types

type CSVData struct {
	Matrix      Matrix   // The numerical data
	Headers     []string // Column names (if present)
	RowNames    []string // Row names (if present)
	MissingMask [][]bool // Track NaN locations (true = missing)
	Rows        int      // Number of data rows
	Columns     int      // Number of data columns
}

type CSVFormat struct {
	FieldDelimiter   rune     // Field separator: ',', ';', '\t'
	DecimalSeparator rune     // Decimal separator: '.', ','
	HasHeaders       bool     // First row contains column names
	HasRowNames      bool     // First column contains row names
	NullValues       []string // Strings to treat as missing values
}

type CSVParser struct {
	// contains filtered or unexported fields
}

type DiagnosticLimits struct {
	T2Limit95 float64 `json:"t2_limit_95,omitempty"`
	T2Limit99 float64 `json:"t2_limit_99,omitempty"`
	QLimit95  float64 `json:"q_limit_95,omitempty"`
	QLimit99  float64 `json:"q_limit_99,omitempty"`
}

type EigencorrelationResult struct {
	Correlations map[string][]float64 `json:"correlations"` // Variable name -> correlations with each PC
	PValues      map[string][]float64 `json:"pValues"`      // Variable name -> p-values
	Variables    []string             `json:"variables"`    // Order of variables
	Components   []string             `json:"components"`   // PC labels
	Method       string               `json:"method"`       // Correlation method used
}

type EllipseParams struct {
	// Center coordinates of the ellipse (typically the mean of scores)
	CenterX float64
	CenterY float64

	// Semi-major and semi-minor axes lengths
	MajorAxis float64
	MinorAxis float64

	// Rotation angle in radians
	Angle float64

	// Confidence level (e.g., 0.95 for 95% confidence)
	ConfidenceLevel float64
}

type FeatureData struct {
	Names    []string  `json:"names"`    // Feature names from input
	Loadings Matrix    `json:"loadings"` // Loadings (c × k)
	Means    []float64 `json:"means"`    // Original means (k)
	StdDevs  []float64 `json:"stddevs"`  // Original std devs (k)
}

type JSONFloat64 float64

type Matrix [][]float64

type MetricsCalculator interface {
	// CalculateMetrics computes all diagnostic metrics for the given PCA result
	// Parameters:
	//   - result: The PCA result containing scores, loadings, and eigenvalues
	//   - data: The original data matrix (observations x variables)
	//   - config: Configuration options for metrics calculation
	// Returns:
	//   - PCAMetrics containing all calculated metrics
	//   - error if calculation fails
	CalculateMetrics(result *PCAResult, data Matrix, config MetricsConfig) (*PCAMetrics, error)

	// DetectOutliers identifies outliers based on Hotelling's T² statistic
	// Parameters:
	//   - metrics: Previously calculated PCA metrics
	//   - significance: Significance level for outlier detection (e.g., 0.01)
	// Returns:
	//   - Boolean array indicating outliers (true = outlier)
	DetectOutliers(metrics *PCAMetrics, significance float64) []bool

	// CalculateContributions computes variable contributions to each PC
	// Parameters:
	//   - result: The PCA result containing loadings
	//   - data: The original data matrix
	// Returns:
	//   - Matrix of contributions (variables x components)
	CalculateContributions(result *PCAResult, data Matrix) [][]float64
}

type MetricsConfig struct {
	// NumComponents specifies how many components to use for metrics calculation
	// If 0, uses all available components from the PCA result
	NumComponents int

	// SignificanceLevel for outlier detection (e.g., 0.01 for 1% significance)
	// Default is 0.01 if not specified
	SignificanceLevel float64

	// CalculateContributions determines whether to compute variable contributions
	// Default is true
	CalculateContributions bool

