Liking cljdoc? Tell your friends :D

Testing

Polylith encourages a test-centric approach when working with code. New tests are easy to write, and you can avoid mocking in most cases because you can access all components from your workspace.

You write tests for your bricks (bases and components) and deployable projects. The poly tool always runs brick tests from the context of the deployable projects that reference them. This convention verifies that the bricks tests pass using the classpath formulated from the project and all its referenced bricks.

The development project offers a way to run tests from your REPL and a temporary home (and test vehicle) for new bricks that aren’t yet part of any deployable project

Let’s continue with our tutorial, we last left off exploring tagging.

The info command provides a great way to see what tests poly will run. Rerun poly info:

info annotated
See the Flags doc for a general description of flags.

Notice that poly has marked the user component with * (asterisk) because as part of our tutorial, you changed its core namespace.

Notice the stx flags under the cl column for both the user and cli bricks. The x flags mean poly test will run tests for these bricks from the command-line project.

There is no * after the cli brick, so why is it marked to be tested with the x flag? Well, even though the cli brick hasn’t changed, it depends on the user brick, which has changed, so poly needs to retest it.

Notice the absence of the x flag in the st- flags under the dev column for both the user and cli bricks. Bricks are not tested from the development by default.

Before you run the test command, edit the interface-test namespace in the user component. Replace the dummy-test with a real test:

components/user/test/se/example/user/interface_test.clj
(ns se.example.user.interface-test
  (:require [clojure.test :refer :all]
            [se.example.user.interface :as user]))

(deftest hello--when-called-with-a-name--then-return-hello-phrase
  (is (= "Hello Lisa!"
         (user/hello "Lisa"))))

To explore how poly handles a failing test, change the core namespace in the user component:

components/user/src/se/example/user/core.clj
(ns se.example.user.core)

(defn hello [name]
  (str "Hello " name "!!")) (1)
1Add an extra !

Cursive users: You can run the test from your IDE:

  • Make sure the namespace is loaded, e.g. via the menu (or keyboard shortcuts) Tools > REPL > Load File in REPL.

  • Or, place the cursor under the test and run: Tools > REPL > Run Test under carret in REPL

failing test

Cursive users:

You need to instruct Cursive to run tests from the root module (example in this case) by performing the following steps:

  • Select Edit Configuration…​ in the menu with configurations at the top of the IDE.

  • Select Edit configuration templates…​ at the bottom left corner.

  • Select clojure.test in the left menu.

  • Set Module under Common Options to example and press OK.

You should now be able to run tests from the IDE.

In some of the examples that follow, the poly exit code is relevant. In those cases, at the end of the output, we’ll show the exit code like so:

# Exit code: 0

This line is not part of the actual poly output and is only a convention to convey the process exit code to you. As is the standard convention, an exit code of 0 conveys success.

Run the test command:

poly test

You should see output like this:

Projects to run tests from: command-line

Running tests for the command-line project using test runner: Polylith built-in clojure.test runner... (1)
Running tests from the command-line project, including 2 bricks: user, cli (2)

Testing se.example.cli.core-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Testing se.example.user.interface-test

FAIL in (hello--when-called-with-a-name--then-return-hello-phrase) (interface_test.clj:6) (3)
expected: (= "Hello Lisa!" (user/hello "Lisa"))
  actual: (not (= "Hello Lisa!" "Hello Lisa!!"))

