##<test-fixture>
The <test-fixture>
element can simplify the exercise of consistently
resetting a test suite's DOM.
See: Documentation.
To use it, wrap the test suite's DOM as a template:
<test-fixture id="SomeElementFixture">
<template>
<some-element id="SomeElementForTesting"></some-element>
</template>
</test-fixture>
Now, the <test-fixture>
element can be used to generate a copy of its
template:
<script>
describe('<some-element>', function () {
var someElement;
beforeEach(function () {
document.getElementById('SomeElementFixture').create();
someElement = document.getElementById('SomeElementForTesting');
});
});
</script>
Fixtured elements can be cleaned up by calling restore
on the <test-fixture>
:
afterEach(function () {
document.getElementById('SomeElementFixture').restore();
// <some-element id='SomeElementForTesting'> has been removed
});
<test-fixture>
will create fixtures from all of its immediate <template>
children. The DOM structure of fixture templates can be as simple or as complex
as the situation calls for.
Even simpler usage in Mocha
In Mocha, usage can be simplified even further. Include test-fixture-mocha.js
after Mocha in the <head>
of your document and then fixture elements like so:
<script>
describe('<some-element>', function () {
var someElement;
beforeEach(function () {
someElement = fixture('SomeElementFixture');
});
});
</script>
Fixtured elements will be automatically restored in the afterEach
phase of the
current Mocha Suite
.
Data-bound templates
Data-binding systems are also supported, as long as your (custom) template
elements define a stamp(model)
method that returns a document fragment. This
allows you to stamp out templates w/ custom models for your fixtures.
For example, using Polymer's dom-bind
:
<test-fixture id="bound">
<dom-bind>
<template>
<span>{{greeting}}</span>
</template>
</dom-bind>
</test-fixture>
You can pass an optional context argument to create()
or fixture()
to pass
the model:
var bound = fixture('bound', {greeting: 'ohai thurr'});
The problem being addressed
Consider the following web-component-tester
test suite:
<!doctype html>
<html>
<head>
<title>some-element test suite</title>
</head>
<body>
<some-element id="SomeElementForTesting"></some-element>
<script type="module">
import '../some-element.js';
describe('<some-element>', function () {
var someElement;
beforeEach(function () {
someElement = document.getElementById('SomeElementForTesting');
});
it('can receive property `foo`', function () {
someElement.foo = 'bar';
expect(someElement.foo).to.be.equal('bar');
});
it('has a default `foo` value of `undefined`', function () {
expect(someElement.foo).to.be.equal(undefined);
});
});
</script>
</body>
</html>
In this contrived example, the suite will pass or fail depending on which order
the tests are run in. It is non-deterministic because someElement
has
internal state that is not properly reset at the end of each test.
It would be trivial in the above example to simply reset someElement.foo
to
the expected default value of undefined
in an afterEach
hook. However, for
non-contrived test suites that target complex elements, this can result in a
large quantity of ever-growing set-up and tear-down boilerplate.