	// CalculateConfidenceEllipse determines whether to compute ellipse parameters
	// Default is true
	CalculateConfidenceEllipse bool
}

type MetricsData struct {
	HotellingT2 []float64 `json:"hotelling_t2"`
	Mahalanobis []float64 `json:"mahalanobis"`
	RSS         []float64 `json:"rss"`
	IsOutlier   []bool    `json:"is_outlier"`
}

type MissingValueInfo struct {
	ColumnIndices   []int       // Columns that contain missing values
	RowsAffected    []int       // Rows that contain missing values in selected columns
	TotalMissing    int         // Total number of missing values
	MissingByColumn map[int]int // Missing count per column
}

type MissingValueStrategy string

type ModelComponents struct {
	Loadings               Matrix    `json:"loadings"`
	ExplainedVariance      []float64 `json:"explained_variance"`
	ExplainedVarianceRatio []float64 `json:"explained_variance_ratio"`
	CumulativeVariance     []float64 `json:"cumulative_variance"`
	ComponentLabels        []string  `json:"component_labels"`
	FeatureLabels          []string  `json:"feature_labels"`
}

type ModelConfig struct {
	Method          string               `json:"method"`
	NComponents     int                  `json:"n_components"`
	MissingStrategy MissingValueStrategy `json:"missing_strategy"`
	ExcludedRows    []int                `json:"excluded_rows,omitempty"`
	ExcludedColumns []int                `json:"excluded_columns,omitempty"`
	// Kernel PCA parameters
	KernelType   string  `json:"kernel_type,omitempty"`
	KernelGamma  float64 `json:"kernel_gamma,omitempty"`
	KernelDegree int     `json:"kernel_degree,omitempty"`
	KernelCoef0  float64 `json:"kernel_coef0,omitempty"`
}

type ModelMetadata struct {
	Version   string      `json:"version"`
	CreatedAt string      `json:"created_at"`
	Software  string      `json:"software"`
	Config    ModelConfig `json:"config"`
}

type PCAConfig struct {
	Components      int    `json:"components"`
	MeanCenter      bool   `json:"mean_center"`
	StandardScale   bool   `json:"standard_scale"`
	RobustScale     bool   `json:"robust_scale"`               // Robust scaling (median/MAD)
	ScaleOnly       bool   `json:"scale_only"`                 // Variance scaling: divide by std dev without mean centering
	SNV             bool   `json:"snv"`                        // Standard Normal Variate (row-wise normalization)
	VectorNorm      bool   `json:"vector_norm"`                // L2 normalization (row-wise)
	Method          string `json:"method"`                     // "svd", "eigen", "nipals", or "kernel"
	ExcludedRows    []int  `json:"excluded_rows,omitempty"`    // 0-based indices of rows to exclude
	ExcludedColumns []int  `json:"excluded_columns,omitempty"` // 0-based indices of columns to exclude
	// Missing value handling
	MissingStrategy MissingValueStrategy `json:"missing_strategy,omitempty"` // How to handle missing values
	// Kernel PCA specific parameters
	KernelType   string  `json:"kernel_type,omitempty"`   // "rbf", "linear", "poly"
	KernelGamma  float64 `json:"kernel_gamma,omitempty"`  // RBF/Poly parameter
	KernelDegree int     `json:"kernel_degree,omitempty"` // Poly parameter
	KernelCoef0  float64 `json:"kernel_coef0,omitempty"`  // Poly parameter
}

type PCAEngine interface {
	Fit(data Matrix, config PCAConfig) (*PCAResult, error)
	Transform(data Matrix) (Matrix, error)
	FitTransform(data Matrix, config PCAConfig) (*PCAResult, error)
}

type PCAMetadata struct {
	NSamples           int       `json:"n_samples"`
	NFeatures          int       `json:"n_features"`
	NComponents        int       `json:"n_components"`
	Method             string    `json:"method"`
	Preprocessing      string    `json:"preprocessing"`
	ExplainedVariance  []float64 `json:"explained_variance"`
	CumulativeVariance []float64 `json:"cumulative_variance"`
}

type PCAMetrics struct {
	// MahalanobisDistances contains the Mahalanobis distance for each observation
	// in the transformed PC space, measuring multivariate distance from the mean
	MahalanobisDistances []float64

	// HotellingT2 contains Hotelling's T-squared statistic for each observation,
	// which is used for multivariate outlier detection
	HotellingT2 []float64

	// RSS (Residual Sum of Squares) contains the reconstruction error
	// for each observation when using the specified number of components
	RSS []float64

	// QResiduals contains the Q-statistic (SPE - Squared Prediction Error)
	// for each observation, measuring the lack of fit
	QResiduals []float64

	// OutlierMask indicates which observations are considered outliers
	// based on the specified significance level
	OutlierMask []bool

	// ContributionScores contains the contribution of each variable
	// to each principal component (variables x components)
	ContributionScores [][]float64

