<dependency>
<groupId>io.vertx</groupId>
<artifactId>vertx-unit</artifactId>
<version>5.0.0.CR3</version>
<scope>test</scope>
</dependency>
Vertx unit
Asynchronous polyglot unit testing.
Introduction
Vertx Unit is designed for writing asynchronous unit tests with a polyglot API and running these tests in the JVM. Vertx Unit Api borrows from existing test frameworks like JUnit or QUnit and follows the Vert.x practices.
As a consequence Vertx Unit is the natural choice for testing Vert.x applications.
To use vert.x unit, add the following dependency to the dependencies section of your build descriptor:
-
Maven (in your
pom.xml
):
-
Gradle (in your
build.gradle
file):
testCompile ${io.vertx}:${vertx-unit}:5.0.0.CR3
Vert.x unit can be used in different ways and run anywhere your code runs, it is just a matter of reporting the results the right way, this example shows the bare minimum test suite:
TestSuite suite = TestSuite.create("the_test_suite");
suite.test("my_test_case", context -> {
String s = "value";
context.assertEquals("value", s);
});
suite.run();
The run
method will execute the suite and go through all the tests of the suite. The suite can fail or pass, this does not matter if the outer world is not aware of the test result.
When executed, the test suite now reports to the console the steps of the test suite:
Begin test suite the_test_suite Begin test my_test Passed my_test End test suite the_test_suite , run: 1, Failures: 0, Errors: 0
The reporters
option configures the reporters used by the suite runner for reporting the execution of the tests, see the Reporting section for more info.
Writing a test suite
A test suite is a named collection of test case, a test case is a straight callback to execute. The suite can have lifecycle callbacks to execute before and/or after the test cases or the test suite that are used for initializing or disposing services used by the test suite.
TestSuite suite = TestSuite.create("the_test_suite");
suite.test("my_test_case_1", context -> {
// Test 1
});
suite.test("my_test_case_2", context -> {
// Test 2
});
suite.test("my_test_case_3", context -> {
// Test 3
});
The API is fluent and therefore the test cases can be chained:
TestSuite suite = TestSuite.create("the_test_suite");
suite.test("my_test_case_1", context -> {
// Test 1
}).test("my_test_case_2", context -> {
// Test 2
}).test("my_test_case_3", context -> {
// Test 3
});
The test cases declaration order is not guaranteed, so test cases should not rely on the execution of another test case to run. Such practice is considered as a bad one.
Vertx Unit provides before and after callbacks for doing global setup or cleanup:
TestSuite suite = TestSuite.create("the_test_suite");
suite.before(context -> {
// Test suite setup
}).test("my_test_case_1", context -> {
// Test 1
}).test("my_test_case_2", context -> {
// Test 2
}).test("my_test_case_3", context -> {
// Test 3
}).after(context -> {
// Test suite cleanup
});
The declaration order of the method does not matter, the example declares the before callback before the test cases and after callback after the test cases but it could be anywhere, as long as it is done before running the test suite.
The before callback is executed before any tests, when it fails, the test suite execution will stop and the failure is reported. The after callback is the last callback executed by the testsuite, unless the before callback reporter a failure.
Likewise, Vertx Unit provides the beforeEach and afterEach callback that do the same but are executed for each test case:
TestSuite suite = TestSuite.create("the_test_suite");
suite.beforeEach(context -> {
// Test case setup
}).test("my_test_case_1", context -> {
// Test 1
}).test("my_test_case_2", context -> {
// Test 2
}).test("my_test_case_3", context -> {
// Test 3
}).afterEach(context -> {
// Test case cleanup
});
The beforeEach callback is executed before each test case, when it fails, the test case is not executed and the failure is reported. The afterEach callback is the executed just after the test case callback, unless the beforeEach callback reported a failure.
