@insiderfinance/totalfinance 0.1.0

package/modules/math/dist/linalg.d.ts

@@ -0,0 +1,251 @@
+/**
+ * Dense linear algebra for small matrices (spec §8.3).
+ *
+ * Cholesky factorization, a symmetric Jacobi eigensolver, covariance-matrix construction, and
+ * nearest positive-semidefinite / correlation repair (eigenvalue clipping). Pure TypeScript
+ * reference implementations; acceleration is a later concern.
+ */
+import { type Diagnostics } from '../../core/dist/index.js';
+import type { PopulationOptions } from './statistics.js';
+export type Matrix = number[][];
+/** Cholesky factorization: lower-triangular `L` with `A = L Lᵀ`. Throws if `A` is not PD. */
+export declare function cholesky(matrix: Matrix): Matrix;
+/** Solve `A x = b` given the Cholesky factor `L` of `A`. */
+export declare function choleskySolve(L: Matrix, b: number[]): number[];
+/** Covariance matrix of a set of equal-length variables (`series[k]` is variable k). */
+export declare function covarianceMatrix(series: number[][], options?: PopulationOptions): Matrix;
+/** Options for {@link ledoitWolfShrinkage}. */
+export interface LedoitWolfOptions {
+    /**
+     * Shrinkage target: `'identity'` (the scaled identity `μI`, the default and original),
+     * `'constant-correlation'` (the Ledoit–Wolf 2004 target — each variable's own variance with the
+     * **average** sample correlation on every off-diagonal, which fits correlated asset returns far better), or
+     * `'single-index'` (the Ledoit–Wolf 2003 market-model target — a one-factor structure built from each
+     * variable's covariance with the equal-weighted market, the best-specified target for stock returns).
+     */
+    target?: 'identity' | 'constant-correlation' | 'single-index';
+    /**
+     * Observed market return series (length `T`, chronological) for the `'single-index'` target.
+     * When omitted the equal-weight cross-sectional average of `series` is used as the market proxy
+     * (Ledoit–Wolf 2003's construction); {@link estimateCovariance} discloses which was used.
+     */
+    market?: ArrayLike<number>;
+}
+/** The Ledoit–Wolf shrunk covariance and the intensity that produced it. */
+export interface LedoitWolfResult {
+    /** Shrunk covariance `Σ* = δ·F + (1−δ)·S` (`F` = the chosen target) — symmetric positive-definite even when `T < p`. */
+    covariance: Matrix;
+    /** Optimal shrinkage intensity `δ ∈ [0, 1]`. */
+    shrinkage: number;
+    /** Raw sample covariance `S` (`1/T` / MLE normalization). */
+    sampleCovariance: Matrix;
+    /** Average variance `μ = tr(S)/p` — the scaled-identity target's diagonal. */
+    averageVariance: number;
+    /** Number of observations `T`. */
+    observations: number;
+    /** The shrinkage target used. */
+    target: 'identity' | 'constant-correlation' | 'single-index';
+    /** Average off-diagonal sample correlation `r̄` — the constant-correlation target's correlation (that target only). */
+    averageCorrelation?: number;
+    /** Variance of the equal-weighted market proxy `varmkt` — the single-index target's factor variance (that target only). */
+    marketVariance?: number;
+}
+export declare function ledoitWolfShrinkage(series: number[][], options?: LedoitWolfOptions): LedoitWolfResult;
+/** Covariance estimator selector for {@link estimateCovariance}. */
+export type CovarianceMethod = 'sample' | 'ledoit-wolf' | 'ridge' | 'ewma' | 'auto';
+/** Options for {@link estimateCovariance}. */
+export interface EstimateCovarianceOptions {
+    /** Estimator; default `'auto'`. */
+    method?: CovarianceMethod;
+    /** Ridge diagonal load as a fraction of the average variance (method `'ridge'`). Default 0.1; > 0. */
+    ridge?: number;
+    /** Condition-number threshold above which `'auto'` shrinks rather than using the sample. Default 1e4. */
+    conditionThreshold?: number;
+    /** Population (`1/T`) vs sample (`1/(T−1)`) normalization. Default `true` (matches Ledoit–Wolf). */
+    population?: boolean;
+    /** EWMA decay `0 < λ < 1` (method `'ewma'`). Default 0.94 (RiskMetrics). */
+    lambda?: number;
+    /** EWMA half-life in periods (method `'ewma'`); `λ = 2^{−1/halfLife}` when `lambda` is omitted. */
+    halfLife?: number;
+    /**
+     * Ledoit–Wolf shrinkage target when the method resolves to `'ledoit-wolf'` (explicitly or via `'auto'`).
