We can further speed up the execution of our tests by caching the dependencies between builds. In most of our code changes the external dependencies remain stable. Yet, in the javascript job above we keep downloading all dependencies for each build. It is easy to cache files between builds using the restore_cache and save_cache steps. We only need a stable cache key: a file which changes if and only if our dependencies change. Luckily, the yarn.lock file does exactly this.
jobs:
javascript:
docker:
- image: node:8.6.0
steps:
- restore_cache:
key: node-modules-{{ checksum "yarn.lock" }}
- run:
name: Install dependencies
command: yarn install
- save_cache:
key: node-modules-{{ checksum "yarn.lock" }}
paths:
- node_modules
- run:
name: Run javascript tests
command: yarn test
One can apply the same trick to cache dependencies of other programming languages. Look for a suitable cache key. For example, if you use vendoring in your Golang application you can use the vendor.json file. For a Ruby application you can use the Gemfile.lock file.
Exercise 2.1: Improve the performance of your test by caching the dependencies using the restore_cache and save_cache steps.
If you run the job for the first time. The cache key will not be present and the yarn install command will run as usual. The following save_cache step will cache all files present in the node_modules folder. Now, if you run the job again for the second time the cache key will be present and the yarn install command will finish much faster.
Both the restore_cache / save_cache and persist_to_workspace / attach_workspace steps save and restore files for later use. Typically, the cache is used to store files between multiple workflow runs, while the workspace is used to store files between multiple jobs of a single workflow:
