Skip to content

exercism/sml

Repository files navigation

Exercism Standard ML Track

configlet sml / ci

Exercism exercises in Standard ML.

Setup

Even though there are multiple Standard ML implementations, we'll stick to PolyML.

Please read INSTALLATION.md for more info.

Contributing Guide

Any type of contribution is more than welcome!

Before opening a pull request please have look into Contributors Pull Request Guide.

Contributing a new exercise

Usually an exercise is derived from one of the exercises in the problem-specifications repository. If you want to contribute a completely new exercise, consider opening a pull request there to make it available to all tracks.

There is a comprehensive guide on how to add an exercise to one of the exercism tracks. It is advisable that you skim over the text. You don't have to remember everything, we recall the essentials here anyway. We also provide the track-specific details here. Down below we describe tooling which helps you with the boilerplate.

Basically, adding an exercise means to do the following things.

  • Register the exercise in {{ repo-path }}/config.json,
  • Use tooling to generate exercise-folder exercises/practice/{{ slug }}/,
  • Provide a solution for the exercise in example.sml,
  • Check if the linter (configlet) is satisfied.

Register exercise in {{ repo-path }}/config.json

This step needs to be done manually. You have to add a block looking like that under exercises/practice:

{
  "exercises": {
    "practice": [
      {
        "slug": "flatten-array", // the slug from `problem-specifications`
        "name": "Flatten Array",
        "uuid": "fb0a030d-33bc-4066-a30a-1b8b02cc42f1", // use `configlet uuid` to generate this
        "practices": [],
        "prerequisites": [],
        "difficulty": 1,
        "topics": []
      },
      // more exercises ...
    ],
    // more stuff ...
  },
  // more stuff ...
}

To generate a unique uuid just execute the following on the command line:

$ bin/fetch-configlet # to fetch the latest version of configlet
$ bin/configlet uuid  # paste the output into the `uuid` field.

Generate exercise-folder exercises/practice/{{ slug }}/

A folder similar to this one has to be created:

exercises/practice/flatten-array/
├── flatten-array.sml          # stub-file displayed to studend on website
├── testlib.sml                # copy of the test-library (track-specific)
├── test.sml                   # actual test-suite
├── .docs
│   ├── instructions.append.md # optional track-specific file augmenting `instructions.md`
│   └── instructions.md        # contains (track-independent) description of the exercise
└── .meta
    ├── config.json            # you may add yourself as author here
    ├── example.sml            # solution proving that the test-suite can actually be satisfied
    └── tests.toml             # specifies which tests from `problem-specifications` are implemented

The easiest way to autogenerate the boilerplate is to execute configlet sync in combination with the track specific tool bin/generate. The former is responsible for required files with "generic" content, and the latter for required files with sml-specific content

$ bin/configlet sync -yu --tests include --docs --filepaths --metadata -e {{ slug }}
$ bin/generate {{ slug }}

Note on bin/generate: You need Python 3.5+. It may fail with some exercises. Reasons: canonical-data.json does not exist, or type mismatch (in these situation you can use --force option). In those cases you will have to create the files manually.

IMPORTANT: Currently the test-framework expects example.sml to be inside .meta/, which is not the case right after execution of bin/generate. As a workaround you should move the file manually and update the path in .meta/config.json accordingly.

Provide a solution in example.sml

The most creative part is to write a valid solution example.sml which passes the test-suite. To verify that your solution is valid just execute the tests like so

$ make test-{{ slug }}

Alternatively you can verify your solution by

$ # You might need `sudo` since docker is invoked:
$ bin/test {{ slug }}

Under the hood this uses docker to run the tests in the sml-test-runner image. This is essentially what happens if a student submits their own solution.

Linting

Finally check if the linter is satisfied with your work

$ bin/configlet lint

Exercise Tests

You can execute tests by

$ make test            # all tests
$ make test-{{ slug }} # single test

Mainstream languages usually have one or more popular test-frameworks. Standard ML is not blessed with this convenience. Therefore the track implements its own "test-framework" lib/testlib.sml. It gets the job done. For technical reasons each exercise must provide its own copy of the testlib.

Any updates to lib/testlib.sml have to be synced to all exercises by

$ make redeploy-testlib

We don't want to deal with multiple versions of testlib. Hence the redeploy script is the only way to update the testlib of any exercise.