test_naming.md

• The inclusion of _it_ seems unnecessary in the context of these isolated tests, UNLESS additional clarity is desired, in which case this should include the function name under test, which seems pretty common.

it is often used in JS tests/specs in the sense:

describe( 'disableBlockEditorForNavigationPostType', () => {
    // Some suites use "it"
    it( 'should only disable block editor for navigation post types',      () => {})
    // Some suites use "test"
    test( 'it should only disable block editor for navigation post types', () => {})
    // For suites using "test", removing "it" is common
    test( 'should only disable block editor for navigation post types',    () => {})
})

Converted to PHPUnit, the third example above would look like:

class Tests_Editor_DisableBlockEditorForNavigationPostType extends WP_UnitTestCase {
    public function test_should_only_disable_block_editor_for_navigation_post_types() {
    }
}

Each of these migrated tests have the same test method name, test_it_correctly_handles_different_post_types(), which seems awkward compared with others in the project. Maybe it's just me, but an alternative might be to simplify it to test_handle_different_post_types().

My issue with handle - and admittedly, I've used it myself in the past, is that it doesn't describe the expected behaviour.

i.e. test_handle_different_post_types() vs test_should_only_disable_block_editor_for_navigation_post_types().

By using test_should_ and test_should_not_ prefixes where possible, that means:

• Raw output:

Tests_Editor_DisableBlockEditorForNavigationPostType::test_should_only_disable_block_editor_for_navigation_post_types
Tests_Editor_DisableBlockEditorForNavigationPostType::test_should_not_disable_block_editor_for_missing_post_types

• --testdox output:

_Editor_DisableBlockEditorForNavigationPostType
 ✔ Should only disable block editor for navigation post types
 ✔ Should not disable block editor for missing post types

Theoretically, _should_ may not be necessary, but it gives a clear indication of the expected result before the reader even knows the remaining context, and we'll often internally say "Okay, this function should only / should not do something", so this saves the reader time.

+1 on the suggestion to use test_should_ and test_should_not_. Also, thanks, @costdev for pointing out the existing docs guideline concerning test names.

@anton-vlasenko, I agree the naming convention could be made more obvious in the docs. While no consensus/handbook update is necessary for us to use the test_should_ convention here, I think it's worth looking into a future docs update that recommends the converse test_should_not_, as well as points out the clearly established convention for including the method name (particularly in combined test class files) that @costdev alluded to. I'm working on Test Handbook updates that could incorporate this.

IMO, modern IDEs make it easy to understand which class a particular method belongs to.

@anton-vlasenko It's mainly for clarity that we can make the test method names as unique as possible. When running tests locally or in CI/CD, the self-descriptive names could save time in understanding the failure 🙂.

Just an update that I've made this Gist public after realising that I had created it as a "Secret" Gist - My bad.

It's mainly for clarity that we can make the test method names as unique as possible. When running tests locally or in CI/CD, the self-descriptive names could save time in understanding the failure 🙂.

IMO, since the class name already shows in the list of failures, that's not an issue for me.

However, running a specific test becomes a little more awkward via CLI.

Consider:

phpunit --filter should_return_false_for_multiple_post_types

It's quite possible that this test method name could be used in multiple test classes.

So, you would either have to have an IDE that allows you to click to run a specific test, or use one of the following:

# Meaning manually typing the class name or copying/pasting the class and method names separately.
phpunit --filter Test_Class_Name::should_return_false_for_multiple_post_types

# Not ideal.
phpunit --group <ticket_no> --filter should_return_false_for_multiple_post_types

# Not possible due to file name conventions not matching test class names in a way that PHPUnit understands (I don't know if there's a way around this already?)
phpunit /path/to/test/file.php

On a personal note, I prefer just using should_return_false_for_multiple_post_types without the tested method name, but I can't think of a way that isn't more difficult when running tests.

I've renamed the test methods.
But I didn't create test_should_not methods.
Why?
That would lead to code duplication.
The test classes would need to have the same code for test_should and test_should_not methods.
If the intention is to save time in understanding the failure, I think we are fine because these particular tests use named data sets and unique error messages.
@ironprogrammer @costdev Please review the code and let me know what you think.
WordPress/wordpress-develop#3005