Your friendly BDD testing library for Haxe!
- Install the lib:
haxelib install buddy
- Create a test file called Main.hx:
using buddy.Should;
class Main extends buddy.SingleSuite {
public function new() {
// A test suite:
describe("Using Buddy", {
var experience = "?";
var mood = "?";
beforeEach({
experience = "great";
});
it("should be a great testing experience", {
experience.should.be("great");
});
it("should make the tester really happy", {
mood.should.be("happy");
});
afterEach({
mood = "happy";
});
});
}
}
- Make a quick test:
haxe -x Main -lib buddy
..
Using Buddy
should be a great testing experience (Passed)
should make the tester really happy (Passed)
2 specs, 0 failures, 0 pending
But please don't stop there. Try using it on other targets than Neko, Buddy supports them all on both Windows and Linux, even Hashlink if you're using Haxe 4. The only thing you need to remember is to add -D nodejs
to your hxml if you're targeting Node.js, but that is usually taken care of by libraries like hxnodejs.
Buddy was built from the ground up to have great support for async testing, so it's fully compatible with Node.js and handles ajax requests with ease. To use it, just create the specification with a function that takes one argument (targeting javascript now):
using buddy.Should;
class Main extends buddy.SingleSuite {
public function new() {
describe("Using Buddy asynchronously", {
var mood = "?";
// Add function(done) here to enable async testing:
beforeAll(function(done) {
haxe.Timer.delay(function() {
mood = "thrilled";
done(); // Call the done() function when the async operation is complete.
}, 100);
});
// Can be added to "it" and "after" as well if needed.
it("can be described in a certain word", {
mood.should.be("thrilled");
});
});
}
}
The default timeout is 5000 ms, after which the spec will automatically fail if done()
hasn't been called. If you want to change the timeout, set the property timeoutMs
in the BuddySuite
before the actual it()
specification, or in the before/after block. Here's an example:
using buddy.Should;
class Main extends buddy.SingleSuite {
public function new() {
describe("Using Buddy asynchronously", {
timeoutMs = 100;
it("should fail specs after a timeout set before it()", function(done) {
// This test will fail after 100 ms.
haxe.Timer.delay(done, 200);
});
});
}
}
You can set timeoutMs
to 0 to disable the timeout check. Note: When using function(done)
, on some targets the timeout check will run in a separate thread. Also, timeouts and asynchronous behavior aren't supported when targeting PHP or interp.
To setup tests, you can use beforeAll
, beforeEach
, afterEach
and afterAll
:
using buddy.Should;
class BeforeAfterTest extends buddy.SingleSuite {
public function new() {
describe("Using before/after", {
// vars will be set one time only, before any tests run.
var test = 0;
// Will run once as the first thing in the current describe block
beforeAll({
test++;
});
// Will run before each "it" in the current and before each "it" in any nested describes.
beforeEach({
test++;
});
it("should be a convenient way to set up tests", {
test.should.be(2);
});
describe("When nesting describes", {
beforeEach({
test++;
});
it("should run all before/afterEach defined here or above", {
test.should.be(3);
});
afterEach({
test--;
});
});
it("should run in correct order too", {
test.should.be(2);
});
// Will run after each "it" in the current and before each "it" in any nested describes.
afterEach({
test--;
});
// Will run once as the last thing in the current describe block
afterAll({
test--;
});
});
}
}
As you've seen in the examples, testing if specifications are correct is as simple as adding using Buddy.should
to the package and then use the should
extension for the identifier you want to test. The following assertions are supported:
a.should.be(b)
- Tests equality for value types (Bool
, Float
, Int
, Int64
, and the immutable String
) and identity for the other (reference) types.
a.should.beType(b)
- Tests if a
is of type b
. Basically a wrapper around Std.is
.
a.should.beLessThan(b)
a.should.beLessThanOrEqualTo(b)
a.should.beGreaterThan(b)
a.should.beGreaterThanOrEqualTo(b)
Same as Int plus
a.should.beCloseTo(b, p = 2)
- a
should be close to b
with p
decimals precision, so you can easily compare floats without worrying about precision issues.
