Get Faster Feedback From Your Cypress Tests Running On CircleCI

How to run the modified Cypress.io specs first when using CircleCI workflows

Nobody likes slow tests. Pushing a pull request commit and waiting for 30 minutes for the CI to work through all the specs, just to find out your changed spec has failed ... is bad. In this blog post, I will show how to run only the new and changed Cypress.io specs first. This blog post shows it using CircleCI workflows. A similar approach could be used with any CI provider. In fact, in my previous post Get Faster Feedback From Your Cypress Tests Running On GitHub Actions I have described how to achieve the same using GitHub Actions.

🎁 You can find the source code for this blog post in the repo bahmutov/chat.io

The initial workflow

Typically, we check out the source code, install NPM dependencies, start the application and run all tests using Cypress CircleCI Orb. For out application, we also need MongoDB and Redis services, and we can start them up using Docker images.

.circleci/config.yml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
version: 2.1
orbs:
# https://github.com/cypress-io/circleci-orb
cypress: cypress-io/[email protected]

executors:
with-services:
# to run MongoDB and Redis on CircleCI, follow the instructions
# https://glebbahmutov.com/blog/testing-mongo-with-cypress/
docker:
# image used to install source code,
# run our server and run Cypress tests
- image: cypress/base:14.16.0
environment:
# for connecting to the Mongo and Redis services
MONGODB: mongodb://root:[email protected]:27017/
SESSION_SECRET: SoMeSeCrEtStrInG

# image used to run Mongo in a separate container
- image: mongo:4.4.5
environment:
MONGO_INITDB_ROOT_USERNAME: root
MONGO_INITDB_ROOT_PASSWORD: rootPass1234
# image for running Redis
- image: redis:alpine

workflows:
build:
jobs:
- cypress/run:
name: Run all tests
executor: with-services
start: npm start
wait-on: 'http://localhost:3000'
record: true
group: 'All Circle Tests'
no-workspace: true

Parallelization

Great start. What if we want to run tests faster? We can split them across N containers, each executing some specs through Cypress parallelization. Read the blog post Parallel or not for more examples, but in general we will have one job to install dependencies and pass them to the test jobs via workspace.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
version: 2.1
orbs:
# https://github.com/cypress-io/circleci-orb
cypress: cypress-io/[email protected]

executors:
with-services:
# to run MongoDB and Redis on CircleCI, follow the instructions
# https://glebbahmutov.com/blog/testing-mongo-with-cypress/
docker:
# image used to install source code,
# run our server and run Cypress tests
- image: cypress/base:14.16.0
environment:
# for connecting to the Mongo and Redis services
MONGODB: mongodb://root:[email protected]:27017/
SESSION_SECRET: SoMeSeCrEtStrInG

# image used to run Mongo in a separate container
- image: mongo:4.4.5
environment:
MONGO_INITDB_ROOT_USERNAME: root
MONGO_INITDB_ROOT_PASSWORD: rootPass1234
# image for running Redis
- image: redis:alpine

workflows:
build:
jobs:
# every testing job needs to install dependencies and Cypress
- cypress/install:
name: Install
# to install dependencies, we do not need other services
# so let's use the executor close to what the tests will use
# https://github.com/cypress-io/circleci-orb/blob/master/docs/executors.md
executor: cypress/base-14

- cypress/run:
name: Run all tests
executor: with-services
requires:
- Install
start: npm start
wait-on: 'http://localhost:3000'
record: true
# split all specs across 4 machines
# https://on.cypress.io/parallelization
parallel: true
parallelism: 4
group: 'All Circle Tests'
no-workspace: true

Tip: you can run tests even faster by using CircleCI RAM disk feature, read the blog post Start CircleCI Machines Faster by Using RAM Disk.

Run changed specs first

But what if we want to open a pull request with changed / added Cypress specs and run them first? That way we find out if our changes are correct very quickly. We still will run all the tests to ensure nothing else is broken, of course.

Here is how we can do this: we will define a middle job between the install and "Run all tests" jobs. This middle job will only run the changed Cypress specs on pull requests. If the job is running for regular commits, the job will halt itself, skipping right to the "Run all tests" job. Here is the definition for the job:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
# run only the changed Cypress tests if any
# and only if we are running in a pull request
- cypress/run:
name: Run changed Cypress tests
executor: with-services
requires:
- Install
post-checkout:
- run:
name: Stop if not a pull request
command: |
if [ -z "$CIRCLE_PULL_REQUEST" ]; then
echo "Not a pull request, exiting..."
circleci-agent step halt
fi
- run:
name: Stop if no changed specs
# also stop if there are too many changed specs
command: |
git diff --name-only origin/main
# make sure to NOT fail if there are no changed specs
n=$(git diff --name-only origin/main | { grep cypress/integration || true; } | wc -l | tr -d ' ')
echo ""
echo "number of added or changed Cypress specs ${n}"
echo ""
if [ ${n} -lt 1 ]; then
echo "No Cypress specs changed, exiting..."
circleci-agent step halt
fi

if [ ${n} -gt 4 ]; then
echo "Too many Cypress specs changed, will run them all in the next job..."
circleci-agent step halt
fi

# to "trick" the Orb into NOT installing again
# (the dependencies were installed in the Install job)
# we use the dummy "install-command"
install-command: echo Already installed
start: npm start
wait-on: 'http://localhost:3000'
no-workspace: true
command: |
git diff --name-only origin/main
# by now we know there are changed Cypress specs
# so we know grep command will not fail
specs=$(git diff --name-only origin/main | grep cypress/integration | tr '\n' ',')
echo ""
echo "Changed and added Cypress specs"
echo ${specs}
echo ""

# we have to form the Cypress run command ourselves
npx cypress run --record --group "1. Changed specs" --spec ${specs}

Let me explain the individual parts of the above job definition.

