Snapshot testing for Optimization Detective

[westonruter]

With https://github.com/WordPress/performance/pull/1405, the Optimization Detective tests were refactored to be easier to maintain by putting the test cases into separate files. However, when changes occur to the actual HTML output they remain tedious to update. We should consider using phpunit-snapshot-assertions to make this easier. There is even an assertMatchesHtmlSnapshot() assertion we can use instead of assertEquals which does normalization which would reduce the need to use remove_initial_tabs. (Nevertheless, the HtmlDriver is using PHP DOM which has limitations.)

Relevant code that this would touch (in addition to the test-cases subdirectories):

1. https://github.com/WordPress/performance/blob/e34711dbf286a16523b3c3d35462e0e79ea4c6f9/plugins/embed-optimizer/tests/test-optimization-detective.php#L40-L80
2. https://github.com/WordPress/performance/blob/e34711dbf286a16523b3c3d35462e0e79ea4c6f9/plugins/image-prioritizer/tests/test-helper.php#L32-L65
3. https://github.com/WordPress/performance/blob/e34711dbf286a16523b3c3d35462e0e79ea4c6f9/plugins/optimization-detective/tests/test-optimization.php#L188-L245
4. https://github.com/WordPress/performance/blob/e34711dbf286a16523b3c3d35462e0e79ea4c6f9/plugins/auto-sizes/tests/test-optimization-detective.php#L43-L122

[swissspidy]

There is even an assertMatchesHtmlSnapshot() assertion we can use instead of assertEquals which does normalization which would reduce the need to use remove_initial_tabs. (Nevertheless, the HtmlDriver is using PHP DOM which has limitations.)

As a start we could also think about writing our own custom assertions first, or write a custom driver for that package.

[westonruter]

It's interesting that assertMatchesHtmlSnapshot() ends up using assertEquals on the serialized HTML, rather than comparing DOMDocument instances.

I recall that @dmsnell also maybe has implemented such an HTML-equivalency algorithm using HTML (Tag?) Processor?

[dmsnell]

@westonruter we have primitive forms with the Tag Processor, but I think that the HTML Processor will, for the most part, support all^† HTML in 6.7, making it possible to propose something in Core for a new assertion.

PHP 8.4 will have a spec-compliant HTML5 parser based on lexbor which will also do the trick and support everything, but in most cases I believe we will have an effective tool with the HTML API, but it doesn't exist yet in a proposal.