a.should.contain(substr)
- Test if a
contains a given substring.
a.should.match(regexp)
- Test if a
matches a regular expression (EReg
).
a.should.startWith(substr)
- Test if a
starts with a given substring.
a.should.endWith(substr)
- Test if a
ends with a given substring.
a.should.beOn(date)
- Test if a
is on a given date
a.should.beOnStr(string)
- Test if a
is on a date specified by a string in the Date.fromString accepted formats.
a.should.beBefore(date)
- Test if a
is before a given date.
a.should.beBeforeStr(string)
- Same as above, but specified by a string.
a.should.beAfter(date)
- Test if a
is after a given date.
a.should.beAfterStr(string)
- Same as above, but specified by a string.
a.should.contain(b)
- Test if an Iterable contains b
.
a.should.containAll(b)
- Test if an Iterable contains all objects in Iterable b
.
a.should.containExactly(b)
- Test if an Iterable contains exactly the same objects as in Iterable b
and in the same order.
a.should.equal(b)
- Makes a deep equality check, using Type.enumEq
. A warning will be given when enums are compared by should.be
, since the result of that comparison is undefined.
Testing if a function throws an exception is made easy using the special bind
field which exists for every function.
If the function signature is String -> Void
then apply the string argument like this:
a.bind("test").should.throwValue("error")
a.bind("test").should.throwType(String)
a.bind("test").should.throwAnything()
You can also test an anonymous function directly:
(function() { throw "error"; }).should.throwType(String)
The throw methods will return the exception object, so it can be tested further. This works synchonously only.
Every assertion can be negated using not
which is present on all should
fields:
a.should.not.contain("test")
A test can be failed using the fail(o : Dynamic) : Void
method available in a BuddySuite
. The test will fail with the string value of o
as a message. If you're testing asynchronously you can pass the fail
method to the error handler. Here are some examples:
it("should fail manually when using fail()", {
fail("Totally on purpose.");
});
it("should fail if a promise fails", function(done) {
request.getJson("/some/url")
.then(function(r) {
r.statusCode.should.be(200);
done();
})
.catchError(fail);
});
it("should also fail when throwing an exception", {
throw "But only synchronously!";
});
If you want to test if some part of your code fails to compile, guess what, there is a macro for that:
import buddy.CompilationShould;
class Main extends buddy.SingleSuite
{
public function new() {
describe("Using CompilationShould", {
it("should pass if an expression won't compile", {
CompilationShould.failFor(this.will.not.compile);
});
});
}
}
The method will return a string representation of the compilation failure, or an empty string if compilation succeded, in case you want to test it further.
Exceptions in it
will be handled as above, but if something goes wrong in a before/after
section, Buddy will stop executing the whole suite. It will also count as a failure.
If you're getting an early runtime error, you might want to disable the trace capture that buddy uses. You can do that globally by putting BuddySuite.useDefaultTrace = true
in the beginning of a test class. Then you'll see the traces immediately instead of in the reporter.
Please note that putting code that should be tested outside a describe
, it
or any before/after
block can result in undefined behavior.
Since BDD is also made for non-programmers to use, a common development style is to write empty, or pending tests, and let a programmer implement them later. To do this, just write a string in the it
methods, nothing else. Our previous test class would then look like this:
Main.hx
using buddy.Should;
class Main extends buddy.SingleSuite
{
public function new() {
describe("Using Buddy", {
it("should be a great testing experience");
it("should really make the tester happy");
});
}
}
And the output would be:
PP
Using Buddy
should be a great testing experience (Pending)
should really make the tester happy (Pending)
2 specs, 0 failures, 2 pending
There is also a pending(reason : String)
method available to make a spec pending, similar to fail
.
Classes, suites and specs can all be marked with @include
and @exclude
metadata.
@include
will only run the tests that are marked, removing everything else.@exclude
does the opposite, it removes the marked ones.
If you have a huge test suite, it can be convenient to mark the suite you're currently working on with @include
.
