Modules display ID validation does not work with Paragraphs anymore

Just looking at this a bit, I don't think isEmpty() was actually validating though. There is a validate() method on the grantparent class of ViewsReferenceItem, so perhaps we should be extending that method.

I think I worded it incorrectly. The isEmpty() method was used to check whether the display_id was empty before the validation took place.
Which class do you mean? I couldn't find a validate() method in the class that ViewsReferenceItem extends.

nilesh.addweb → made their first commit to this issue’s fork.

Thanks for opening MR, is this ready for review?

Yes, ready for review. we can use validateDisplayId method which is present in ViewsReferenceTrait.php.
Please review MR !67.

Thanks for the progress on this! Tested this out, couple issues:

• This only works if the paragraph is the first one
• The validation from setErrorByName doesn't actually set the error to the display id field

@scott_euser, Thanks for review, I have updated MR.

I was getting the following warning:

Warning: Undefined array key "subform" in Drupal\viewsreference\Plugin\Field\FieldWidget\ViewsReferenceSelectWidget::validateDisplayId() (line 216 of modules/contrib/viewsreference/src/Plugin/Field/FieldWidget/ViewsReferenceTrait.php).
Drupal\viewsreference\Plugin\Field\FieldWidget\ViewsReferenceSelectWidget::validateDisplayId() (Line: 53)
Drupal\viewsreference\Plugin\Field\FieldWidget\ViewsReferenceSelectWidget::validateElement()

That's why I added isset() to $field_value['subform'].

Besides that it workes for me. But I think we could try a different approach by using the Validation API. I will open a new MR with the changes I made. It does not work 100% correctly right now. It works great with the autocomplete field. When you use the select list the error shows up but all fields of the widget somehow show the red border when the error appears. I'm not sure if this might be a core problem or if the way the widget is defined might be the problem or something else. Maybe I'm missing something. There is not much documentation on this.
It would be great if someone could take a look at this. We should also write tests for this scenario.

In case anyone needs a working patch for this issue, I created one from the first merge request.

Before I retest would be good to:

1. Hide the branch that isn't in use (its confusing to have 2)
2. Fix test coverage

Thanks!