Continuous integration

Deno's built-in tools make it easy to set up Continuous Integration (CI) pipelines for your projects. Testing, linting and formatting your code can all be done with the corresponding commands deno test, deno lint and deno fmt. In addition, you can generate code coverage reports from test results with deno coverage in pipelines.

Setting up a basic pipeline Jump to heading#

You can set up basic pipelines for Deno projects in GitHub Actions. The concepts explained on this page largely apply to other CI providers as well, such as Azure Pipelines, CircleCI or GitLab.

Building a pipeline for Deno generally starts with checking out the repository and installing Deno:

name: Build

on: push

jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      - uses: denoland/setup-deno@v2
        with:
          deno-version: v2.x # Run with latest stable Deno.

To expand the workflow, add any of the deno subcommands that you might need:

# Check if the code is formatted according to Deno's default
# formatting conventions.
- run: deno fmt --check

# Scan the code for syntax errors and style issues. If
# you want to use a custom linter configuration you can add a configuration file with --config <myconfig>
- run: deno lint

# Run all test files in the repository and collect code coverage. The example
# runs with all permissions, but it is recommended to run with the minimal permissions your program needs (for example --allow-read).
- run: deno test --allow-all --coverage=cov/

# This generates a report from the collected coverage in `deno test --coverage`. It is
# stored as a .lcov file which integrates well with services such as Codecov, Coveralls and Travis CI.
- run: deno coverage --lcov cov/ > cov.lcov

Cross-platform workflows Jump to heading#

As a Deno module maintainer, you probably want to know that your code works on all of the major operating systems in use today: Linux, MacOS and Windows. A cross-platform workflow can be achieved by running a matrix of parallel jobs, each one running the build on a different OS:

jobs:
  build:
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
        os: [ubuntu-latest, macos-latest, windows-latest]
    steps:
      - run: deno test --allow-all --coverage cov/

git config --system core.autocrlf false
git config --system core.eol lf

If you are working with experimental or unstable Deno APIs, you can include a matrix job running the canary version of Deno. This can help to spot breaking changes early on:

jobs:
  build:
    runs-on: ${{ matrix.os }}
    continue-on-error: ${{ matrix.canary }} # Continue in case the canary run does not succeed
    strategy:
      matrix:
        os: [ubuntu-latest, macos-latest, windows-latest]
        deno-version: [v1.x]
        canary: [false]
        include:
          - deno-version: canary
            os: ubuntu-latest
            canary: true

Speeding up Deno pipelines Jump to heading#

Reducing repetition Jump to heading#

In cross-platform runs, certain steps of a pipeline do not need to run for each OS necessarily. For example, generating the same test coverage report on Linux, MacOS and Windows is a bit redundant. You can use the if conditional keyword of GitHub Actions in these cases. The example below shows how to run code coverage generation and upload steps only on the ubuntu (Linux) runner:

- name: Generate coverage report
  if: matrix.os == 'ubuntu-latest'
  run: deno coverage --lcov cov > cov.lcov

- name: Upload coverage to Coveralls.io
  if: matrix.os == 'ubuntu-latest'
  # Any code coverage service can be used, Coveralls.io is used here as an example.
  uses: coverallsapp/github-action@master
  with:
    github-token: ${{ secrets.GITHUB_TOKEN }} # Generated by GitHub.
    path-to-lcov: cov.lcov

Caching dependencies Jump to heading#

As a project grows in size, more and more dependencies tend to be included. Deno will download these dependencies during testing and if a workflow is run many times a day, this can become a time-consuming process. A common solution to speed things up is to cache dependencies so that they do not need to be downloaded anew.

Deno stores dependencies locally in a cache directory. In a pipeline the cache can be preserved between workflows by setting the DENO_DIR environment variable and adding a caching step to the workflow:

# Set DENO_DIR to an absolute or relative path on the runner.
env:
  DENO_DIR: my_cache_directory

steps:
  - name: Cache Deno dependencies
    uses: actions/cache@v2
    with:
      path: ${{ env.DENO_DIR }}
      key: my_cache_key

At first, when this workflow runs the cache is still empty and commands like deno test will still have to download dependencies, but when the job succeeds the contents of DENO_DIR are saved and any subsequent runs can restore them from cache instead of re-downloading.

There is still an issue in the workflow above: at the moment the name of the cache key is hardcoded to my_cache_key, which is going to restore the same cache every time, even if one or more dependencies are updated. This can lead to older versions being used in the pipeline even though you have updated some dependencies. The solution is to generate a different key each time the cache needs to be updated, which can be achieved by using a lockfile and by using the hashFiles function provided by GitHub Actions:

key: ${{ hashFiles('deno.lock') }}

To make this work you will also need a have a lockfile in your Deno project, which is discussed in detail here. Now, if the contents of deno.lock are changed, a new cache will be made and used in subsequent pipeline runs thereafter.

To demonstrate, let's say you have a project that uses the logger from @std/log:

import * as log from "jsr:@std/[email protected]";

In order to increment this version, you can update the import statement and then reload the cache and update the lockfile locally:

deno install --reload --lock=deno.lock --frozen=false --entrypoint deps.ts

You should see changes in the lockfile's contents after running this. When this is committed and run through the pipeline, you should then see the hashFiles function saving a new cache and using it in any runs that follow.

Clearing the cache Jump to heading#

Occasionally you may run into a cache that has been corrupted or malformed, which can happen for various reasons. It is possible to clear a cache from the GitHub Actions UI, or you can simply change the name of the cache key. A practical way of doing so without having to forcefully change your lockfile is to add a variable to the cache key name, which can be stored as a GitHub secret and can be changed if a new cache is needed:

key: ${{ secrets.CACHE_VERSION }}-${{ hashFiles('deno.lock') }}