Ran 1 tests containing 1 assertions.
1 failures, 0 errors. (3)

Test results: 0 passes, 1 failures, 0 errors. (3)

# Exit code: 1 (3)
1The poly tool runs tests in the context of each project.
2The referenced bricks are tested within the context of the project.
3Notice evidence of the failing test

Adapt your test to match the new behavior:

components/user/test/se/example/user/interface_test.clj
(ns se.example.user.interface-test
  (:require [clojure.test :refer :all]
            [se.example.user.interface :as user]))

(deftest hello--when-called-with-a-name--then-return-hello-phrase
  (is (= "Hello Lisa!!" (1)
         (user/hello "Lisa"))))
1Edit to expect the extra ! so the test will pass

Cursive users: If you rerun the test from the REPL, it will now turn green:

successful test

Rerun test with poly:

poly test

You should see output like this:

Projects to run tests from: command-line

Running tests for the command-line project using test runner: Polylith built-in clojure.test runner...
Running tests from the command-line project, including 2 bricks: user, cli

Testing se.example.cli.core-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Testing se.example.user.interface-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors. (1)

Test results: 1 passes, 0 failures, 0 errors. (1)

Execution time: 1 seconds

# Exit code: 0 (1)
1Notice evidence of the now-passing test

Brick tests are not evaluated for execution from the development project by default. Specify the :dev argument to include the development project.

Try it out with the info command to see the impact of what will be tested:

poly info :dev
info dev annotated

Notice under the dev column, poly marked both the user and the cli bricks for test execution with the x flag.

You don’t need to bring in dev for test execution when your projects reference all your bricks. But sometimes, you’ll create a brick before a project. In this case, you can include the brick for testing from dev.
When a brick is marked for testing from multiple projects, poly will run its tests in the context of each of those projects.

Test Filters

Filter on Projects

You can narrow the number of projects to test by specifying, e.g., project:dev or project:cl:dev.

You can use full project names or aliases:

  • project:development is equivalent to project:dev

  • project:command-line:development is the same as project:cl:dev

If you only specify project:dev, then poly only includes the development project:

poly info project:dev
info project dev annotated

Notice:

  • the absence of x in the st- flags under the cl column

  • the presence of x in the stx flags under the dev column.

You can consider poly info :dev as a shorthand for selecting all projects. The equivalent, when specifying the project argument, requires specifying all projects:

poly info project:cl:dev
info project cl dev annotated
We’ll show later that the project argument also applies to project tests.

Filter on Bricks

You can also filter which bricks to include for test execution. If you’ve been following the tutorial, your workspace looks like this:

poly info
info filter on bricks annotated

The x flags under the cl column mean poly will test both bricks from the command-line project.

If you filter on the cli brick:

poly info brick:cli
info brick cli annotated

Notice that poly has marked only the cli brick for testing.

Let’s pretend that no bricks are marked by poly for testing:

info no changes annotated

Rerunning poly info brick:cli again gives the exact same result:

info brick cli no changes annotated

The poly tool applies the brick:cli filter argument after it has evaluated cli for test execution.

If you want to force the cli brick tests to run, you need to pass in :all-bricks (or :all, if you also want to execute the project tests):

poly info brick:cli :all-bricks
info brick cli no changes all bricks annotated

Notice the x in stx flags; you have forced poly to mark the cli brick for testing.

You can specify multiple bricks, e.g., brick:cli:user.

You can exclude all bricks with the brick:- argument, which can be useful when combined with :project or :all to execute only the project tests.

Project Tests

Before we proceed, let’s add a test to the command-line project.

Add a test directory to the command-line project:

example
├── projects
│   └── command-line
│       └── test

Then add the test path to projects/command-line/deps.edn:

 :aliases {:test {:extra-paths ["test"] (1)
                  :extra-deps  {}}
1Add test path

Now add this same path to your ./deps.edn:

            :test {:extra-paths ["components/user/test"
                                 "bases/cli/test"
                                 "projects/command-line/test"]} (1)
1Add projects/command-line/test path

Finally, add a project.command-line.dummy-test namespace to the command-line project:

example
├── projects
│   └── command-line
│       └── test
│           └── project
│               └──command_line
│                  └──dummy_test.clj
(ns project.command-line.dummy_test (1)
  (:require [clojure.test :refer :all]))

(deftest dummy-test
  (is (= 1 1)))
1If you’ve been following our tutorial, you might notice we did not begin with our top namespace se.example. We could have chosen se.example.project.command-line, but note that this would conflict if we also had project brick. To avoid conflicts with bricks and keep things short and simple, we’ve opted for project.command-line here. Also, because poly executes each project in isolation, the choice of namespace is less critical.
Normally, when you write tests in Clojure, you match the test namespace to the namespace it is testing. This strategy gives your tests access to private vars in the tested namespace. The poly tool guarantees encapsulation, which makes the usage of private vars unnecessary, allowing for more flexibility in test namespace choices. See Interface for more details.

Rerun poly info:

info project dir annotated

Notice poly has marked the command-line project as changed with a *:

  • status flags of -t- to tell us that the project now has a test directory.

  • dev flags of -t- mean the project is referenced by the development project

But why no x flag? Well, poly doesn’t execute project tests to by default. You must specify :project (or :all) to also include projects:

poly info :project
info project annotated

Notice the x in the -tx flags under the status column; this means poly has marked the command-line project for testing. Let’s verify by running the tests:

poly test :project

You should see output like this:

Projects to run tests from: command-line

Running tests for the command-line project using test runner: Polylith built-in clojure.test runner...
Running tests from the command-line project, including 2 bricks and 1 project: user, cli, command-line (1)

Testing se.example.cli.core-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Testing se.example.user.interface-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Testing project.command-line.dummy_test (2)

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Execution time: 2 seconds

# Exit code: 0
1Notice and 1 project
2Our command-line project tests are included!

They passed!

Test approaches

As you have just seen, you can add tests at two levels: the brick and the project.

We recommend project tests for:

  • Slower tests. Tests that take over 100 milliseconds (or whatever threshold you choose) are good candidates.

  • Tailor-made tests that are unique per project.

Brick tests are for faster tests. Fast-running brick tests keep your feedback loop short during development. Remember, poly test only runs brick tests, not project tests.

But does that mean we recommend only putting unit tests in your bricks? No. As long as the tests are fast (e.g., by using in-memory databases), you should put them in the bricks they belong to.

Before we continue, let’s commit the work we have done so far and mark our example workspace as stable:

git add --all
git commit -m "Added tests"
git tag -f stable-lisa

Rerun poly info, you should see output like:

info added tests

The * signs are gone, and no x flags means poly has marked nothing for testing.

The poly tool only executes tests for a brick if it has directly or indirectly changed. A way to force it to test all bricks is to pass in :all-bricks:

poly info :all-bricks
info all bricks annotated

Notice that poly has marked all the bricks for testing under deployable project cl.

To also run brick tests from the development project, specify :dev:

poly info :all-bricks :dev
info all bricks dev annotated
This is for demonstration purposes only. The poly tool has already marked all of your bricks for testing under the command-line (alias cl) project. Retesting your bricks the development (alias dev) project is questionable.

To include all brick and project tests (except development) you can type:

poly info :all
info all annotated

To also include development, type:

poly info :all :dev
info all dev annotated
Because projects and bricks were already marked for testing, adding :dev in this case is questionable. You’ll typically use the development project to test new bricks you’ve not yet added to any deployable project.

Now let’s see if it actually all works:

poly test :all :dev
Projects to run tests from: command-line, development

Running tests for the command-line project using test runner: Polylith built-in clojure.test runner... (1)
Running tests from the command-line project, including 2 bricks and 1 project: user, cli, command-line

Testing se.example.cli.core-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Testing se.example.user.interface-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Testing project.command-line.dummy_test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.
Running tests for the development project using test runner: Polylith built-in clojure.test runner... (2)
Running tests from the development project, including 2 bricks and 1 project: user, cli, command-line

Testing se.example.cli.core-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Testing se.example.user.interface-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Execution time: 1 seconds

# Exit code: 0
1Tests run from command-line project
2And rerun from development project (notice absence of command-line project tests when run from development)

Looks like it worked!

Test Setup and Teardown

Sometimes, tests require some setup before being run and some teardown (or cleanup) after being run.

If multiple projects needed the same test setup/teardown, you’d put this support in a component to make it shareable. We only have one project, so we’ll put the test setup/teardown in the command-line project.

Let’s create a test-setup namespace in the command-line project’s test directory and add setup and teardown functions:

example
├── projects
│   └── command-line
│       └── test
│           └── project
│               └──command_line
│                  └──test_setup.clj
(ns project.command-line.test-setup
  (:require [clojure.test :refer :all]))

(defn setup [project-name]
  (println (str "--- test setup for " project-name " ---")))

(defn teardown [project-name]
  (println (str "--- test teardown for " project-name " ---")))

You need to keep two things in mind:

  • Make sure your functions are accessible (in this case, from the command-line project)

  • Make sure the functions take exactly one argument, the project name

Specify your new functions in ./workspace.edn for the command-line project:

 ...
 :projects {"development" {:alias "dev"}
            "command-line" {:alias "cl"
                            :test {:setup-fn project.command-line.test-setup/setup
                                   :teardown-fn project.command-line.test-setup/teardown}}}}
In practice, if you don’t need a :teardown-fn, you can omit it.

Rerun your tests:

poly test :all
Projects to run tests from: command-line

Running test setup for the command-line project: project.command-line.test-setup/test-setup
--- test setup for command-line --- (1)

Running tests for the command-line project using test runner: Polylith built-in clojure.test runner...
Running tests from the command-line project, including 2 bricks and 1 project: user, cli, command-line

Testing se.example.cli.core-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Testing se.example.user.interface-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.

Testing project.command-line.test-setup

Ran 0 tests containing 0 assertions.
0 failures, 0 errors.

Test results: 0 passes, 0 failures, 0 errors.

Testing project.command-line.dummy_test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Test results: 1 passes, 0 failures, 0 errors.
Running test teardown for the command-line project: project.command-line.test-setup/test-teardown
--- test teardown for command-line --- (2)


Execution time: 2 seconds

# Exit code: 0
1The setup
2The teardown

Nice, it worked!

Include and Exclude Bricks by Configuration

There is a way to restrict what tests to run for a project by giving a list of bricks to include and/or exclude in workspace.edn, e.g.:

{...
 :projects {"mytool" {:alias "t"
                      :test {:include []}}
            "myservice" {:alias "s"
                         :test {:exclude ["cli" "user"]}}
            ...

This configuration tells poly to include no brick tests for project mytool and exclude cli and user brick tests for project myservice.

The :include keyword is optional and assumed.

You may wonder when this could be useful. A good example is the polylith codebase itself, where the workspace.edn looks similar to this:

...
  :projects {"poly" {:alias "poly"} (1)
             "polyx" {:alias "polyx" :test []} (2)
             "development" {:alias "dev"} (3)
             ...
1all brick tests for deployable projects are included by default
2all brick tests excluded via :test []
3no tests are included for the development project by default

Our motivation is to speed up the test execution time. We felt comfortable with this strategy because:

  • poly brick tests give full brick coverage

  • polyx uses these bricks in the same way poly does

When you exclude a brick via configuration, poly will never include it for testing, even when you specify :project or :all arguments.

Summary

Let’s summarise the different ways to run the tests.

By default, poly test runs tests for each deployable project’s bricks.

CommandBrick tests? (once for each deployable project)Deployable project tests?Selection

poly test

yes

no

only brick tests impacted by change

poly test :project

yes

yes

only brick and project tests impacted by change

poly test :all‑bricks

yes

no

forces all brick tests

poly test :all

yes

yes

forces all brick and project tests

By specifying the ':dev' argument, you can also tell poly to include brick tests from the development project.

