mathlib-fp

A collection of focused Free Pascal mathematics domains organised as a single source tree and distribution.

See the roadmap for the quality-first path toward a comprehensive native Free Pascal numerical package.

New to Free Pascal numerical code? Read the beginner guide, then choose a tested task from the beginner recipes. The recipes lead with the double-real, allocating path and link the advanced controls separately.

The cheatsheet is one page of canonical entry points; the task index covers five common tasks per domain.

See the comparison of Pascal numerical libraries for how mathlib-fp relates to other projects. The release maintenance view summarises each domain's stable surface, evidence, and limitations at a glance.

Getting the source

Get the source directly with git:

git clone --depth 1 --branch v1.9.9 \

  https://github.com/ikelaiah/mathlib-fp.git

or as a source archive with wget:

wget \

  https://github.com/ikelaiah/mathlib-fp/archive/refs/tags/v1.9.9.tar.gz

or with curl:

curl -L -o mathlib-fp.tar.gz \

  https://github.com/ikelaiah/mathlib-fp/archive/refs/tags/v1.9.9.tar.gz

> No package manager or install step is required for direct source use; add > src/ to the Free Pascal unit path, for example fpc -Fusrc ....

Direct source use is canonical; the Lazarus package remains a tested, versioned alternative.

Releases

• mathlib-fp 1.9.9 release notes — final 1.9.x convergence: closed 1.10.0 capability manifest, final snapshot and diff, provenance audit, and governance policies.
• 1.9.9 convergence manifest — the closed 1.10.0 declarations, the no-deprecation decision, and explicit deferrals.
• 1.9.9 final snapshot and diff — the final candidate public API and its exact empty compiled diff from 1.9.0.
• 1.9.9 provenance audit — algorithm, fixture, and licence provenance for every stable unit.
• 1.9.9 governance — 2.x maintenance, support, deprecation, new-domain gate, and security window policies.
• 1.9.9 PR notes — implementation and review boundary.
• 1.9.9 qualification report — local and candidate archive evidence status.

• mathlib-fp 1.9.8 release notes — representative multi-domain sensor, modelling, and probability/finance workflows with reproducible clean-archive evidence.
• 1.9.8 workflow qualification — the three qualified end-to-end workflows, their domains, limits, and evidence.
• 1.9.8 PR notes — implementation and review boundary.
• 1.9.8 qualification report — local and candidate archive evidence status.

• mathlib-fp 1.9.7 release notes — executable side-by-side migration, external-library mappings, and alias decisions.
• 1.9.7 migration rehearsal — all-domain consumers, source edits, package boundary, and NumLib/LMath guidance.
• 1.9.7 PR notes — implementation and review boundary.
• 1.9.7 qualification report — local and candidate archive evidence status.

• mathlib-fp 1.9.6 release notes — evidence-backed platform tiers, archive qualification, and offline distribution.
• 1.9.6 portability evidence — native ABI/ binary probe, source-assumption audit, exact target checks, and limitations.
• 1.9.6 PR notes — implementation and review boundary.
• 1.9.6 qualification report — local and exact checksummed clean-archive release gates.

• mathlib-fp 1.9.5 release notes — checked performance and allocation evidence plus a profiled DSP improvement.
• 1.9.5 performance evidence — benchmark conditions, claim-to-row mapping, exact gates, timing review, and limits.
• 1.9.5 PR notes — implementation and review boundary.
• 1.9.5 qualification report — local and clean-archive release gates.

• mathlib-fp 1.9.4 release notes — stable-family numerical evidence, bounded accuracy/failure claims, and offline validation.
• 1.9.4 numerical evidence — evidence categories, budgets, provenance, mutation sampling, and limits.
• 1.9.4 PR notes — review boundary and required checks.
• 1.9.4 qualification report — release gate and clean-archive evidence requirements.

• mathlib-fp 1.9.3 release notes — complete 2.0 API decision, curated common paths, all-domain conventions, compatibility closure, and exact proposed diff.
• 1.9.3 PR notes — review boundary, implementation discipline, completion-gate mapping, and PR-readiness checklist.
• 1.9.3 qualification report — exact declaration, common-program, compatibility, diff, local, and clean-archive gates.
• Curated 2.0 common paths, complete conventions, and exact proposed diff.

• mathlib-fp 1.9.2 release notes — beginner guide, task recipes, stable-domain learning routes, and checked problem-oriented search.
• 1.9.2 PR notes — beginner-learning-path review boundary and automated completion-gate evidence.
• 1.9.2 qualification report — automated preflight, API freeze, and clean-archive requirements.
• 1.9.2 automated beginner journeys — the deterministic evidence replacing external participant collection.

