Reporting
SimpleTest pretty much follows the MVC-ish pattern
(Model-View-Controller).
The reporter classes are the view and the model is your
test cases and their hiearchy.
The controller is mostly hidden from the user of
SimpleTest unless you want to change how the test cases
are actually run, in which case it is possible to
override the runner objects from within the test case.
As usual with MVC, the controller is mostly undefined
and there are other places to control the test run.
Table of Contents
Reporting results in HTML
The default HTML display is minimal in the extreme.
It reports success and failure with the conventional red and
green bars and shows a breadcrumb trail of test groups
for every failed assertion.
Here's a fail...
File test
Fail: createnewfile->True assertion failed.
1/1 test cases complete.
0 passes, 1 fails and 0 exceptions.
And here all tests passed...
File test
1/1 test cases complete.
1 passes, 0 fails and 0 exceptions.
The good news is that there are several points in the display
hiearchy for subclassing.
For web page based displays there is the
HtmlReporter class with the following
signature...
public makeDry(boolean $is_dry) { ...
}
public void paintFail(string $message) { ...
}
public void paintPass(string $message) { ...
}
protected string getCss() { ...
}
}
Here is what some of these methods mean. First the display methods
that you will probably want to override...
HtmlReporter(string $encoding)
is the constructor.
Note that the unit test sets up the link to the display
rather than the other way around.
The display is a mostly passive receiver of test events.
This allows easy adaption of the display for other test
systems beside unit tests, such as monitoring servers.
The encoding is the character encoding you wish to
display the test output in.
In order to correctly render debug output when
using the web tester, this should match the encoding
of the site you are trying to test.
The available character set strings are described in
the PHP html_entities()
function.
void paintHeader(string $test_name)
is called once at the very start of the test when the first
start event arrives.
The first start event is usually delivered by the top level group
test and so this is where $test_name
comes from.
It paints the page title, CSS, body tag, etc.
It returns nothing (void).
void paintFooter(string $test_name)
Called at the very end of the test to close any tags opened
by the page header.
By default it also displays the red/green bar and the final
count of results.
Actually the end of the test happens when a test end event
comes in with the same name as the one that started it all
at the same level.
The tests nest you see.
Closing the last test finishes the display.
void paintMethodStart(string $test_name)
is called at the start of each test method.
The name normally comes from method name.
The other test start events behave the same way except
that the group test one tells the reporter how large
it is in number of held test cases.
This is so that the reporter can display a progress bar
as the runner churns through the test cases.
void paintMethodEnd(string $test_name)
backs out of the test started with the same name.
void paintFail(string $message)
paints a failure.
By default it just displays the word fail, a breadcrumbs trail
showing the current test nesting and the message issued by
the assertion.
void paintPass(string $message)
by default does nothing.
string getCss()
Returns the CSS styles as a string for the page header
method.
Additional styles have to be appended here if you are
not overriding the page header.
You will want to use this method in an overriden page header
if you want to include the original CSS.
There are also some accessors to get information on the current
state of the test suite.
Use these to enrich the display...
array getTestList()
is the first convenience method for subclasses.
Lists the current nesting of the tests as a list
of test names.
The first, top level test case, is first in the
list and the current test method will be last.
integer getPassCount()
returns the number of passes chalked up so far.
Needed for the display at the end.
integer getFailCount()
is likewise the number of fails so far.
integer getExceptionCount()
is likewise the number of errors so far.
integer getTestCaseCount()
is the total number of test cases in the test run.
This includes the grouping tests themselves.
integer getTestCaseProgress()
is the number of test cases completed so far.
One simple modification is to get the HtmlReporter to display
the passes as well as the failures and errors...
function paintPass($message) {
parent::paintPass($message);
print "<span class=\"pass\">Pass</span>: ";
$breadcrumb = $this->getTestList();
print
implode("->", $breadcrumb);
print "->$message<br />\n";
}
protected function getCss() {
return parent::getCss() . ' .pass { color: green; }';
}
}
One method that was glossed over was the makeDry()
method.
If you run this method, with no parameters, on the reporter
before the test suite is run no actual test methods
will be called.
You will still get the events of entering and leaving the
test methods and test cases, but no passes or failures etc,
because the test code will not actually be executed.
The reason for this is to allow for more sophistcated
GUI displays that allow the selection of individual test
cases.
In order to build a list of possible tests they need a
report on the test structure for drawing, say a tree view
of the test suite.
With a reporter set to dry run that just sends drawing events
this is easily accomplished.
Extending the reporter
Rather than simply modifying the existing display, you might want to
produce a whole new HTML look, or even generate text or XML.
Rather than override every method in
HtmlReporter we can take one
step up the class hiearchy to SimpleReporter
in the simple_test.php source file.