Asserting
Vertx Unit provides the TestContext
object for doing assertions in test cases. The context object provides the usual methods when dealing with assertions.
assertEquals
Assert two objects are equals, works for basic types or json types.
suite.test("my_test_case", context -> {
context.assertEquals(10, callbackCount);
});
There is also an overloaded version for providing a message:
suite.test("my_test_case", context -> {
context.assertEquals(10, callbackCount, "Should have been 10 instead of " + callbackCount);
});
Usually each assertion provides an overloaded version.
assertNotEquals
The counter part of assertEquals.
suite.test("my_test_case", context -> {
context.assertNotEquals(10, callbackCount);
});
assertNull
Assert an object is null, works for basic types or json types.
suite.test("my_test_case", context -> {
context.assertNull(null);
});
assertNotNull
The counter part of assertNull.
suite.test("my_test_case", context -> {
context.assertNotNull("not null!");
});
assertInRange
The assertInRange
targets real numbers.
suite.test("my_test_case", context -> {
// Assert that 0.1 is equals to 0.2 +/- 0.5
context.assertInRange(0.1, 0.2, 0.5);
});
assertTrue and assertFalse
Asserts the value of a boolean expression.
suite.test("my_test_case", context -> {
context.assertTrue(var);
context.assertFalse(value > 10);
});
Failing
Last but not least, test provides a fail method that will throw an assertion error:
suite.test("my_test_case", context -> {
context.fail("That should never happen");
// Following statements won't be executed
});
The failure can either be a string as seen previously or an error. The error object depends on the target language, for Java or Groovy it can be any class extending Throwable- , for JavaScript it is an _error, for Ruby it is an Exception.
Using third-party assertion framework
It is also possible to use any other assertion framework, like the popular hamcrest and assertj. The recommended way to go is to use verify
and perform the assertions within the supplied Handler. This way, asynchronous testing termination will be correctly handled.
suite.test("my_test_case", context -> context.verify(v -> {
// Using here Assert from junit, could be assertj, hamcrest or any other
// Even manually throwing an AssertionError.
Assert.assertNotNull("not null!");
Assert.assertEquals(10, callbackCount);
}));
Asynchronous testing
The previous examples supposed that test cases were terminated after their respective callbacks, this is the default behavior of a test case callback. Often it is desirable to terminate the test after the test case callback, for instance:
suite.test("my_test_case", context -> {
Async async = context.async();
eventBus.consumer("the-address", msg -> {
(2)
async.complete();
});
(1)
});
1 | The callback exits but the test case is not terminated |
2 | The event callback from the bus terminates the test |
Creating an Async
object with the async
method marks the executed test case as non terminated. The test case terminates when the complete
method is invoked.
When the complete callback is not invoked, the test case fails after a certain timeout. |
Several Async
objects can be created during the same test case, all of them must be completed to terminate the test.
suite.test("my_test_case", context -> {
HttpClient client = vertx.createHttpClient();
client.request(HttpMethod.GET, 8080, "localhost", "/").onComplete(context.asyncAssertSuccess(req -> {
req.send().onComplete(context.asyncAssertSuccess(resp -> {
context.assertEquals(200, resp.statusCode());
}));
}));
Async async = context.async();
vertx.eventBus().consumer("the-address", msg -> {
async.complete();
});
});
Async objects can also be used in before or after callbacks, it can be very convenient in a before callback to implement a setup that depends on one or several asynchronous results:
suite.before(context -> {
Async async = context.async();
HttpServer server = vertx.createHttpServer();
server.requestHandler(requestHandler);
server.listen(8080).onComplete(ar -> {
context.assertTrue(ar.succeeded());
async.complete();
});
});
It is possible to wait until the completion of a specific Async
, similar to Java’s count-down latch:
Async async = context.async();
HttpServer server = vertx.createHttpServer();
server.requestHandler(requestHandler);
server.listen(8080).onComplete(ar -> {
context.assertTrue(ar.succeeded());
async.complete();
});
// Wait until completion
async.awaitSuccess();
this should not be executed from the event loop! |
Async can also be created with an initial count value, it completes when the count-down reaches zero using countDown
:
Async async = context.async(2);
HttpServer server = vertx.createHttpServer();
server.requestHandler(requestHandler);
server.listen(8080).onComplete(ar -> {
context.assertTrue(ar.succeeded());
async.countDown();
});
vertx.setTimer(1000, id -> {
async.complete();
});
// Wait until completion of the timer and the http request
async.awaitSuccess();
Calling complete()
on an async completes the async as usual, it actually sets the value to 0
.