• mathlib-fp 1.9.1 release notes — feature-frozen correctness stabilisation, versioned web/offline documentation, output-aware execution checks, and first-use feedback.
• 1.9.1 PR notes — review boundary, compatibility evidence, and completion-gate mapping.
• 1.9.1 qualification report — normal, optimised, checked/heap, examples, documentation, package, and clean-archive evidence.
• Versioned web documentation and focused 1.9 feedback.
• mathlib-fp 1.9.0 release notes — typed structured/sparse storage, matrix-free Krylov solvers, reusable factors, partial eigensystems, and the candidate-2.0 runway.
• 1.9.0 PR notes — review scope, compatibility, completion-gate mapping, risks, and explicit exclusions.
• 1.9.0 qualification report — dense oracles, termination outcomes, bounded-storage measurements, and target gates.
• 1.9 API audit, generated declaration reference, candidate 2.0 contract, and migration preview.
• mathlib-fp 1.8.0 release notes — applied DSP, bounded streaming statistics, reproducible analysis, state-space filtering, numerical interchange, and portable performance.
• 1.8.0 PR notes — review scope, compatibility, release qualification, risks, and explicitly open roadmap work.
• 1.8.0 qualification report — numerical oracles, memory bounds, persistence rejection, benchmark comparison, and targets.
• mathlib-fp 1.7.0 release notes — interpolation, fitting, adaptive integration/ODEs, differentiation, and convex optimisation.
• 1.7.0 PR notes — review scope, compatibility, local verification, risks, and explicitly deferred work.
• 1.7.0 qualification report — modelling accuracy, diagnostics, reentrancy, examples, and target checks.
• mathlib-fp 1.6.0 release notes — typed dense QR/CPQR, SVD, eigensystems, and inspectable direct-solve diagnostics.
• 1.6.0 qualification report — algorithm-specific residual, reconstruction, target, example, and benchmark evidence.
• mathlib-fp 1.5.0 release notes — typed contiguous single/double real/complex matrices, kernels, and direct solve.
• 1.5.0 qualification report — target configurations, numerical evidence, benchmarks, dependencies, and gaps.
• mathlib-fp 1.4.0 release notes — GeometryLib vector arithmetic and scale-safe normalization.
• mathlib-fp 1.3.0 release notes — complex and vector foundation.
• mathlib-fp 1.2.3 release notes — numerical correctness, special-function accuracy, and robust probability tails.
• mathlib-fp 1.2.2 release notes — complete newcomer example coverage and cross-platform example builds.
• mathlib-fp 1.2.1 release notes — terminology and public API naming consistency.
• mathlib-fp 1.2.0 release notes — first public release.

Terminology

mathlib-fp uses the following terms consistently:

“Kit” describes an API class, not a domain or unit. Supporting units containing constants, types, low-level functions, or exception declarations do not need an artificial Kit class.

Domains

Public API naming inventory

Dependency Graph

MathBase

├── AlgebraLib

├── FinanceLib

├── StatsLib

├── EngineeringLib

├── NumericsLib

├── ProbabilityLib

├── CombinatoricsLib

├── OptimizationLib

├── TimeSeriesLib

├── MLLib

├── InterchangeLib (optional adapters over selected domains)

└── GeometryLib

Common Base Types

All domains share the types defined in MathBase.SharedTypes:

TIntegerArray  = array of Integer;

TDoubleArray   = array of Double;

TSingleArray   = array of Single;

TExtendedArray = array of Extended;

TDoublePair    = record Lower, Upper: Double; end;

TComplexArray  = array of TComplex;  // MathBase.Complex

TSingleComplexArray = array of TSingleComplex;

Design Principles

• Kit classes normally use static class methods for stateless calculations.
• TMatrixKit also implements IMatrix; it is the established matrix factory and concrete implementation as well as the algebra Kit class.
• The typed dense API is the contiguous single/double real/complex path. Its migration guide names copy and allocation costs.
• Collection APIs use TDoubleArray, TIntegerArray, or documented matrix aliases.
• Optional ADecimals parameters round scalar results where documented.
• Invalid inputs raise typed exceptions such as EFinanceError, EStatsError, EMatrixError, EProbabilityError, or the domain-specific equivalent.

Find an algorithm

The problem-oriented beginner recipes are the shortest route when you know the task but not its Pascal identifier.

See the supported platform matrix for compiler, target, and precision qualifications.