A do nothing display, a blank canvas for your own creation, would
be...
require_once('simpletest/simpletest.php');
function paintHeader($test_name) { }
function paintFooter($test_name) { }
function paintStart($test_name, $size) {
parent::paintStart($test_name, $size);
}
function paintEnd($test_name, $size) {
parent::paintEnd($test_name, $size);
}
function paintPass($message) {
parent::paintPass($message);
}
function paintFail($message) {
parent::paintFail($message);
}
function paintError($message) {
parent::paintError($message);
}
function paintException($exception) {
parent::paintException($exception);
}
}
No output would come from this class until you add it.
The catch with using this low level class is that you must
explicitely invoke it in the test script.
The "autorun" facility will not be able to use
it's runime context (whether it's running in a web browser
or the command line) to select the reporter.
You explicitely invoke the test runner like so...
<?php
require_once('simpletest/autorun.php');
$test->addFile('tests/file_test.php');
$test->run(new MyReporter());
?>
...perhaps like this...
<?php
require_once('simpletest/simpletest.php');
require_once('my_reporter.php');
function __construct() {
parent::__construct();
$this->addFile('tests/file_test.php');
}
}
$test = new MyTest();
$test->run(new MyReporter());
?>
We'll show how to fit in with "autorun" later.
The command line reporter
SimpleTest also ships with a minimal command line reporter.
The interface mimics JUnit to some extent, but paints the
failure messages as they arrive.
To use the command line reporter explicitely, substitute it
for the HTML version...
<?php
require_once('simpletest/autorun.php');
$test->addFile('tests/file_test.php');
?>
Then invoke the test suite from the command line...
php file_test.php
You will need the command line version of PHP installed
of course.
A passing test suite looks like this...
File test
OK
Test cases run: 1/1, Passes: 1, Failures: 0, Exceptions: 0
A failure triggers a display like this...
File test
1) True assertion failed.
in createNewFile
FAILURES!!!
Test cases run: 1/1, Passes: 0, Failures: 1, Exceptions: 0
One of the main reasons for using a command line driven
test suite is of using the tester as part of some automated
process.
To function properly in shell scripts the test script should
return a non-zero exit code on failure.
If a test suite fails the value false
is returned from the SimpleTest::run()
method.
We can use that result to exit the script with the desired return
code...
<?php
require_once('simpletest/autorun.php');
$test->addFile('tests/file_test.php');
?>
Of course we wouldn't really want to create two test scripts,
a command line one and a web browser one, for each test suite.
The command line reporter includes a method to sniff out the
run time environment...
<?php
require_once('simpletest/autorun.php');
$test->addFile('tests/file_test.php');
}
?>
This is the form used within SimpleTest itself.
When you use the "autorun.php", and no
test has been run by the end, this is pretty much
the code that SimpleTest will run for you implicitely.
In other words, this is gives the same result...
<?php
require_once('simpletest/autorun.php');
function __construct() {
parent::__construct();
$this->addFile('tests/file_test.php');
}
}
?>
Remote testing
SimpleTest ships with an XmlReporter class
used for internal communication.
When run the output looks like...
<?xml version="1.0"?>
<run>
<group size="4">
<name>Remote tests</name>
<group size="4">
<name>Visual test with 48 passes, 48 fails and 4 exceptions</name>
<case>
<name>testofunittestcaseoutput</name>
<test>
<name>testofresults</name>
<pass>This assertion passed</pass>
<fail>This assertion failed</fail>
</test>
<test>
...
</test>
</case>
</group>
</group>
</run>
To get your normal test cases to produce this format, on the
command line add the --xml flag.
php my_test.php --xml
You can do teh same thing in the web browser by adding the
URL parameter xml=1.
Any true value will do.
You can consume this format with the parser
supplied as part of SimpleTest itself.
This is called SimpleTestXmlParser and
resides in xml.php within the SimpleTest package...
<?php
require_once('simpletest/xml.php');
...
$parser->parse($test_output);
?>
The $test_output should be the XML format
from the XML reporter, and could come from say a command
line run of a test case.
The parser sends events to the reporter just like any
other test run.
There are some odd occasions where this is actually useful.
Most likely it's when you want to isolate a problematic crash
prone test.
You can collect the XML output using the backtick operator
from another test.
In that way it runs in it's own process...
<?php
require_once('simpletest/xml.php');
} else {
}
$parser->parse(`php flakey_test.php --xml`);
?>
Another use is breaking up large test suites.
A problem with large test suites is thet they can exhaust
the default 16Mb memory limit on a PHP process.
By having the test groups output in XML and run in
separate processes, the output can be reparsed to
aggregate the results into a much smaller footprint top level
test.
Because the XML output can come from anywhere, this opens
up the possibility of aggregating test runs from remote
servers.
A test case already exists to do this within the SimpleTest
framework, but it is currently experimental...
<?php
require_once('../remote.php');
require_once('simpletest/autorun.php');
$test_url = ...;
$dry_url = ...;
function __construct() {
$test_url = ...
parent::__construct($test_url, $test_url . ' --dry');
}
}
?>
The RemoteTestCase takes the actual location
of the test runner, basically a web page in XML format.
It also takes the URL of a reporter set to do a dry run.
This is so that progress can be reported upward correctly.
The RemoteTestCase can be added to test suites
just like any other test suite.