Fix phpstan findings in recent pipelines

Please review my MR https://git.drupalcode.org/project/search_api/-/merge_requests/350

I was checking my pipelines and saw one of my issues failed and ended up on this issue. Thanks for tackling this.

Some feedback that I see:

• You are also fixing phpunit, so updating the title and description and link to a job with failing phpunit tests should make this issue clearer.
• Adding @ in front of trigger_error doesn't seem related to phpstan or phpunit failed jobs (but perhaps I'm missing something)
• The change to src/Plugin/views/field/SearchApiEntity.php seems unrelated to phpstan report linked
• Some changes mean the module will only support Drupal 11.3 and up, so a .info.yml file change will be needed I think. Another option is to selectively ignore certain items e.g. the CacheableMetadata to ensure compatibility with older Drupal versions (for now)

That's good feedback, @daften! I left the change to SearchApiEntity.php in place because it is flagged by phpcs, if not by phpstan.

I hope I was now able to fix all the warnings/errors reported by static analysis. For the PHPUnit failures I’ll follow up in #3624636: Fix failing pipelines.

FYI, since the latest commit here (d6ef078a) the Drupal AI module's pipelines fail on Drupal 11.4 with:

TypeError: SearchApiViewsHooks::__construct(): Argument #9 ($memoryCache) must be of type MemoryCacheInterface, MemoryBackend given

It looks like cache.static returns a plain MemoryBackend, which doesn't implement MemoryCacheInterface. Type-hinting CacheBackendInterface might be enough, since only get() and set() are used.

Failing job: https://git.drupalcode.org/project/ai/-/jobs/12712634

Thanks! Diff fixes the fatal error with the type-hint change.

Quick AI check on it and it flagged this though:

Separately, getHandlerMapping() and getFieldHandlerMapping() compare the raw get() return value with is_array(), but CacheBackendInterface::get() returns an object or FALSE, so the static cache never hits. Suggest $mapping = $this->memoryCache->get($cid)->data ?? NULL; (and the same for $mappings).

I have not checked it but could check it asap if it sounds wrong (school pickup time!)

Merged a bit too hastily, but now static analysis should pass again for all tested configurations and no new bugs should have been introduced.
Thanks, everyone!

Oops, sorry, didn’t see the last comments before posting.
The comments about getHandlerMapping() and getFieldHandlerMapping() seem worth pursuing.

Np looks good to me, thank you!

Thanks for reviewing, merged!
Hopefully at least nothing is worse off than before now …