+     * `'identity'` (default), `'constant-correlation'` (fits correlated returns better), or `'single-index'`
+     * (the market-model target, best-specified for stock returns).
+     */
+    ledoitWolfTarget?: 'identity' | 'constant-correlation' | 'single-index';
+    /**
+     * Observed market return series for the `'single-index'` target (length `T`). Omitted → the
+     * equal-weight cross-sectional proxy is used and disclosed via `assumptions.marketProxy`.
+     */
+    market?: ArrayLike<number>;
+}
+/** A well-conditioned covariance estimate plus its conditioning diagnostics. */
+export interface CovarianceEstimate {
+    /** The estimated covariance. */
+    covariance: Matrix;
+    /** The estimator actually used (`'auto'` resolves to one of these). */
+    method: 'sample' | 'ledoit-wolf' | 'ridge' | 'ewma';
+    /** Number of variables `p`. */
+    variables: number;
+    /** Number of observations `T`. */
+    observations: number;
+    /** Smallest eigenvalue of the returned covariance. */
+    minEigenvalue: number;
+    /** Largest eigenvalue of the returned covariance. */
+    maxEigenvalue: number;
+    /** `maxEigenvalue / minEigenvalue` (`Infinity` when singular). */
+    conditionNumber: number;
+    /** Whether the returned covariance is positive-definite (invertible). */
+    isPositiveDefinite: boolean;
+    /** Number of eigenvalues `> 1e-9 · maxEigenvalue` — the effective rank. */
+    effectiveRank: number;
+    /** Ledoit–Wolf shrinkage intensity, present when `method` is `'ledoit-wolf'`. */
+    shrinkage?: number;
+    /** The EWMA decay used — present only for method `'ewma'`. */
+    lambda?: number;
+    /** Kish effective sample size `1/Σwₜ²` — present only for method `'ewma'`. */
+    effectiveObservations?: number;
+    assumptions: {
+        conventionsVersion: string;
+        method: string;
+        population: boolean;
+        /** Single-index target only: the caller's observed market series vs the equal-weight proxy. */
+        marketProxy?: 'observed-series' | 'equal-weight';
+    };
+    diagnostics: Diagnostics;
+}
+/**
+ * The covariance-estimation front door: turn a returns matrix into a well-conditioned (usually
+ * invertible) covariance and report its **conditioning** so a caller knows whether it is safe to invert.
+ * `'sample'` and `'ewma'` are the raw estimators (never altered; flagged + `converged: false` when
+ * singular); `'ledoit-wolf'` and `'ridge'` GUARANTEE an SPD result — degenerate/collinear inputs are
+ * eigenvalue-floored with a `math.covariance_floored` warning rather than returned singular; `'auto'`
+ * (default) uses the sample when it is SPD and well-conditioned, else shrinks (inheriting the SPD
+ * guarantee). `series[k]` is variable
+ * k's observations (p × T), matching {@link covarianceMatrix}. See `docs/specs/estimate-covariance.md`.
+ */
+export declare function estimateCovariance(series: number[][], options?: EstimateCovarianceOptions): CovarianceEstimate;
+export interface EigenResult {
+    /** Eigenvalues (not sorted). */
+    values: number[];
+    /** Eigenvectors as columns of the matrix. */
+    vectors: Matrix;
+}
+/**
+ * Symmetric eigendecomposition via cyclic Jacobi rotations. The tolerance is RELATIVE to the input's
+ * off-diagonal Frobenius norm, so a matrix scaled by any constant converges identically (an absolute
+ * tolerance would spin to `maximumIterations` on a large-scaled covariance matrix). Throws `ConvergenceError`
+ * (`linalg.no_convergence`) if the sweeps do not reach the tolerance within `maximumIterations`.
+ *
+ * The input must be SQUARE and finite. It used to be neither checked nor needed: every loop is
+ * indexed by `input.length`, so a `2×3` matrix quietly returned the eigenvalues of its leading `2×2`
+ * block — the right shape, the wrong matrix, and no way for the caller to tell. (Non-symmetric square
+ * input is still accepted and still reports non-convergence through `ConvergenceError` when the
+ * rotations cannot diagonalize it; use {@link eigenvalues} for the general case.)
+ */
+export declare function jacobiEigen(input: Matrix, options?: {
+    maximumIterations?: number;
+    tolerance?: number;
+}): EigenResult;
+/** Nearest positive-semidefinite matrix (eigenvalue clipping). */
+export declare function nearestPsd(matrix: Matrix): Matrix;
+/**
+ * A valid correlation matrix near `matrix`: ONE eigenvalue-clipping pass to PSD ({@link nearestPsd}),
+ * then a rescale to a unit diagonal.
+ *
+ * Not the Frobenius-NEAREST correlation matrix. That is Higham's (2002) alternating-projections
+ * problem, iterated to convergence between the PSD cone and the unit-diagonal set; the rescale here
+ * is a single projection that can push the result slightly off the PSD cone again (the output is PSD
+ * to within the rescale, not by construction). It is the standard cheap repair — good enough to make
+ * a noisy correlation estimate usable for Cholesky/simulation — and it is deterministic and O(n³),
+ * but a caller who needs the true minimizer needs Higham's iteration, not this.
+ */
+export declare function nearestCorrelation(matrix: Matrix): Matrix;
+/** The `n × n` identity matrix. */
+export declare function identity(size: number): Matrix;
+/** Transpose of an `m × n` matrix → `n × m`. */
+export declare function transpose(matrix: Matrix): Matrix;
+/** Matrix product `A (m×k) · B (k×n)` → `m×n`. Throws on a dimension mismatch. */
+export declare function matrixMultiply(left: Matrix, right: Matrix): Matrix;
+/** Matrix–vector product `A (m×n) · x (n)` → `m`. */
+export declare function matrixVectorProduct(matrix: Matrix, vector: number[]): number[];
+export interface LuResult {
+    /** Combined lower/upper factors (unit-diagonal `L` below, `U` on and above the diagonal). */
+    lu: Matrix;
+    /** Row permutation: `pivotIndices[i]` is the source row moved into row `i`. */
+    pivotIndices: number[];
+    /** Sign of the permutation (+1 or −1), for the determinant. */
+    sign: number;
+}
+/** LU decomposition with partial pivoting: `P·A = L·U`. Throws on a singular matrix. */
+export declare function luDecompose(matrix: Matrix): LuResult;
+/** Solve `A·x = b` from a precomputed {@link luDecompose} result. */
+export declare function luSolve(result: LuResult, b: number[]): number[];
+/** Determinant via LU; returns 0 for a singular matrix instead of throwing. */
+export declare function determinant(matrix: Matrix): number;
+export interface QrResult {
+    /** Orthonormal `Q` (`m × n`, thin). */
+    q: Matrix;
+    /** Upper-triangular `R` (`n × n`). */
+    r: Matrix;
+}
+/** Householder QR of an `m × n` matrix (`m ≥ n`): `A = Q·R`, `Q` thin with orthonormal columns. */
+export declare function qrDecompose(matrix: Matrix): QrResult;
+/**
+ * Least-squares solve of `A·x ≈ b` (`m ≥ n`) via QR back-substitution.
+ *
+ * Rejects a rank-deficient (numerically singular) system rather than silently returning zeroed or
+ * blown-up coefficients: if the smallest `|Rᵢᵢ|` pivot falls below `rcond × max|Rᵢᵢ|`, the columns
+ * of `A` are collinear and the least-squares solution is not well defined (design law #4 — no silent
+ * garbage). Raise `rcond` to reject more aggressively; lower it to admit borderline-conditioned fits.
