tour3-pipeline_5bb915692c7d3a04dd5b5d5f.md

After creating your first project, you've initialized it with a working demo pipeline. Now we'll see how we can define our own pipelines on Semaphore.

Let's say that our CI build consists of 4 steps:

1. Compile
2. Run unit tests
3. Run integration tests
4. Run performance tests

Semaphore can run CI jobs in parallel, so we can use that to run integration and performance tests simultaneously and save us time. We'll keep the unit tests separate since we want them to finish fast, and halt the build in case of failure.

Recall what we've covered in concepts. Sequential phases maps to blocks. Parallel stuff maps to jobs within a block's task. Commands within a job run sequentially.

Let's start with two sequential blocks:

# .semaphore/semaphore.yml
blocks:
  - name: "Set things up"
    task:
      jobs:
        - name: Compile
          commands:
            - checkout
            - echo "Linting code"
            - echo "Compiling code"
  - name: "Unit tests"
    task:
      jobs:
        - name: Unit
          commands:
            - checkout
            - echo "Running unit tests"

With this configuration, commands in the "Compile" job run sequentially. If all of them pass, the "Set things up" block passes, and the pipeline proceeds to run the "Unit tests" block.

Notice how each block includes the checkout command. This is Semaphore's built-in command which loads your code in the job's environment.

Defining blocks with parallel jobs

Let's add another block with two parallel jobs:

# .semaphore/semaphore.yml
blocks:
  - name: "Set things up"
    task:
      jobs:
        - name: Compile
          commands:
            - checkout
            - echo "Linting code"
            - echo "Compiling code"
  - name: "Unit tests"
    task:
      jobs:
        - name: Unit
          commands:
            - checkout
            - echo "Running unit tests"
  - name: "E2E tests"
    task:
      prologue:
        commands:
          - checkout
      jobs:
        # these two jobs run in parallel, if "Unit tests" block passes:
        # (also, comments are ok ;)
        - name: "Integration tests"
          commands:
            - echo "Running integration tests"
        - name: "Performance tests"
          commands:
            - echo "Running performance tests"

In the new "E2E tests" block, commands in prologue run before each job, so that we don't repeat ourselves. Prologue is the right place for setup tasks like checking out or building the code. Blocks also support a list of commands that run after each job, called epilogue. This is covered in depth in the Pipeline YAML reference.

Next steps

Running a CI build in practice usually requires us to install and run dependencies, set environment variables, etc. Semaphore has several features which make this very easy. Let's move on to customizing the build environment.