Travis CI¶
In this section we are going to use Travis CI as our continuous-integration service. Travis CI is mostly used for building and running tests for projects hosted at GitHub. It supports different programming languages and for our particular case, it supports the Crystal language.
Note
If you are new to continuous integration (or you want to refresh the basic concepts) we may start reading the core concepts guide.
Now let's see some examples!
Build and run specs¶
Using latest
and nightly
¶
A first (and very basic) Travis CI config file could be:
.travis.yml
language: crystal
That's it! With this config file, Travis CI by default will run crystal spec
.
Now, we just need to go to Travis CI dashboard to add the GitHub repository.
Let's see another example:
.travis.yml
language: crystal
crystal:
- latest
- nightly
script:
- crystal spec
- crystal tool format --check
With this configuration, Travis CI will run the tests using both Crystal latest
and nightly
releases on every push to a branch on your Github repository.
Note
When creating a Crystal project using crystal init
, Crystal creates a .travis.yml
file for us.
Using a specific Crystal release¶
Let's suppose we want to pin a specific Crystal release (maybe we want to make sure the shard compiles and works with that version) for example Crystal 0.31.1.
Travis CI only provides runners to latest
and nightly
releases directly and so, we need to install the requested Crystal release manually. For this we are going to use Docker.
First we need to add Docker as a service in .travis.yml
, and then we can use docker
commands in our build steps, like this:
.travis.yml
language: minimal
services:
- docker
script:
- docker run -v $PWD:/src -w /src crystallang/crystal:0.31.1 crystal spec
Note
A list with the different official Crystal docker images is available at DockerHub.
Using latest
, nightly
and a specific Crystal release all together!¶
Supported runners can be combined with Docker-based runners using a Build Matrix. This will allow us to run tests against latest
and nightly
and pinned releases.
Here is the example:
.travis.yml
matrix:
include:
- language: crystal
crystal:
- latest
script:
- crystal spec
- language: crystal
crystal:
- nightly
script:
- crystal spec
- language: bash
services:
- docker
script:
- docker run -v $PWD:/src -w /src crystallang/crystal:0.31.1 crystal spec
Installing shards packages¶
In native runners (language: crystal
), Travis CI already automatically installs shards dependencies using shards install
. To improve build performance we may add caching on top of that.
Using Docker¶
In a Docker-based runner we need to run shards install
explicitly, like this:
.travis.yml
language: bash
services:
- docker
script:
- docker run -v $PWD:/src -w /src crystallang/crystal:0.31.1 shards install
- docker run -v $PWD:/src -w /src crystallang/crystal:0.31.1 crystal spec
Note
Since the shards will be installed in ./lib/
folder, it will be preserved for the second docker run command.
Installing binary dependencies¶
Our application or maybe some shards may required libraries and packages. This binary dependencies may be installed using different methods. Here we are going to show an example using the Apt command (since the Docker image we are using is based on Ubuntu)
Here is a first example installing the libsqlite3
development package using the APT addon:
.travis.yml
language: crystal
crystal:
- latest
before_install:
- sudo apt-get -y install libsqlite3-dev
addons:
apt:
update: true
script:
- crystal spec
Using Docker¶
We are going to build a new docker image based on crystallang/crystal, and in this new image we will be installing the binary dependencies.
To accomplish this we are going to use a Dockerfile:
Dockerfile
FROM crystallang/crystal:latest
# install binary dependencies:
RUN apt-get update && apt-get install -y libsqlite3-dev
And here is the Travis CI configuration file:
.travis.yml
language: bash
services:
- docker
before_install:
# build image using Dockerfile:
- docker build -t testing .
script:
# run specs in the container
- docker run -v $PWD:/src -w /src testing crystal spec
Note
Dockerfile arguments can be used to use the same Dockerfile for latest, nightly or a specific version.
Using services¶
Travis CI may start services as requested.
For example, we can start a MySQL database service by adding a services:
section to our .travis.yml
:
.travis.yml
language: crystal
crystal:
- latest
services:
- mysql
script:
- crystal spec
Here is the new test file for testing against the database:
spec/simple_db_spec.cr
require "./spec_helper"
require "mysql"
it "connects to the database" do
DB.connect ENV["DATABASE_URL"] do |cnn|
cnn.query_one("SELECT 'foo'", as: String).should eq "foo"
end
end
When pushing this changes Travis CI will report the following error: Unknown database 'test' (Exception)
, showing that we need to configure the MySQL service and also setup the database:
.travis.yml
language: crystal
crystal:
- latest
env:
global:
- DATABASE_NAME=test
- DATABASE_URL=mysql://root@localhost/$DATABASE_NAME
services:
- mysql
before_install:
- mysql -e "CREATE DATABASE IF NOT EXISTS $DATABASE_NAME;"
- mysql -u root --password="" $DATABASE_NAME < db/schema.sql
script:
- crystal spec
We are using a schema.sql
script to create a more readable .travis.yml
. The file ./db/schema.sql
looks like this:
schema.sql
CREATE TABLE ... etc ...
Pushing these changes will trigger Travis CI and the build should be successful!
Caching¶
If we read Travis CI job log, we will find that every time the job runs, Travis CI needs to fetch the libraries needed to run the application:
Fetching https://github.com/crystal-lang/crystal-mysql.git
Fetching https://github.com/crystal-lang/crystal-db.git
This takes time and, on the other hand, these libraries might not change as often as our application, so it looks like we may cache them and save time.
Travis CI uses caching to improve some parts of the building path. Here is the new configuration file with cache enabled:
.travis.yml
language: crystal
crystal:
- latest
cache: shards
script:
- crystal spec
Let's push these changes. Travis CI will run, and it will install dependencies, but then it will cache the shards cache folder which, usually, is ~/.cache/shards
. The following runs will use the cached dependencies.