In Akamai tag purger apply formatting after the tags has been altered by PURGE_CREATION

This is a great catch, thank you for the patch! I'd really like a test for something being altered/ added though?

Okay, if you can't get round to making a test - I'll see if I can find time over the holidays.

In addition the feedback from @pobster above, I've noticed one other issue with this solution:

     $formatter = \Drupal::service('akamai.helper.cachetagformatter');
     foreach ($invalidations as $invalidation) {
       $invalidation->setState(InvalidationInterface::PROCESSING);
-      $tag = $formatter->format($invalidation->getExpression());
+      $tag = $invalidation->getExpression();
       if (mb_strlen($tag) > 128) {
         $this->logger->warning('Cache Tag %tag has exceeded the Akamai 128 character tag maximum length.', ['%tag' => $tag]);
       }

Since the formatted version is what is getting sent to Akamai, the mb_strlen($tag) check needs to occur after the formatting, not before.

In its present state CacheTagFormatter::format() shouldn't modify the length of the cache tag at all, but we can't guarantee that will always be the case, especially if #3154613: Add cache-tag prefix is introduced.

So we need to move the mb_strlen($tag) check inside this loop instead:

+    // Apply the formatting after the tags have been altered.
+    $formatted_tags_to_clear = [];
+    foreach ($tags_to_clear as $tag) {
+      $formatted_tags_to_clear[] = $formatter->format($tag);
+    }

This patch #6 is not working for the latest version 5.x, so created new patch with same code changes mentioned inside the #6. This is working for me.

inside comment #15 i had missed 1 line to remove so adding new patch for the same.

I'm afraid it's still going to need a specific test though.

Plus, see this previous comment:

Since the formatted version is what is getting sent to Akamai, the mb_strlen($tag) check needs to occur after the formatting, not before.

Disclosure used AI to help right the unit test.

But during review found that typecasting to (string) means we don't need to update the one test

Sure, LGTM