Asynchronous assertions
TestContext
provides useful methods that provides powerful constructs for async testing:
The asyncAssertSuccess
method returns a Handler<AsyncResult<T>>
instance that acts like Async
, resolving the Async
on success and failing the test on failure with the failure cause.
Async async = context.async();
vertx.deployVerticle("my.verticle").onComplete(ar -> {
if (ar.succeeded()) {
async.complete();
} else {
context.fail(ar.cause());
}
});
// Can be replaced by
vertx.deployVerticle("my.verticle").onComplete(context.asyncAssertSuccess());
The asyncAssertSuccess
method returns a Handler<AsyncResult<T>>
instance that acts like Async
, invoking the delegating Handler<T>
on success and failing the test on failure with the failure cause.
AtomicBoolean started = new AtomicBoolean();
Async async = context.async();
vertx.deployVerticle(new AbstractVerticle() {
public void start() throws Exception {
started.set(true);
}
}).onComplete(ar -> {
if (ar.succeeded()) {
context.assertTrue(started.get());
async.complete();
} else {
context.fail(ar.cause());
}
});
// Can be replaced by
vertx.deployVerticle("my.verticle").onComplete(context.asyncAssertSuccess(id -> {
context.assertTrue(started.get());
}));
The async is completed when the Handler
exits, unless new asyncs were created during the invocation, which can be handy to chain asynchronous behaviors:
Async async = context.async();
vertx.deployVerticle("my.verticle").onComplete(ar1 -> {
if (ar1.succeeded()) {
vertx.deployVerticle("my.otherverticle").onComplete(ar2 -> {
if (ar2.succeeded()) {
async.complete();
} else {
context.fail(ar2.cause());
}
});
} else {
context.fail(ar1.cause());
}
});
// Can be replaced by
vertx.deployVerticle("my.verticle").onComplete(context.asyncAssertSuccess(id ->
vertx.deployVerticle("my_otherverticle").onComplete(context.asyncAssertSuccess())
));
The asyncAssertFailure
method returns a Handler<AsyncResult<T>>
instance that acts like Async
, resolving the Async
on failure and failing the test on success.
Async async = context.async();
vertx.deployVerticle("my.verticle").onComplete(ar -> {
if (ar.succeeded()) {
context.fail();
} else {
async.complete();
}
});
// Can be replaced by
vertx.deployVerticle("my.verticle").onComplete(context.asyncAssertFailure());
The asyncAssertFailure
method returns a Handler<AsyncResult<T>>
instance that acts like Async
, invoking the delegating Handler<Throwable>
on failure and failing the test on success.
Async async = context.async();
vertx.deployVerticle("my.verticle").onComplete(ar -> {
if (ar.succeeded()) {
context.fail();
} else {
context.assertTrue(ar.cause() instanceof IllegalArgumentException);
async.complete();
}
});
// Can be replaced by
vertx.deployVerticle("my.verticle").onComplete(context.asyncAssertFailure(cause -> {
context.assertTrue(cause instanceof IllegalArgumentException);
}));
The async is completed when the Handler
exits, unless new asyncs were created during the invocation.
Repeating test
When a test fails randomly or not often, for instance a race condition, it is convenient to run the same test multiple times to increase the failure likelihood of the test.
TestSuite.create("my_suite").test("my_test", 1000, context -> {
// This will be executed 1000 times
});
When declared, beforeEach and afterEach callbacks will be executed as many times as the test is executed.
test repetition are executed sequentially |
Sharing objects
The TestContext
has get
/put
/remove
operations for sharing state between callbacks.