Install

1
2
3
4
name: Run changed Cypress tests
executor: with-services
requires:
- Install

The modified specs job uses the installed dependencies from the workspace created by the "Install" job, just like "Run all specs" job.

Only run on pull requests

Each cypress/run can execute steps after the checkout step (even if we are using the workspace from the "Install" job and do not check out the code again). In our case, we want to skip the job if this is not a pull request:

1
2
3
4
5
6
7
8
post-checkout:
- run:
name: Stop if not a pull request
command: |
if [ -z "$CIRCLE_PULL_REQUEST" ]; then
echo "Not a pull request, exiting..."
circleci-agent step halt
fi

CircleCI has many built-in environment variables, you can see them in each job by expanding the step.

Built-in CircleCI environment variables

Count changes Cypress specs

Next we want to find all changed Cypress specs (modified or added). We can use the git diff --name-only origin/main command to list all changed files in the current branch vs the default branch "main". The output is a single filename per line, something like

1
2
3
4
# let's say we only have two modified files
$ git diff --name-only origin/main
cypress/integration/a.js
README.md

Next, we can pipe the output through Unix utility commands to count the number of lines.

1
2
3
4
5
# grep filters non-Cypress filenames
# we need to use "|| true;" to ignore exit code 1 from grep if there are no specs
# wc -l counts each spec (one per line)
# tr -d ' ' removes all white space characters that "wc" outputs
n=$(git diff --name-only origin/main | { grep cypress/integration || true; } | wc -l | tr -d ' ')

The variable n will have a number greater or equal to zero - that's the number of changed specs in the current pull request. We can skip the remaining job commands if the number is zero. We also can skip running the changed specs if a lot of specs have changed - it would be more efficient to run all specs using parallelization in this case.

1
2
3
4
5
6
7
8
9
10
11
12
echo ""
echo "number of added or changed Cypress specs ${n}"
echo ""
if [ ${n} -lt 1 ]; then
echo "No Cypress specs changed, exiting..."
circleci-agent step halt
fi

if [ ${n} -gt 4 ]; then
echo "Too many Cypress specs changed, will run them all in the next job..."
circleci-agent step halt
fi

Run changed specs

Now that we know we have a small number of changed Cypress specs, let's run them.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
# to "trick" the Orb into NOT installing again
# (the dependencies were installed in the Install job)
# we use the dummy "install-command"
install-command: echo Already installed
start: npm start
wait-on: 'http://localhost:3000'
no-workspace: true
command: |
git diff --name-only origin/main
# by now we know there are changed Cypress specs
# so we know grep command will not fail
specs=$(git diff --name-only origin/main | grep cypress/integration | tr '\n' ',')
echo ""
echo "Changed and added Cypress specs"
echo ${specs}
echo ""

# we have to form the Cypress run command ourselves
npx cypress run --record --group "1. Changed specs" --spec ${specs}

We skip the NPM install step using install-command: ... parameter. All dependencies have been installed already. We do need to start the application and wait for it to respond before running Cypress. We get the specs filenames and join them into a single string using Unix commands:

1
2
# tr '\n' ',' converts multiple lines into a single comma-separated string
specs=$(git diff --name-only origin/main | grep cypress/integration | tr '\n' ',')

We already know there are Cypress specs, thus no need to ignore the grep exit code using || true; syntax. The output before running the tests will be something like:

1
2
3
Changed and added Cypress specs
cypress/integration/data-across-specs/creates-room.js,
number of added or changed Cypress specs 1

Finally, we need to form the cypress run command ourselves, since we need to pass the changed specs using the --spec command line argument.

1
npx cypress run --record --group "1. Changed specs" --spec ${specs}

Tip: I give indices to the group names like --group "1. Changed specs" and group: '2. All Circle Tests' because we are recording the test results on Cypress Dashboard. By prefixing the group names with numbers, the Dashboard run lists them in the expected order. You can see the recordings yourself at this public project

Changed specs before all specs

Results

You can find the complete CircleCI config file here and see the CircleCI runs here. When there are no changed Cypress specs, the middle job finishes quickly

Pull request without changed specs

But when there is at least one changed Cypress spec, it gets to run first, providing very fast feedback to the team

Tested changed specs first

See also

Update 1: support, plugin and utils files

A pull request can modify other non-spec files inside the cypress folder, which will be included because they might match the grep filter. It is fine - Cypress still filters the --spec ... list using its testFiles and ignoreTestFiles configuration patterns.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
# changed files
cypress/integration/cannot-find-room.js
cypress/integration/data-across-specs/utils.js

number of added or changed Cypress specs 2
# Cypress command output

┌────────────────────────────────────────────────────────────────────────────────────────────────┐
│ Cypress: 8.5.0 │
│ Browser: Electron 91 (headless) │
│ Specs: 1 found (cannot-find-room.js) │
│ Searched: cypress/integration/cannot-find-room.js, cypress/integration/data-across-specs/uti │
│ ls.js, │
│ Params: Tag: false, Group: 1. Changed specs, Parallel: false │
│ Run URL: https://dashboard.cypress.io/projects/f1j79r/runs/151 │
└────────────────────────────────────────────────────────────────────────────────────────────────┘

Notice that Cypress found one spec only, ignoring the utils.js file because it is ignored by the cypress.json file setting "ignoreTestFiles": ["utils.js", "README.md"].

Update 2: CircleCI config syntax

The CircleCI config file syntax can be intimidating. Luckily, you can check if the file is correct by running the CircleCI CLI utility locally to verify the file syntax.

1
2
$ circleci config validate .circleci/config.yml
Config file at .circleci/config.yml is valid.