+ */
+export declare function qrSolve(matrix: Matrix, b: number[], options?: {
+    rcond?: number;
+}): number[];
+export interface SvdResult {
+    /** Left singular vectors as columns (`m × n`, thin). */
+    u: Matrix;
+    /** Singular values, descending. */
+    s: number[];
+    /** Right singular vectors as columns (`n × n`). */
+    v: Matrix;
+}
+/**
+ * Thin singular value decomposition `A = U·diag(s)·Vᵀ` via the one-sided Jacobi (Hestenes) method —
+ * accurate and dependency-free for the small matrices TotalFinance handles. `s` is sorted descending.
+ */
+export declare function svd(matrix: Matrix, options?: {
+    maxSweeps?: number;
+    tolerance?: number;
+}): SvdResult;
+/** Moore–Penrose pseudoinverse `A⁺` via SVD (singular values below `rcond·σ_max` are dropped). */
+export declare function pseudoInverse(matrix: Matrix, options?: {
+    rcond?: number;
+}): Matrix;
+/** A complex number `re + im·i`. */
+export interface Complex {
+    re: number;
+    im: number;
+}
+export interface GeneralEigenResult {
+    /** Eigenvalues (real, or complex-conjugate pairs), in real-Schur order. */
+    values: Complex[];
+    /**
+     * Eigenvector for each real eigenvalue (via inverse iteration), aligned to `values`; `null` for
+     * complex eigenvalues (their eigenvectors are complex and not returned in this real interface).
+     */
+    vectors: (number[] | null)[];
+}
+/**
+ * Eigenvalues of a general real matrix via the **Francis double-shift QR** algorithm on its Hessenberg
+ * form (the EISPACK/JAMA `hqr2` reduction to real Schur form). Returns real eigenvalues and
+ * complex-conjugate pairs. For symmetric matrices prefer {@link jacobiEigen} (which also gives vectors).
+ */
+export declare function eigenvalues(matrix: Matrix): Complex[];
+/**
+ * Eigenvalues and (for real eigenvalues) eigenvectors of a general real matrix. Eigenvalues come from
+ * the Francis QR algorithm ({@link eigenvalues}); each real eigenvalue's eigenvector is recovered by
+ * inverse iteration. Complex eigenvalues report a `null` vector in this real-valued interface.
+ */
+export declare function eigen(matrix: Matrix): GeneralEigenResult;
+//# sourceMappingURL=linalg.d.ts.map