Any object added during the before callback is available in any other callbacks. Each test case will operate on a copy of the shared state, so updates will only be visible for a test case.
TestSuite.create("my_suite").before(context -> {
// host is available for all test cases
context.put("host", "localhost");
}).beforeEach(context -> {
// Generate a random port for each test
int port = helper.randomPort();
// Get host
String host = context.get("host");
// Setup server
Async async = context.async();
HttpServer server = vertx.createHttpServer();
server.requestHandler(req -> {
req.response().setStatusCode(200).end();
});
server.listen(port, host).onComplete(ar -> {
context.assertTrue(ar.succeeded());
context.put("port", port);
async.complete();
});
}).test("my_test", context -> {
// Get the shared state
int port = context.get("port");
String host = context.get("host");
// Do request
HttpClient client = vertx.createHttpClient();
client.request(HttpMethod.GET, port, host, "/resource").onComplete(context.asyncAssertSuccess(req -> {
req.send().onComplete(context.asyncAssertSuccess(resp -> {
context.assertEquals(200, resp.statusCode());
}));
}));
});
sharing any object is only supported in Java, other languages can share only basic or json types. Other objects should be shared using the features of that language. |
Running
When a test suite is created, it won’t be executed until the run
method is called.
suite.run();
The test suite can also be run with a specified Vertx
instance:
suite.run(vertx);
When running with a Vertx
instance, the test suite is executed using the Vertx event loop, see the Event loop section for more details.
Several test suites can be executed in the same verticle, Vert.x Unit waits until completion of all suite executed. |
Test suite completion
No assumptions can be made about when the test suite will be completed, and if some code needs to be executed after the test suite, it should either be in the test suite after callback or as callback of the Completion
:
TestCompletion completion = suite.run(vertx);
// Simple completion callback
completion.handler(ar -> {
if (ar.succeeded()) {
System.out.println("Test suite passed!");
} else {
System.out.println("Test suite failed:");
ar.cause().printStackTrace();
}
});
The Completion
object provides also a resolve
method that takes a Promise
object, this Promise
will be notified of the test suite execution:
TestCompletion completion = suite.run();
// When the suite completes, the promise is resolved
completion.resolve(startPromise);
This allow to easily create a test verticle whose deployment is the test suite execution, allowing the code that deploys it to be easily aware of the success or failure.
The completion object can also be used like a latch to block until the test suite completes. This should be used when the thread running the test suite is not the same than the current thread:
Completion completion = suite.run();
// Wait until the test suite completes
completion.await();
The await
throws an exception when the thread is interrupted or a timeout is fired.
The awaitSuccess
is a variation that throws an exception when the test suite fails.
Completion completion = suite.run();
// Wait until the test suite succeeds otherwise throw an exception
completion.awaitSuccess();
Time out
Each test case of a test suite must execute before a certain timeout is reached. The default timeout is of 2 minutes, it can be changed using test options:
TestOptions options = new TestOptions().setTimeout(10000);
// Run with a 10 seconds time out
suite.run(options);
Event loop
Vertx Unit execution is a list of tasks to execute, the execution of each task is driven by the completion of the previous task. These tasks should leverage Vert.x event loop when possible but that depends on the current execution context (i.e the test suite is executed in a main
or embedded in a Verticle
) and wether or not a Vertx
instance is configured.
The setUseEventLoop
configures the usage of the event loop:
useEventLoop:null | useEventLoop:true | useEventLoop:false | |
---|---|---|---|
| use vertx event loop | use vertx event loop | force no event loop |
in a | use current event loop | use current event loop | force no event loop |
in a main | use no event loop | raise an error | use no event loop |
The default useEventLoop
value is null
, that means that it will uses an event loop when possible and fallback to no event loop when no one is available.
Reporting
Reporting is an important piece of a test suite, Vertx Unit can be configured to run with different kind of reporters.
