TerraHS

A purely functional Haskell library for geospatial programming and map algebra, with no foreign dependencies.

Overview

Repository: https://github.com/LambdaGeo/terrahs

TerraHS is a from-scratch rewrite of the original TerraHS (2006–2009), a Haskell binding to the C++ GIS library TerraLib developed at Brazil’s National Institute for Space Research (INPE). This version keeps the ideas — geometry, spatial predicates, and a generalized map algebra — while dropping the FFI dependency entirely: everything is plain Haskell, or built on well-established pure-Haskell libraries for file I/O.

It is the earliest link in the research trajectory that leads to DisSModel: a 2005 exploration of scientific models as verifiable, executable artifacts, now revisited as a modern, dependency-free library.

The library serves two purposes at once:

  1. Teaching material — each module illustrates a functional-programming concept applied to a real problem: algebraic data types, type classes for ad-hoc polymorphism, smart constructors for enforcing invariants, hand-written parser combinators, and a small map-algebra DSL built from first principles.
  2. A research tool — the map algebra (TerraHS.Algebra.Coverage) is a faithful reconstruction of the algebra proposed in the original TerraHS Master’s thesis (INPE, 2006), generalizing Tomlin’s (1990) classic map algebra with arbitrary spatial predicates.
Haskell No FFI Cabal

Features

  • Geometry — Point, Line, Polygon, bounding boxes, and a Geometry type class unifying area/perimeter/centroid/envelope across all three.
  • Topology — bounding-box overlap, point-on-line, point-in-polygon (ray casting), and line-crosses-polygon predicates.
  • Simplification — Ramer-Douglas-Peucker line/polygon simplification, pure arithmetic on coordinates, no extra dependency.
  • File I/O — read/write WKT and GeoJSON; read ESRI Shapefiles (.shp + .dbf, UTF-8 or Latin-1), pairing geometry with attributes.
  • Map algebra — two complementary implementations:
    • TerraHS.Algebra.Coverage, a direct reconstruction of the original thesis’s algebra: a discrete Coverage (domain → values) with local, focal, and zonal operators generalized around arbitrary spatial predicates.
    • TerraHS.Algebra.Funct / TerraHS.Algebra.Field, a classic Tomlin-style raster algebra (local, focal, zonal, global operators over a 2D grid).
  • Dynamic spatial models — TerraHS.CA (terrahs-ca), a small, generic cellular-automaton machine built on a Store comonad; demonstrated with Game of Life, diffusion, and forest-fire examples.
  • PNG rendering — TerraHS.Render.PNG (terrahs-render, via JuicyPixels, pure Haskell/no FFI) draws any Coverage at its real geometric position.

Installation

TerraHS is a standard Cabal package (cabal-version: 3.0), with no system dependencies beyond a working GHC and Cabal toolchain (GHC ≥ 8.10, Cabal ≥ 3.0 recommended). The recommended way to get that toolchain is GHCup:

curl --proto '=https' --tlsv1.2 -sSf https://get-ghcup.haskell.org | sh

Then build the library and examples with Cabal:

git clone https://github.com/LambdaGeo/terrahs
cd terrahs
cabal build

Part of the DisSModel Ecosystem

TerraHS traces the same research trajectory as DisSModel and QGISSPARQL: geospatial models built for reuse, verification, and reproducibility — here worked out in a purely functional setting.