package/modules/math/dist/linalg.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"linalg.d.ts","sourceRoot":"","sources":["../src/linalg.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAEH,OAAO,EAKL,KAAK,WAAW,EAMjB,MAAM,0BAAoB,CAAC;AAC5B,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,iBAAiB,CAAC;AAQzD,MAAM,MAAM,MAAM,GAAG,MAAM,EAAE,EAAE,CAAC;AAEhC,6FAA6F;AAC7F,wBAAgB,QAAQ,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,CAwB/C;AAED,4DAA4D;AAC5D,wBAAgB,aAAa,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,EAAE,GAAG,MAAM,EAAE,CAe9D;AAED,wFAAwF;AACxF,wBAAgB,gBAAgB,CAAC,MAAM,EAAE,MAAM,EAAE,EAAE,EAAE,OAAO,CAAC,EAAE,iBAAiB,GAAG,MAAM,CAiBxF;AAsDD,+CAA+C;AAC/C,MAAM,WAAW,iBAAiB;IAChC;;;;;;OAMG;IACH,MAAM,CAAC,EAAE,UAAU,GAAG,sBAAsB,GAAG,cAAc,CAAC;IAC9D;;;;OAIG;IACH,MAAM,CAAC,EAAE,SAAS,CAAC,MAAM,CAAC,CAAC;CAC5B;AAED,4EAA4E;AAC5E,MAAM,WAAW,gBAAgB;IAC/B,wHAAwH;IACxH,UAAU,EAAE,MAAM,CAAC;IACnB,gDAAgD;IAChD,SAAS,EAAE,MAAM,CAAC;IAClB,6DAA6D;IAC7D,gBAAgB,EAAE,MAAM,CAAC;IACzB,8EAA8E;IAC9E,eAAe,EAAE,MAAM,CAAC;IACxB,kCAAkC;IAClC,YAAY,EAAE,MAAM,CAAC;IACrB,iCAAiC;IACjC,MAAM,EAAE,UAAU,GAAG,sBAAsB,GAAG,cAAc,CAAC;IAC7D,uHAAuH;IACvH,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,2HAA2H;IAC3H,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AA6BD,wBAAgB,mBAAmB,CACjC,MAAM,EAAE,MAAM,EAAE,EAAE,EAClB,OAAO,GAAE,iBAAsB,GAC9B,gBAAgB,CAwFlB;AA4PD,oEAAoE;AACpE,MAAM,MAAM,gBAAgB,GAAG,QAAQ,GAAG,aAAa,GAAG,OAAO,GAAG,MAAM,GAAG,MAAM,CAAC;AAEpF,8CAA8C;AAC9C,MAAM,WAAW,yBAAyB;IACxC,mCAAmC;IACnC,MAAM,CAAC,EAAE,gBAAgB,CAAC;IAC1B,sGAAsG;IACtG,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,yGAAyG;IACzG,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,oGAAoG;IACpG,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB,4EAA4E;IAC5E,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,mGAAmG;IACnG,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB;;;;OAIG;IACH,gBAAgB,CAAC,EAAE,UAAU,GAAG,sBAAsB,GAAG,cAAc,CAAC;IACxE;;;OAGG;IACH,MAAM,CAAC,EAAE,SAAS,CAAC,MAAM,CAAC,CAAC;CAC5B;AAED,gFAAgF;AAChF,MAAM,WAAW,kBAAkB;IACjC,gCAAgC;IAChC,UAAU,EAAE,MAAM,CAAC;IACnB,uEAAuE;IACvE,MAAM,EAAE,QAAQ,GAAG,aAAa,GAAG,OAAO,GAAG,MAAM,CAAC;IACpD,+BAA+B;IAC/B,SAAS,EAAE,MAAM,CAAC;IAClB,kCAAkC;IAClC,YAAY,EAAE,MAAM,CAAC;IACrB,sDAAsD;IACtD,aAAa,EAAE,MAAM,CAAC;IACtB,qDAAqD;IACrD,aAAa,EAAE,MAAM,CAAC;IACtB,kEAAkE;IAClE,eAAe,EAAE,MAAM,CAAC;IACxB,yEAAyE;IACzE,kBAAkB,EAAE,OAAO,CAAC;IAC5B,2EAA2E;IAC3E,aAAa,EAAE,MAAM,CAAC;IACtB,iFAAiF;IACjF,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,8DAA8D;IAC9D,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,8EAA8E;IAC9E,qBAAqB,CAAC,EAAE,MAAM,CAAC;IAC/B,WAAW,EAAE;QACX,kBAAkB,EAAE,MAAM,CAAC;QAC3B,MAAM,EAAE,MAAM,CAAC;QACf,UAAU,EAAE,OAAO,CAAC;QACpB,+FAA+F;QAC/F,WAAW,CAAC,EAAE,iBAAiB,GAAG,cAAc,CAAC;KAClD,CAAC;IACF,WAAW,EAAE,WAAW,CAAC;CAC1B;AA8DD;;;;;;;;;GASG;AACH,wBAAgB,kBAAkB,CAChC,MAAM,EAAE,MAAM,EAAE,EAAE,EAClB,OAAO,GAAE,yBAA8B,GACtC,kBAAkB,CA2OpB;AAED,MAAM,WAAW,WAAW;IAC1B,gCAAgC;IAChC,MAAM,EAAE,MAAM,EAAE,CAAC;IACjB,6CAA6C;IAC7C,OAAO,EAAE,MAAM,CAAC;CACjB;AAED;;;;;;;;;;;GAWG;AACH,wBAAgB,WAAW,CACzB,KAAK,EAAE,MAAM,EACb,OAAO,GAAE;IAAE,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAAC,SAAS,CAAC,EAAE,MAAM,CAAA;CAAO,GAC/D,WAAW,CAkGb;AAeD,kEAAkE;AAClE,wBAAgB,UAAU,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,CAIjD;AAED;;;;;;;;;;GAUG;AACH,wBAAgB,kBAAkB,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,CAYzD;AAID,mCAAmC;AACnC,wBAAgB,QAAQ,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,CAI7C;AAED,gDAAgD;AAChD,wBAAgB,SAAS,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,CAMhD;AAED,kFAAkF;AAClF,wBAAgB,cAAc,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,CAsBlE;AAED,qDAAqD;AACrD,wBAAgB,mBAAmB,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,GAAG,MAAM,EAAE,CAoB9E;AAID,MAAM,WAAW,QAAQ;IACvB,6FAA6F;IAC7F,EAAE,EAAE,MAAM,CAAC;IACX,+EAA+E;IAC/E,YAAY,EAAE,MAAM,EAAE,CAAC;IACvB,+DAA+D;IAC/D,IAAI,EAAE,MAAM,CAAC;CACd;AAED,wFAAwF;AACxF,wBAAgB,WAAW,CAAC,MAAM,EAAE,MAAM,GAAG,QAAQ,CA2CpD;AAuFD,qEAAqE;AACrE,wBAAgB,OAAO,CAAC,MAAM,EAAE,QAAQ,EAAE,CAAC,EAAE,MAAM,EAAE,GAAG,MAAM,EAAE,CAmC/D;AAED,+EAA+E;AAC/E,wBAAgB,WAAW,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,CAWlD;AAID,MAAM,WAAW,QAAQ;IACvB,uCAAuC;IACvC,CAAC,EAAE,MAAM,CAAC;IACV,sCAAsC;IACtC,CAAC,EAAE,MAAM,CAAC;CACX;AAED,mGAAmG;AACnG,wBAAgB,WAAW,CAAC,MAAM,EAAE,MAAM,GAAG,QAAQ,CA6CpD;AAED;;;;;;;GAOG;AACH,wBAAgB,OAAO,CAAC,MAAM,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,EAAE,EAAE,OAAO,GAAE;IAAE,KAAK,CAAC,EAAE,MAAM,CAAA;CAAO,GAAG,MAAM,EAAE,CAmC/F;AAID,MAAM,WAAW,SAAS;IACxB,wDAAwD;IACxD,CAAC,EAAE,MAAM,CAAC;IACV,mCAAmC;IACnC,CAAC,EAAE,MAAM,EAAE,CAAC;IACZ,mDAAmD;IACnD,CAAC,EAAE,MAAM,CAAC;CACX;AAED;;;GAGG;AACH,wBAAgB,GAAG,CACjB,MAAM,EAAE,MAAM,EACd,OAAO,GAAE;IAAE,SAAS,CAAC,EAAE,MAAM,CAAC;IAAC,SAAS,CAAC,EAAE,MAAM,CAAA;CAAO,GACvD,SAAS,CA4FX;AAED,kGAAkG;AAClG,wBAAgB,aAAa,CAAC,MAAM,EAAE,MAAM,EAAE,OAAO,GAAE;IAAE,KAAK,CAAC,EAAE,MAAM,CAAA;CAAO,GAAG,MAAM,CA4BtF;AAID,oCAAoC;AACpC,MAAM,WAAW,OAAO;IACtB,EAAE,EAAE,MAAM,CAAC;IACX,EAAE,EAAE,MAAM,CAAC;CACZ;AAED,MAAM,WAAW,kBAAkB;IACjC,2EAA2E;IAC3E,MAAM,EAAE,OAAO,EAAE,CAAC;IAClB;;;OAGG;IACH,OAAO,EAAE,CAAC,MAAM,EAAE,GAAG,IAAI,CAAC,EAAE,CAAC;CAC9B;AAsCD;;;;GAIG;AACH,wBAAgB,WAAW,CAAC,MAAM,EAAE,MAAM,GAAG,OAAO,EAAE,CA4LrD;AA6BD;;;;GAIG;AACH,wBAAgB,KAAK,CAAC,MAAM,EAAE,MAAM,GAAG,kBAAkB,CAOxD"}