CommandBrick tests? (once for each project including the development project)Project tests? (including the development project 1)Selection

poly test :dev

yes

no

only brick tests impacted by change

poly test :project :dev

yes

yes

only brick and project tests impacted by change

poly test :all‑bricks :dev

yes

no

forces all brick tests

poly test :all :dev

yes

yes

forces all brick and project tests

Table notes:

  1. The development project does not typically include any tests

You can explicitly select projects via e.g, project:proj1 or project:proj1:proj2.

You can filter bricks to run the tests for with e.g., brick:b1 or brick:b1:b2.

Remember that the info command is an excellent way to get an overview of what tests poly will run.

How Test Execution Works

From the REPL

The primary purpose of the development project is to allow you to work with all of your code from your IDE using a single REPL. To meet that goal, you must set up your project in a way that is compatible with tool.deps and your IDE integration. One example of this compatibility setup is adding test paths explicitly in ./deps.edn to give access to the tests from your REPL.

The ./deps.edn config file sets up all your paths and dependencies. The :dev and :test aliases (and sometimes profile aliases) informs tools.deps what source code and libraries should be accessible from your IDE and REPL. When you’ve set this up correctly, you can run your tests from your REPL, which will have access to all the test and src code.

Libraries you reference as default dependencies are automatically accessible when you run tests. You should reference libraries you only need for testing under the test alias.

From poly

When you run the test command, poly will detect which components, bases and projects have been affected since the last stable point in time. Based on this information, poly will:

  • for each affected project:

    • run tests for the affected bricks (components and bases) referenced by the project

    • run tests belonging to the project (if you’ve specified :project or :all)

The poly tool executes this set of tests in an isolated classloader, which speeds up the test execution and reflects the production classpath. The test command includes libraries (and their transitive dependencies) from both default dependencies and :test aliases. You can also run tests from the development project, but that’s not its primary purpose.

Tests fail fast. If you run tests on projects A, B, C, and D, when a test in project B fails, the whole test run stops at project B. The poly tool won’t run tests for projects C and D. Failing fast also applies to test setup and teardown, should they fail for whatever reason.

The poly test command uses the sum of all library dependencies for components and bases, either indirectly via :local/root or directly via :deps and :extra-deps. If a library is defined more than once in the set of bricks and projects, then the latest version of that library is used if not overridden by :override-deps in the project.

A project does not need to respecify libraries specified by its referenced bricks. It will typically specify dependencies common to all bricks, e.g., org.clojure/clojure.

Sometimes, you’ll depend on libraries not hosted in the default Maven repositories. You can specify custom maven repositories in a brick. Everything that depends on the brick will pick up the custom Maven repositories.

For example, the poly tool’s datomic-ions brick specifies a custom Maven repository for datomic libraries. You can verify that the brick picks up the maven repository by executing poly ws get:components:datomic-ions:maven-repos:

{"datomic-cloud" {:url "s3://datomic-releases-1fc2183a/maven/releases"}}

And that the invoicing project uses it by executing poly ws get:projects:invoicing:maven-repos:

{"central" {:url "https://repo1.maven.org/maven2/"},
 "clojars" {:url "https://repo.clojars.org/"},
 "datomic-cloud" {:url "s3://datomic-releases-1fc2183a/maven/releases"}}

Every project using the datomic-ions brick will inherit the datomic-cloud maven repository.

Troubleshooting

If your tests don’t work for whatever reason, you can pass in :verbose to see the configuration and paths poly uses when executing the tests:

poly test :verbose
# config:
{:mvn/repos {"central" {:url "https://repo1.maven.org/maven2/"}, ...
# paths: (1)
["bases/cli/resources" "bases/cli/src" "components/user-remote/resources" ...
1Represents the JVM classpath

Long-Lived Shell Sessions

For long-running shell sessions, after running the test command many times, you may eventually get classloader errors.

Solutions:

  • Quit, then restart the poly shell

  • Run tests outside of the shell, e.g, poly test

  • Switch to an external test runner.

Can you improve this documentation?Edit on GitHub

cljdoc is a website building & hosting documentation for Clojure/Script libraries

× close