By default no reporter is configured, when running a test suite, test options can be provided to configure one or several:
ReportOptions consoleReport = new ReportOptions().
setTo("console");
// Report junit files to the current directory
ReportOptions junitReport = new ReportOptions().
setTo("file:.").
setFormat("junit");
suite.run(new TestOptions().
addReporter(consoleReport).
addReporter(junitReport)
);
File reporting
Reports to a file, a Vertx
instance must be provided:
- to
-
file
:
dir name - format
-
simple or junit
- example
-
file:.
The file reporter will create files in the configured directory, the files will be named after the test suite name executed and the format (i.e simple creates txt files and junit creates xml files).
Log reporting
Reports to a logger, a Vertx
instance must be provided:
- to
-
log
:
logger name - example
-
log:mylogger
Event bus reporting
Reports events to the event bus, a Vertx
instance must be provided:
- to
-
bus
:
event bus address - example
-
bus:the-address
It allow to decouple the execution of the test suite from the reporting.
The messages sent over the event bus can be collected by the EventBusCollector
and achieve custom reporting:
EventBusCollector collector = EventBusCollector.create(
vertx,
new ReportingOptions().addReporter(
new ReportOptions().setTo("file:report.xml").setFormat("junit")));
collector.register("the-address");
Vertx integration
By default, assertions and failures must be done on the TestContext
and throwing an assertion error works only when called by Vert.x Unit:
suite.test("my_test_case", ctx -> {
// The failure will be reported by Vert.x Unit
throw new RuntimeException("it failed!");
});
In a regular Vert.x callback, the failure will be ignored:
suite.test("test-server", testContext -> {
HttpServer server = vertx.createHttpServer().requestHandler(req -> {
if (req.path().equals("/somepath")) {
throw new AssertionError("Wrong path!");
}
req.response().end();
});
});
Since Vert.x 3.3, a global exception handler can be set to report the event loop uncaught exceptions:
suite.before(testContext -> {
// Report uncaught exceptions as Vert.x Unit failures
vertx.exceptionHandler(testContext.exceptionHandler());
});
suite.test("test-server", testContext -> {
HttpServer server = vertx.createHttpServer().requestHandler(req -> {
if (req.path().equals("/somepath")) {
throw new AssertionError("Wrong path!");
}
req.response().end();
});
});
The exception handler is set during the before phase, the TestContext
is shared between each before, test and after phase. So the exception handler obtained during the before phase is correct.
Junit integration
Although Vertx Unit is polyglot and not based on JUnit, it is possible to run a Vertx Unit test suite or a test case from JUnit, allowing you to integrate your tests with JUnit and your build system or IDE.
@RunWith(VertxUnitRunner.class)
public class JUnitTestSuite {
@Test
public void testSomething(TestContext context) {
context.assertFalse(false);
}
}
The VertxUnitRunner
uses the junit annotations for introspecting the class and create a test suite after the class. The methods should declare a TestContext
argument, if they don’t it is fine too. However the TestContext
is the only way to retrieve the associated Vertx instance of perform asynchronous tests.
The JUnit integration is also available for the Groovy language with the io.vertx.groovy.ext.unit.junit.VertxUnitRunner
runner.
Running a test on a Vert.x context
By default the thread invoking the test methods is the JUnit thread. The RunTestOnContext
JUnit rule can be used to alter this behavior for running these test methods with a Vert.x event loop thread.
Thus there must be some care when state is shared between test methods and Vert.x handlers as they won’t be on the same thread, e.g incrementing a counter in a Vert.x handler and asserting the counter in the test method. One way to solve this is to use proper synchronization, another is to execute test methods on a Vert.x context that will be propagated to the created handlers.
For this purpose the RunTestOnContext
rule needs a Vertx
instance. Such instance can be provided, otherwise the rule will manage an instance under the hood. Such instance can be retrieved when the test is running, making this rule a way to manage a Vertx
instance as well.
