Crate ledge_core 

Numerical kernel for Ledge.

Ledge solves convex QPs whose Hessian is a factor covariance plus a diagonal. Its ADMM x-update uses the Sherman-Morrison-Woodbury identity, so it factors a system sized by factors plus explicit linear constraints rather than by the number of assets.

Structs

BatchAccount

One account in a batch: its own portfolio problem plus its per-date steps.

BenchmarkRecord

One measured solver/instance result.

BenchmarkRunner

Executes the same generated instance through registered adapters.

ConvergenceDiagnostics

Heuristic diagnosis of a solve that stopped before convergence.

DualCertificate

An unbounded-descent certificate of dual infeasibility, normalized to unit infinity norm.

DualCertificateResiduals

Independently recomputed residuals of a DualCertificate.

DualVariables

Dual multipliers associated with each constraint block.

FactorQuad

A positive-semidefinite quadratic represented as (Q = F\Omega F^\mathsf{T} + \operatorname{diag}(d)).

GeneratedInstance

Generated problem plus a known feasible reference point.

KktResiduals

Infinity-norm KKT diagnostics.

L1Term

A piecewise-linear proportional cost sum_i costs[i] * |x[i] - anchor[i]| added to the smooth objective (for portfolios: exact L1 turnover around the previous weights).

LinearConstraints

A block of constraints matrix * x = rhs or matrix * x <= rhs.

Matrix

A contiguous row-major dense matrix.

PortfolioProblem

A mean-variance portfolio problem backed by a factor covariance.

PortfolioSequence

A rolling solve sequence over a fixed portfolio structure.

PrimalCertificate

A Farkas certificate of primal infeasibility, normalized to unit infinity norm.

PrimalCertificateResiduals

Independently recomputed residuals of a PrimalCertificate.

QpProblem

Standard-form convex QP accepted by Ledge.

RebalanceStep

Per-date data changes for one PortfolioSequence::solve_next call.

Solution

Solver result and diagnostics.

Solver

Structure-exploiting convex QP solver.

SolverSettings

ADMM settings.

SyntheticConfig

Configuration for a reproducible synthetic portfolio QP.

WarmStart

Optional initial primal and dual iterates.

Workspace

Reusable solve state for a fixed problem structure.

Enums

Certificate

Proof that a solve stopped because the problem itself is pathological.

FactorCovariance

Factor covariance matrix (\Omega).

GeneratorError

Synthetic generator errors.

KktError

Errors from the standalone KKT checker.

MatrixError

Errors raised while constructing a Matrix.

PortfolioError

Errors from high-level portfolio construction and solution.

ProblemError

Input validation errors.

SolveStatus

Solver termination status.

SolverError

Errors that prevent solver setup.

Traits

ComparisonSolver

Solver-neutral adapter for benchmark integrations.

Functions

check_dual_certificate

Recomputes the defining residuals of a dual infeasibility certificate on the original problem data.

check_kkt

Evaluates primal feasibility, stationarity/dual feasibility, and complementarity.

check_primal_certificate

Recomputes the defining residuals of a primal infeasibility certificate on the original problem data.

generate_synthetic

Generates a deterministic, feasible, long-only factor-model portfolio QP.

solve_batch

Solves every account’s rolling sequence, in parallel over the account axis when the rayon cargo feature is enabled (roadmap 3.2).

solve_mean_variance_factor

Solves a factor mean-variance portfolio with optional custom settings.

solve_sequence

Solves a whole rolling sequence in one call: one Solution per step, in order (roadmap 2.5).

Type Aliases

AccountResult

Per-account outcome of solve_batch: all step solutions in order, or the error that stopped that account.