1}%", f.name, f.self_percent(total_time)))
                    .collect(),
            });
        }

        // Detect regex hotspots
        let regex_funcs: Vec<_> = analysis
            .functions
            .iter()
            .filter(|f| {
                let name_lower = f.name.to_lowercase();
                name_lower.contains("regexp")
                    || name_lower.contains("regex")
                    || f.name.contains("match")
            })
            .collect();

        let regex_time: u64 = regex_funcs.iter().map(|f| f.self_time).sum();
        let regex_pct = if total_time > 0 {
            (regex_time as f64 / total_time as f64) * 100.0
        } else {
            0.0
        };

        if regex_pct >= 3.0 {
            recommendations.push(Recommendation {
                priority: Priority::Medium,
                effort: Effort::QuickWin,
                issue_type: IssueType::Algorithm,
                title: "Optimize regular expressions".to_string(),
                target: "RegExp operations".to_string(),
                location: regex_funcs
                    .first()
                    .map_or("(multiple)".to_string(), |f| f.location.clone()),
                current_time_us: regex_time,
                estimated_savings_us: regex_time * 2 / 3,
                root_cause: format!("Regular expressions consume {:.1}% of CPU time", regex_pct),
                actions: vec![
                    "Compile regex once outside loops: `const re = /pattern/`".to_string(),
                    "Use simpler string methods when possible (includes, startsWith)".to_string(),
                    "Avoid capturing groups if not needed: `(?:...)` instead of `(...)`"
                        .to_string(),
                    "Consider using non-backtracking patterns".to_string(),
                ],
                code_patterns: vec![
                    "/pattern/.test(str) inside loops".to_string(),
                    "new RegExp() called repeatedly".to_string(),
                    "Complex patterns with backtracking".to_string(),
                ],
                evidence: regex_funcs
                    .iter()
                    .take(3)
                    .map(|f| format!("`{}` - {:.1}%", f.name, f.self_percent(total_time)))
                    .collect(),
            });
        }

        // Detect potential N+1 patterns (same function called many times)
        for func in analysis.functions.iter().take(5) {
            if func.self_samples > 100 && func.avg_time_per_sample() < 100.0 {
                // Many calls, cheap per call
                let name_lower = func.name.to_lowercase();
                if name_lower.contains("get")
                    || name_lower.contains("fetch")
                    || name_lower.contains("load")
                    || name_lower.contains("query")
                    || name_lower.contains("find")
                {
                    let _ = profile; // Silence unused warning
                    recommendations.push(Recommendation {
                        priority: Priority::Medium,
                        effort: Effort::Moderate,
                        issue_type: IssueType::Caching,
                        title: format!("Batch `{}` calls", func.name),
                        target: func.name.clone(),
                        location: func.location.clone(),
                        current_time_us: func.self_time,
                        estimated_savings_us: func.self_time / 2,
                        root_cause: format!(
                            "Called {} times — potential N+1 pattern",
                            func.self_samples
                        ),
                        actions: vec![
                            "Batch multiple calls into a single operation".to_string(),
                            "Use DataLoader pattern for automatic batching".to_string(),
                            "Prefetch data instead of loading on demand".to_string(),
                            "Add caching layer to avoid repeated fetches".to_string(),
                        ],
                        code_patterns: vec![
                            "Loop calling getData(id) — batch to getData(ids)".to_string(),
                            "Multiple awaits in sequence that could be parallel".to_string(),
                        ],
                        evidence: vec![
                            format!("{} calls in profile", func.self_samples),
                            format!("{:.0}μs average per call", func.avg_time_per_sample()),
                        ],
                    });
                    break; // Only one N+1 recommendation
                }
            }
        }
    }

    fn generate_investigations(analysis: &CpuAnalysis, investigations: &mut Vec<String>) {
        // Check for unusual patterns that need manual investigation
        // Only add investigations for categories that are being shown

        // Native code investigation (only if Native category is visible)
        if Self::should_include_category(FrameCategory::Native, analysis)
            && analysis.native_time > analysis.total_time / 5
        {
            investigations.push(
                "High native code time (>20%) — check native addons or C++ bindings".to_string(),
            );
        }

        // V8 internal investigation (only if V8Internal category is visible)
        if Self::should_include_category(FrameCategory::V8Internal, analysis)
            && analysis.category_breakdown.v8_internal > analysis.total_time / 10
        {
            investigations.push(
                "Significant V8 internal time (>10%) — may indicate JIT deoptimization".to_string(),
            );
        }

        if analysis.hot_paths.len() == 1 {
            investigations
                .push("Single dominant code path — check if other paths are expected".to_string());
        }

        if analysis.functions.is_empty() {
            investigations.push(
                "No significant functions found — profile may be too short or app is I/O bound"
                    .to_string(),
            );
        }
    }
}