@RunWith(VertxUnitRunner.class)
public class RunOnContextJUnitTestSuite {
@Rule
public RunTestOnContext rule = new RunTestOnContext();
@Test
public void testSomething(TestContext context) {
// Use the underlying vertx instance
Vertx vertx = rule.vertx();
}
}
The rule can be annotated by @Rule
or @ClassRule
, the former manages a Vert.x instance per test, the later a single Vert.x for the test methods of the class.
keep in mind that you cannot block the event loop when using this rule. Usage of classes like CountDownLatch or similar classes must be done with care. |
Timeout
The Vert.x Unit 2 minutes timeout can be overriden with the timeout
member of the @Test
annotation:
public class JunitTestWithTimeout {
@Test(timeout = 1000l)
public void testSomething(TestContext context) {
//...
}
}
For a more global configuration, the Timeout
rule can be used:
@RunWith(VertxUnitRunner.class)
public class TimeoutTestSuite {
@Rule
public Timeout rule = Timeout.seconds(1);
@Test
public void testSomething(TestContext context) {
//...
}
}
Use io.vertx.ext.unit.junit.Timeout , not org.junit.rules.Timeout . |
The method level timeout overwrites the class level timeout:
@RunWith(VertxUnitRunner.class)
public class TimeoutOverwriteTestSuite {
@Rule
public Timeout rule = Timeout.millis(5_000L);
@Test
public void testQuick(TestContext context) {
//...
}
@Test(timeout = 30_000L)
public void testSlow(TestContext context) {
//...
}
}
The @Test timeout overrides the io.vertx.ext.unit.junit.Timeout rule but not the org.junit.rules.Timeout rule. |
Parameterized tests
JUnit provides useful Parameterized
tests, Vert.x Unit tests can be ran with this particular runner thanks to the VertxUnitRunnerWithParametersFactory
:
@RunWith(Parameterized.class)
@Parameterized.UseParametersRunnerFactory(VertxUnitRunnerWithParametersFactory.class)
public class SimpleParameterizedTest {
@Parameterized.Parameters
public static Iterable<Integer> data() {
return Arrays.asList(0, 1, 2);
}
public SimpleParameterizedTest(int value) {
//...
}
@Test
public void testSomething(TestContext context) {
// Execute test with the current value
}
}
Parameterized tests can also be done in Groovy with the io.vertx.groovy.ext.unit.junit.VertxUnitRunnerWithParametersFactory
.
Repeating a test
When a test fails randomly or not often, for instance a race condition, it is convenient to run the same test multiple times to increase the likelihood failure of the test.
With JUnit a test has to be annotated with @Repeat
to be repeated. The test must also define the RepeatRule
among its rules.
@RunWith(VertxUnitRunner.class)
public class RepeatingTest {
@Rule
public RepeatRule rule = new RepeatRule();
@Repeat(1000)
@Test
public void testSomething(TestContext context) {
// This will be executed 1000 times
}
}
When declared, before and after life cycle will be executed as many times as the test is executed.
test repetition are executed sequentially |
Using with other assertion libraries
Vert.x Unit usability has been greatly improved in Vert.x 3.3. You can now write tests using Hamcrest, AssertJ, Rest Assured, or any assertion library you want. This is made possible by the global exception handler described in Vertx integration.
You can find Java examples of using Vert.x Unit with Hamcrest and AssertJ in the vertx-examples project.
Java language integration
Test suite integration
The Java language provides classes and it is possible to create test suites directly from Java classes with the following mapping rules:
The testSuiteObject
argument methods are inspected and the public, non static methods with TestContext
parameter are retained and mapped to a Vertx Unit test suite via the method name:
-
before
: before callback -
after
: after callback -
beforeEach
: beforeEach callback -
afterEach
: afterEach callback -
when the name starts with test : test case callback named after the method name
public class MyTestSuite {
public void testSomething(TestContext context) {
context.assertFalse(false);
}
}
This class can be turned into a Vertx test suite easily:
TestSuite suite = TestSuite.create(new MyTestSuite());