/haskell

Exercism exercises in Haskell.

Primary LanguageHaskellMIT LicenseMIT

Exercism Haskell Track

Build Status

Exercism exercises in Haskell

Contributing Guide

Asking for help

When in doubt just ask!

In any case, don't be afraid or shy! We promise to be friendly and as helpful as possible.

How to contribute

As a first step we recommend you read the contributing documentation.

Reporting or fixing bugs

Typical examples for a bug: A typo, a missing test case, an unclear or ambiguous problem description.

Reviewing issues and pull requests

If you have a dedicated opinion you are welcome to write a comment for an issue or a pull request. Please be detailed and include reasons, links or arguments to support your opinion.

Porting exercises

The site contains a list of missing exercises. You may implement any of these exercises for the Haskell track. Feel free to refer to the documentation on porting an exercise. Of course you can also add a totally new exercise, but it might be a good idea to first discuss it in one of our forums.

Updating an exercise test suite

Updating a test suite of an existing exercise is special because it usually affects all languages. You can refer to the documentation on updating a test suite and the test suite principles. Note that the whole test suite must run with the sample solution within a couple of seconds.

Repository structure and conventions

The track anatomy documentation is a general description of all the files and directories that are not explicitly described below.

Directory structure

├── .gitignore
├── .travis.yml
├── LICENSE
├── README.md
├── bin
│ └── fetch‐configlet
├── config.json
├── docs
│ ├── ABOUT.md
  ├── EXERCISE_README_INSERT.md
│ ├── INSTALLATION.md
│ ├── LEARNING.md
│ ├── RESOURCES.md
│ └── TESTS.md
└── exercises
  └── accumulate
  │ ├── package.yaml
  │ ├── stack.yaml
  │ ├── examples
  │ │ └── success-standard
  │ │   ├── package.yaml
  │ │   └── src
  │ │     └── Accumuĺate.hs
  │ ├── src
  │ │ └── Accumuĺate.hs
  │ ├── test
  │ │ └── Tests.hs
  │ └── .meta
  │   └── hints.md
  └── allergies
  │ ├── ...
  └── ...
  • config.json: Every exercise has to be registered here. It has a unique name and a difficulty. The sequence order is also the default order in which the exercises are fetched.

Exercise structure

Each exercise has the following structure:

  • stack.yaml has just one line specifying the current Stack snapshot. We use the same resolver for all the exercises.
  • package.yaml is a file in the hpack format that has all dependencies and build instructions for an exercise. One of the properties tracked in package.yaml is the version of the exercise.
  • src/ModuleName.hs is a stub solution.
  • examples/success-<name>/package.yaml contains library dependencies for the example solution. <name> is a unique name for the example - usually "standard" (as in success-standard), but it can be some other name in case of multiple example solutions.
  • examples/success-<name>/src/ModuleName.hs is the source code of the sample solution.
  • test/Tests.hs is the test suite.
  • .meta/hints.md is an optional file containing instructions and/or hints. It is used together with the respective description.md for the exercise from problem-specifications to build the README.md file.

Exercise versioning

Each exercise contains a four-part version in its package.yaml file, MAJOR.MINOR.PATCH.SERIAL.

There are two possibilities for the meaning of the MAJOR.MINOR.PATCH components:

  • Exercises based on a canonical-data.json in problem-specifications should use its version plus a serial number.
  • Exercises that are not based on canonical-data.json should use version 0.1.0 plus a serial number.

The serial number starts at 1 and always increases when the tests are changed, regardless of the changes in other version numbers.

When changing a test suite, the version number should be updated appropriately so that:

  • It is possible for maintainers of this track to tell whether test suites are up to date with https://github.com/exercism/problem-specifications.
  • It is easier for students to determine at-a-glance whether they have the same tests, by comparing version numbers.

This versioning policy was proposed and accepted in exercism#522.

Writing an issue

To report a bug you should create an issue on the exercism/haskell repo.

Writing a pull request

To fix a bug you should create a pull request from a fork on the exercism/haskell repo. If you need help with Git, we have some documentation on Git basics.

Development Dependencies

You should have Stack installed in your system to make contributing to this repository easier.

Stub solution

The stub solution should be as general as possible in order to not exclude any possible solutions. It should take Haskell specifics into account (for example use Maybe instead of a dummy return value). It should not contain any comments (people might forget to remove them), you can use the hints file instead.

The stub solution must compile by itself (with stack build). Ideally, it would also compile together with the test suite (with stack test --no-run-tests). These two conditions are enforced by Travis. If the second condition cannot be met for a good reason, place the explanation in .meta/DONT-TEST-STUB to circumvent the check. The first condition is always enforced and cannot be circumvented.

Example solution

The example solution could be inspiration for other language implementors. It doesn't need to be perfect or very elegant. But it should be efficient enough for the test suite to finish in only a few seconds.

Examples are named <type>-<name>. There are three possible types of examples:

  • success: The example is expected to pass the tests.
    • There must be at least success example per exercise, in order to confirm that it is possible to solve the tests.
    • There may be more than one success example for a given exercise, but these are intended for use when we want to confirm that multiple type signatures for a given solution will compile and pass the tests.
    • We do not intend to use multiple success examples just to showcase a wide variety of possible solutions, since that is not in the goals of this repository.
  • fail: The example is expected to build, but fail the tests.
    • These are intended for use when we want to make sure that the track tests have coverage: Whether the tests find certain classes of incorrect or inefficient solutions.
    • It's suggested that these only be used for tests that are specific to the track. This is under the assumption that tests sourced from problem-specifications have already been judged to have appropriate coverage by the reviewers of the problem-specifications repository.
  • error: The example is expected to fail to build.
    • There is only one intended use of this so far, and that is a single check that a solution without a type signature will fail to build (because CI builds with --pedantic).
    • We do not intend for any additional uses of this type of example.

These example types were proposed and accepted in exercism#397.

Test suite

The test suite should be derived from the respective problem-specifications/exercises/<exercise-name>/canonical-data.json and comply to some formatting and coding standards (to get an idea you may look at some of the existing tests).

Running Tests

In order to be accepted by Travis-CI, every exercise must be registered in config.json, it must compile without warnings and the example solution must pass the tests without failures. Additionally the tests should not run longer than a few seconds.

First you need to provide an example solution.

We provide three scripts in the bin directory of this repository to run the tests. These are the same scripts as those used by Travis CI.

  • test-example path/to/example/dir runs the tests on a single example.
  • test-all-examples path/to/exercise/dir runs the tests on all examples for an exercise.
  • test-stub path/to/exercise/dir checks that the stub for the given exercise compiles.

Running HLint

All code in this repository should be as idiomatic as possible, so we enforce in Travis-CI that it returns No hints when processed by HLint.

It is highly recommended to run hlint on your sources before opening a pull request, so you can fix your code before submitting it for review.

If you are certain that a suggestion given by hlint would make the code worse, you can suppress it with annotations in the source file.