Optimizing unit tests which testing same methods with different parameters - php

Question

I have lots of PHP classes like this:
class Article
{
public function fetchAll($params) {
if ($params["client"] == "mobile") {
return [
"author" => "",
"body" => ""
];
}
return [
"author" => "",
"body" => "",
"publishedOn => """
];
}
}
As you see fetchAll() method returning different fields according to $params["client"].
I need to check existence of these fields. I have 3 options to achieve my goal.
(Note: All codes simplified, dont worry about syntax errors.)
Method 1 : Different test methods for different clients
class ArticleTest extends ...
{
public function testFetchAll() {
$params = [];
$data = $article->fetchAll($params);
$this->assertEqual(array_keys($data), ["author","body","publishedOn"]);
}
public function testFetchAll2() {
$params = ["client" => "mobile"];
$data = $article->fetchAll($params);
$this->assertEqual(array_keys($data), ["author","body"]);
}
}
Cons: Using same codes again and again, it's a really bad practice. No need to explain. Refactoring is really difficult. Also this method will cause dozens of test methods.
Method 2 : Using for-loop for different clients
class ArticleTest extends ...
{
public function testFetchAll() {
for ($i=0; $i<2; $i++) {
$params = [];
if ($i == 1) $params["client"] = "mobile";
$data = $article->fetchAll($params);
if ($i == 0)
$this->assertEqual(array_keys($data), ["author","body","publishedOn"]);
else
$this->assertEqual(array_keys($data), ["author","body"]);
}
}
}
Cons : I'm not sure if using loops on PHPUnit tests is a good idea. Also code readability decreased.
Method 3 : Different test cases / test files for different clients
// articleTest.php
class ArticleTest extends ...
{
public function testFetchAll() {
$params = [];
$data = $article->fetchAll($params);
$this->assertEqual(array_keys($data), ["author","body","publishedOn"]);
}
}
// articleTest2.php
class ArticleTest2 extends ...
{
public function testFetchAll() {
$params = ["client" => "mobile"];
$data = $article->fetchAll($params);
$this->assertEqual(array_keys($data), ["author","body"]);
}
}
Cons: This method causing dozens of test files, and using same codes again and again. Refactoring is really difficult.
Performance Comparison
Method 1 : Time: 127 ms, Memory: 8.00Mb
Method 2 : Time: 125 ms, Memory: 8.00Mb
Method 3 : Time: 96 ms, Memory: 8.00Mb
I'm trying to understand which method is better with lots of classes like this and waiting for a discussion about optimizing unit tests in this situation.

Answer 1

EDIT: Even if your question focus on optimization, there is actually a broadly accepted best practice for your case so that's what I have shown in my answer.
Just use dataProviders. With those you can execute the same test, passing arguments to the test method. The arguments can be variations on the params or / and the expected result.
So you could do this:
public function fetchAllTestData() {
return [
// case 1
[
[], // params
['author', 'body', 'publishedon'] // expectedKeys
],
// case 2
[
['client' => 'mobile'], // params
['author', 'body'] // expectedKeys
],
];
}
/**
* #dataProvider fetchAllTestData
*/
public function testFetchAll(array $params, array $expectedKeys) {
$data = $article->fetchAll($params);
$this->assertEqual($expectedKeys, array_keys($data));
}
This way the test will be executed independently for each dataProvider row. This means that if the first case fails, the second will be tested so you will know if it also fails or not. If you use loops inside the test case, whenever there is a failure, the rest of the cases won't be tested.
Take a look at dataProviders documentation.
NOTE When writing tests please pay attention to the assert methods parameters ordering. Usually the expected data comes first. If you pass them in reversed order the error messages will be misleading and the arrays and objects comparison will tell that there are missing rows when in fact the rows should not be there.