Cabal projects

cabal init

A real project starts with a directory and one command. Cabal asks a few questions and writes the skeleton:

mkdir todo && cd todo && cabal init

Package name? [default: todo] Package version? [default: 0.1.0.0] What does the package build: 1) executable ... Guessing dependencies...

Answer 1 for an executable, accept the defaults, and choose base as the only dependency for now.

The .cabal file

The generated todo.cabal is the project description. Four stanzas matter at this stage:

cabal-version: 3.0 name: todo version: 0.1.0.0 executable todo main-is: Main.hs hs-source-dirs: app build-depends: base >=4.18 && <5 default-language: Haskell2010

build-depends is the dependency list, and the version bounds are a promise about which versions your code was tested against. Keep them loose but present.

Build and run

Four commands cover the whole edit cycle. Cabal caches what it built, so the second run is fast:

cabal build

cabal run todo

cabal repl

ghci> :t main main :: IO ()

cabal repl loads the project modules with the right paths and dependencies, which is why it beats a bare ghci once the project has more than one file.

Project layout

The conventional layout separates the program from the library so that tests can import the logic without running main:

todo/ ├── app/Main.hs ├── src/Todo.hs ├── test/Spec.hs ├── todo.cabal └── cabal.project

The executable depends on the library and does nothing but parse arguments and call it. That single decision is what makes the capstone testable in lesson 49.

Next: Dependencies

Article author: Arthur Isaev