Extending buddy.SingleSuite
is nice and simple, but you can have multiple test classes, and separate them from the main class if you like. Here's how to do it:
Main.hx
import buddy.*;
using buddy.Should;
// Implement "Buddy" and define an array of classes within the brackets:
class Main implements Buddy<[
Tests,
path.to.YourBuddySuite,
AnotherTestSuite,
new SpecialSuite("Constant value", 123)
]> {}
// All test classes should now extend BuddySuite (not SingleSuite)
class Tests extends BuddySuite
{
public function new() {
describe("Using Buddy", {
it("should be a great testing experience");
it("should really make the tester happy");
});
}
}
Enable ANSI color output is easy:
@colorize
class Main extends buddy.SingleSuite {
// ...
}
Or you can do it when compiling with -D buddy-colors
, or disallow it with -D buddy-no-colors
.
The compilation flag will override the metadata, if both are set.
If you have lots of tests, failures may drown in the output. By defining -D buddy-ignore-passing-specs
, only failed and pending test suites will be displayed in the final output.
You can make your own reporter by implementing the buddy.reporting.Reporter interface. Then there are two ways to use it:
@reporter("path.to.your.Reporter")
class Main extends buddy.SingleSuite {
// ...
}
Or do it when compiling with -D reporter=path.to.your.Reporter
.
The compilation flag will override the metadata, if both are set.
buddy.reporting.ConsoleReporter
is the default reporter.
buddy.reporting.ConsoleFileReporter
splits the test progress meter per file, in case you have many test suites in different files.
buddy.reporting.TraceReporter
outputs to trace()
, and is especially useful for CI with flash. If you define -D flash-exit
, the default reporter will be the TraceReporter, and flash will exit if the correct permissions are set. This is tricky to get right, so the easiest way is to use travix.
buddy.reporting.XUnit2Reporting
a reporter for xUnit.net.
Ok, you noticed that it was missing! Using some macro magic, you only need to extend buddy.SingleSuite
, or alternatively implement buddy.Buddy
on your Main class. Then platform-specific code will be generated for waiting until the tests are finished. On all server platforms, exit code 0 will be returned for "all tests passed" and 1 if not, so you can use Buddy in CI tools.
The compiler seems to be a bit too good at optimizing sometimes, especially at the beginning of functions, though this seems to have improved greatly in 3.2. If you have this problem, add a parenthesis after "should", and wrap numbers in parenthesis too.
x.should.be("ok")
-> x.should().be("ok")
123.should.beGreaterThan(100)
-> (123).should.beGreaterThan(100)
Yes, there is special support for utest and general support for all libraries that throws an exception on failure (like mockatoo ). To use utest, just call any utest.Assert
method inside a spec, no need to set up anything else.
It's not possible to do that, since the program has already passed the exception handling code when the exception is thrown. You need to handle asynchronous exceptions yourself and test if something went wrong before calling done
in the spec, or use the fail
method as described in the section "Failing tests".
This usually happens if you're not linking in the correct .ndll
files. An easy fix is to add -lib hxcpp
to your hxml. Another problem could be fixed by adding -D HXCPP_M64
if you're targeting C++ on a 64bit platform (seems to vary between Linux and Win).
Yes, but make sure you know what you're doing, for example some targets requires a wait loop if you have asynchronous tests... Here's a minimal setup for synchronous execution:
import buddy.reporting.ConsoleColorReporter;
class Main {
public static function main() {
var reporter = new ConsoleColorReporter();
var runner = new buddy.SuitesRunner([
new FirstTestSuite(),
new AnotherTestSuite()
], reporter);
runner.run();
#if sys
Sys.exit(runner.statusCode());
#end
}
}
Please make sure that the auto-generated version doesn't work in your case, before doing this.
After my speech about HaxeContracts at WWX2014, I concluded that one does not simply bash unit testing without providing a nice alternative! Having used Jasmine before, I borrowed some if its nice features, and found the perfect aid to implementing async support with the promhx library.
The HaxeContracts library is a nice complement to BDD, so check it out for more information about why most unit testing is waste, and why BDD is a better alternative...!
- Nicer browser reporter
- Your choice! Send me a gmail (ciscoheat) or create an issue here.
Have a good time and much productivity with your new Buddy! :)