	// ConfidenceEllipse contains parameters for drawing confidence ellipses
	// in score plots for outlier visualization
	ConfidenceEllipse EllipseParams
}

type PCAOutputData struct {
	Metadata          ModelMetadata           `json:"metadata"`
	Preprocessing     PreprocessingInfo       `json:"preprocessing"`
	Model             ModelComponents         `json:"model"`
	Results           ResultsData             `json:"results"`
	Diagnostics       DiagnosticLimits        `json:"diagnostics,omitempty"`
	Eigencorrelations *EigencorrelationResult `json:"eigencorrelations,omitempty"`
	PreservedColumns  *PreservedColumns       `json:"preservedColumns,omitempty"`
}

type PCAResult struct {
	Scores               Matrix    `json:"scores"`
	Loadings             Matrix    `json:"loadings"`
	ExplainedVar         []float64 `json:"explained_variance"`
	ExplainedVarRatio    []float64 `json:"explained_variance_ratio"` // Percentage of variance explained
	CumulativeVar        []float64 `json:"cumulative_variance"`
	ComponentLabels      []string  `json:"component_labels"`
	VariableLabels       []string  `json:"variable_labels,omitempty"` // Original variable names
	ComponentsComputed   int       `json:"components_computed"`       // Number of components actually computed
	Method               string    `json:"method"`                    // Method used (svd, nipals, kernel)
	PreprocessingApplied bool      `json:"preprocessing_applied"`     // Whether preprocessing was applied
	// Preprocessing statistics
	Means   []float64 `json:"means,omitempty"`   // Original feature means
	StdDevs []float64 `json:"stddevs,omitempty"` // Original feature std devs
	// Diagnostic metrics
	Metrics []SampleMetrics `json:"metrics,omitempty"` // Per-sample diagnostic metrics
	// Confidence limits for diagnostics
	T2Limit95 float64 `json:"t2_limit_95,omitempty"` // 95% confidence limit for T²
	T2Limit99 float64 `json:"t2_limit_99,omitempty"` // 99% confidence limit for T²
	QLimit95  float64 `json:"q_limit_95,omitempty"`  // 95% confidence limit for Q-residuals
	QLimit99  float64 `json:"q_limit_99,omitempty"`  // 99% confidence limit for Q-residuals
	// Eigencorrelations with metadata
	Eigencorrelations *EigencorrelationResult `json:"eigencorrelations,omitempty"`
	// All eigenvalues (including non-retained) for diagnostic calculations
	AllEigenvalues []float64 `json:"all_eigenvalues,omitempty"`
}

type PreprocessingInfo struct {
	MeanCenter    bool                `json:"mean_center"`
	StandardScale bool                `json:"standard_scale"`
	RobustScale   bool                `json:"robust_scale"`
	ScaleOnly     bool                `json:"scale_only"`
	SNV           bool                `json:"snv"`
	VectorNorm    bool                `json:"vector_norm"`
	Parameters    PreprocessingParams `json:"parameters"`
}

type PreprocessingParams struct {
	FeatureMeans   []float64 `json:"feature_means,omitempty"`
	FeatureStdDevs []float64 `json:"feature_stddevs,omitempty"`
	FeatureMedians []float64 `json:"feature_medians,omitempty"`
	FeatureMADs    []float64 `json:"feature_mads,omitempty"`
	RowMeans       []float64 `json:"row_means,omitempty"`
	RowStdDevs     []float64 `json:"row_stddevs,omitempty"`
}

type PreservedColumns struct {
	Categorical   map[string][]string  `json:"categorical,omitempty"`
	NumericTarget map[string][]float64 `json:"numericTarget,omitempty"`
}

type ResultsData struct {
	Samples SamplesResults `json:"samples"`
}

type SampleData struct {
	Names   []string        `json:"names"`   // Sample names from input
	Scores  Matrix          `json:"scores"`  // PC scores (n × c)
	Metrics []SampleMetrics `json:"metrics"` // Advanced metrics per sample
}

type SampleMetrics struct {
	HotellingT2 float64 `json:"hotelling_t2"`
	Mahalanobis float64 `json:"mahalanobis"`
	RSS         float64 `json:"rss"`
	IsOutlier   bool    `json:"is_outlier"`
}

type SamplesResults struct {
	Names   []string     `json:"names"`
	Scores  Matrix       `json:"scores"`
	Metrics *MetricsData `json:"metrics,omitempty"`
}