highs を Homebrew, apt, Nix, pacman, MacPorts, scoop でインストール

プロジェクトリンク

• ホームページ
• リポジトリ
• 上流ドキュメント
• パッケージマネージャページ

コマンドとエイリアス

• highs

プロジェクトの歴史

HiGHS grew from high-performance research solvers for linear optimization, especially the dual revised simplex work by Qi Huangfu and Julian Hall. The official site describes the codebase as C++11 software with no required third-party utilities for source builds.

The README describes a solver suite rather than a single algorithm: primal and dual revised simplex solvers, an LP interior-point solver, a QP active-set solver, and a MIP branch-and-cut solver. The documentation adds PDLP first-order LP support and explains the executable/library split.

The public tag line shows 1.x releases starting in 2021, then steady expansion through Python packaging, NuGet packaging, interface documentation, MIP work, GPU/PDLP-related development, and HiPO-related builds.

採用の歴史

HiGHS gained a major scientific-Python adoption point when SciPy 1.6.0 added HiGHS methods to `scipy.optimize.linprog` for large sparse problems. SciPy 1.9.0 then made `method='highs'` the default for `linprog` and added mixed-integer linear programming support.

The JuMP ecosystem documents `HiGHS.jl` as a wrapper around the HiGHS solver with both a thin C API wrapper and a MathOptInterface implementation. That gives Julia modelers access to the same solver family through JuMP models.

Packaging now spans both system package managers and language package channels. The README badges and text point to PyPI `highspy`, NuGet `Highs.Native`, release binaries, and source builds, while the input package map shows Homebrew, Debian, MacPorts, Nix, Arch, and Scoop packaging.

使われ方

From the command line, HiGHS reads MPS and CPLEX LP files and solves them with options such as presolve, solver choice, parallel mode, thread count, time limit, and output solution/basis files. A minimal run is `highs model.mps`.

As a library, users can build, modify, solve, and inspect optimization models through the native C++ API or through language bindings. Python users often meet HiGHS through SciPy's `linprog` and `milp` APIs or through the `highspy` wrapper; Julia users commonly meet it through JuMP and HiGHS.jl.

パッケージ好きにとっての重要性

HiGHS matters to package nerds because it is a serious permissively licensed optimization solver with no required third-party dependencies for the core build. That makes it unusually friendly to distributions compared with solver stacks that depend on proprietary binaries or complex external libraries.

It also sits at an important boundary between command-line packages and language ecosystems: the same solver is shipped as a Unix executable, a C/C++ library, a Python package, a Julia solver backend, and a NuGet package.

タイムライン

• 2018: The dual revised simplex work by Huangfu and Hall appears as the key citation used by HiGHS documentation.
• 2021: v1.1.1 appears in the public tag line.
• 2021: SciPy 1.6.0 documents HiGHS methods for `linprog`.
• 2022: SciPy 1.9.0 makes `method='highs'` the default for `linprog` and adds MILP support.
• 2024: HiGHS workshops begin appearing on the official site as community events.
• 2025: v1.10.0 appears in the public tag line during continued solver/interface development.
• 2026: v1.15.0 appears in the public tag line.

Related projects

• SciPy is a major downstream consumer through `scipy.optimize.linprog` and `scipy.optimize.milp`.
• JuMP and MathOptInterface use HiGHS through HiGHS.jl.
• COIN-OR Clp, GLPK, commercial solvers, and other LP/MIP solvers are common benchmark and package-set neighbors.

ソース

• Documentation: https://ergo-code.github.io/HiGHS/dev/
• JuMP HiGHS.jl docs: https://jump.dev/JuMP.jl/stable/packages/HiGHS/
• Project site: https://highs.dev/
• README: https://github.com/ERGO-Code/HiGHS#readme
• Repository tags: https://github.com/ERGO-Code/HiGHS/tags
• SciPy 1.6.0 notes: https://docs.scipy.org/doc/scipy/release/1.6.0-notes.html
• SciPy 1.9.0 notes: https://docs.scipy.org/doc/scipy/release/1.9.0-notes.html