Address jQuery UI position dependency

1. +++ b/core/core.libraries.yml
   @@ -653,13 +653,19 @@ jquery.ui.mouse:
   -  js:
   -    assets/vendor/jquery.ui/ui/position-min.js: { weight: -11.8, minified: true }
   

   Would it be possible to add some integration tests with dialog for example? I think it could be simple since we already have quite extensive test coverage here.

2. +++ b/core/misc/jqueryui.position.popper.adapter.es6.js
   @@ -0,0 +1,369 @@
   +      let tipPosition = {
   

   Should we find a more generic name to call the element than tip?

3. +++ b/core/misc/jqueryui.position.popper.adapter.es6.js
   @@ -0,0 +1,369 @@
   +      if (reference === document.body) {
   

   There's over 100 lines inside this if statement which makes reading this a bit challenging 😅 Would it be possible to move at least some of this outside here?

4. +++ b/core/misc/jqueryui.position.popper.adapter.es6.js
   @@ -0,0 +1,369 @@
   +        if (referenceAttach.vertical !== 'center') {
   ...
   +          // If vertical positioning is centered, use top offset and
   +          // base it on halved client height.
   

   IMO this would be easier to read if the if condition was flipped especially given that the comment explaining this is now in the else statement.

5. +++ b/core/tests/Drupal/Nightwatch/Tests/jQueryUIPositionShimTest.js
   @@ -0,0 +1,980 @@
   +const testScenarios = {
   

   Any thoughts on running these tests also on RTL?

Didn't go through the whole patch yet, only looked at the js shim

Taking a step back I'm not sure what is supposed to happen here.

The jquery ui position is a ~560 lines script depends only on jQuery (and "vesion"), it improve the existing jQuery position method: https://git.drupalcode.org/project/drupal/-/blob/9.2.x/core/assets/vendo...

This patch introduces a ~520 lines script that depends on jquery and popperjs.

To me this suggest the complexity of this script is not in how it's implemented but in what the API supports.

What is the net improvement we're looking for? If we do that simply because we want to deprecate jquery ui, why not take the position script used for shims and remove it when we remove the shims no?

Following a meeting with bnjmnm, lauriii, gabor, and myself, todo:

• Take the code from jquery ui and importing in core as much as-is as possible, keep the tests that have been written here.

Couple of details

Looking good, haven't gone though the tests yet

This is truly impressive work, especially considering you created the Nightwatch tests to match to match the jQuery UI qUnit tests.
To compare I tested the new code in position.es6.js and postion.js against the original qUnit tests and both pass.

Just the with scrollbars qUnit test was missing from the Nightwatch tests, but it looks like you already have the scrollx element in place.

Really appreciated the updated postion.es6.js style, especially the destructuring and prettier formatting for ternary statements.
Also very thoughtful to add comments!

Since the code is being "updated" - why not also fix some of the variable names?
hindex for example could be camelCased then it would not need to be added to dictionary.txt
destructure was unused.

These are all in just in test files and could be camelCased or renamed.

• fractionselement
• fractionsparent
• hindex
• largebox
• parentx
• scrollx
• vindex
• withinbugger
• withinsmaller

These are internal variable names and could also be renamed or camelCased to avoid adding to dictionary.txt

• rhorizontal
• roffset
• rpercent
• rposition
• rvertical

I think the only word that needs to be added is flipfit since that's one of the exposed options in position.

My only concerns were addressed, by adding the with scrollbars test and removing all the unnecessary additions to dictionary.txt, and good call re-adding flipfit back as-is, since that is part of the API. This still passes all the original qUnit tests so I'm confident the Nightwatch tests match.

I also found it particularly notable that you created Nightwatch tests to match the visual tests in jQuery UI- the originals weren't even automated!
Goes by so fast that I had to make a screen recording to be able to appreciate it :)

Confirmed that the code has only cosmetic changes between jquery ui and our version. It's the least amount of work in the least disruptive way while still securing our side of things.
Tests are very exhaustive and most if not all the position permutations are tested.

Huge amount of work, well done.
RTBC +1

Committed 9289782 and pushed to 9.2.x. Thanks!