TextTrimmedFormatter not trimming at periods followed by non-breaking space

I can replicate this issue, the system is trimming up to the period(.) if this is moved further into the sentence the trimming happens after the period

Confirmed this is still relevant in D9.5

Will need an issue summary update when a solution is agreed upon.

This doesn't seem to be related to the html character nbsp but that it's one paragraph and the code uses the period in doctor as a breakpoint.

Not sure best way to avoid certain periods, like Dr, Mr, Mrs, but apply at end of sentences.

So one contrib module that may be worth mentioning is https://www.drupal.org/project/smart_trim

Also playing around with the function there’s a line
$rpos = strpos($reversed, strrev($point));
if I replace as
$rpos = strpos($summary, strrev($point)); OR
$rpos = strpos($reversed, $point);
it works then too but I’m sure using reversed as for a reason

Closed #1482178: text_summary does not break on word boundaries if the first sentence is longer than the length of the summary being produced as duplicate moving over credit.

I don't think it's fair to call #1482178: text_summary does not break on word boundaries if the first sentence is longer than the length of the summary being produced duplicate of this one. To me it feels like they are talking about related, but different things.

The problem we see with text_trimmed formatter is that it breaks works when trimming body field with HTML. We have no   there mostly.

Tested the patch from #20 and it haven't changed anything for us - we still see broken words at the end of the field output where the text was trimmed.

I have a similar issue where when importing content. if space does not follow then period then it will be be not be trimmed as described. This took a little while for me to try and debug but as you can see from the screenshots appear like a bug.

e.g. // If the first paragraph is too long, split at the end of a sentence.

So in text.module you have

  // If the first paragraph is too long, split at the end of a sentence.
  $break_points[] = ['. ' => 1, '! ' => 1, '? ' => 1, '。' => 0, '؟ ' => 1];

Not seen 。before

adding '.' => 0, works for me. e.g.

  $break_points[] = ['. ' => 1, '! ' => 1, '? ' => 1, '。' => 0, '.' => 0, '؟ ' => 1];

For my purposes I can change to following, but the dr. problem would still exist.

e.g.

Please see screen shots attached.

Add separate issue for this https://www.drupal.org/project/drupal/issues/3424444

I'm sorry, but I couldn't reproduce the issue before cloning the MR. The text was trimmed at ~600 characters, not at the "Dr." abbreviation. I'm not sure what the issue is, but notably the steps to reproduce the problem are out of date since a text-with-summary field has to be set up by a tester. Could we have some updated steps to reproduce, please?

Updated the steps, let me know if it still doesn't show.

Thank you. That did help. After following the updated steps I was able to reproduce the issue. I'm not exactly certain where I went wrong, but I know that previously I left the trimming at the default 600 characters.

I performed manual testing with the MR applied. The intent behind the change is that the formatter should trim at non-breaking spaces (NBSP),  , because CKEditor inserts those after a sentence's full stop. I became concerned about two things:

1. Are there other scenarios in which CKEditor inserts an NBSP?
2. What happens if a content editor manually inserts an NBSP?

I searched, but couldn't find an answer to the first question about whether CKEditor inserts them in other scenarios. But I tested the second question myself. I had to edit the sample text from the issue summary slightly due to the positioning of full stops near the 300 character breakpoint, but after a little work I was able to produce text containing a manually inserted NBSP that resulted in trimming not at a sentence boundary. The sample text was this:
Dr. Cras ultricies ligula sed magna dictum porta. Curabitur non nulla sit amet nisl tempus convallis quis ac lectus. Pellentesque in ipsum id orci porta dapibus. Vestibulum ac diam sit amet quam vehicula elementum sed sit amet dui. Nulla porttitor. Cras ultricies ligula sed magna dictum porta. Nulla quis lorem ut libero malesuada feugiat. Proin eget tortor risus. Curabitur non nulla sit amet nisl tempus convallis quis ac lectus. Proin eget tortor risus. Praesent sapien massa, convallis a pellentesque nec, egestas non nisi. Vivamus suscipit tortor eget felis porttitor volutpat. Quisque velit nisi, pretium ut lacinia in, elementum id enim. Curabitur arcu erat, accumsan id imperdiet et, porttitor at sem. Quisque velit nisi, pretium ut lacinia in, elementum id enim. Vestibulum ante ipsum primis in faucibus orci luctus et ultrices posuere cubilia Curae; Donec velit neque, auctor sit amet aliquam vel, ullamcorper sit amet ligula. Curabitur aliquet quam id dui posuere blandit. Curabitur aliquet quam id dui posuere blandit. Curabitur non nulla sit amet nisl tempus convallis quis ac lectus. Pellentesque in ipsum id orci porta dapibus.

It resulted in this trimmed display value (with a 300 character trim length) :

Dr. Cras ultricies ligula sed magna dictum porta. Curabitur non nulla sit amet nisl tempus convallis quis ac lectus. Pellentesque in ipsum id orci porta dapibus. Vestibulum ac diam sit amet quam vehicula elementum sed sit amet dui. Nulla porttitor. Cras ultricies

So my question is: why was ' '=> 6, added to the breakpoints array instead of '. '=> 6,? That could help ensure that the breakpoint is added after a sentence. I tested it on my local.

I applied the feedback as it makes sense and test seems to still pass.

What do you think about changing the text in the test to the examples I gave in #40? That way we ensure there's no regression to that fix.

Also good call made the test update

All right, thank you for considering my feedback. I have nothing further to comment on. It looks good to me.

Setting work for re-enabling cspell. I updated credit while here as well.

I decided to move things around a bit in the test so that only one pair of cspell disable/enable is needed.

The failing test is a core/modules/ckeditor5/tests/src/FunctionalJavascript/CKEditor5Test.php, which I am assuming is unrelated and a random one. I will restart the tests.

That reorganization of the test looks good. The changes make sense for ensuring cspell is checking things that ought to be checked.

I think the issue title and IS might be inaccurate, or at least not up to date? Looks like the MR addresses situations where a period is followed by a non-breaking space, but it's still entirely possible that an abbreviated word like "Dr. " might be where text is trimmed.

Thing the title is accurate enough as it does appear abbreviations like Dr. are triggering this.

Tweaked the summary some

The only question I see here is if #1482178: text_summary does not break on word boundaries if the first sentence is longer than the length of the summary being produced is a duplicate. As I read that issue it is not a duplicate of this one. If that is the case, then that issue needs to be re-opened and credit adjusted here.

Committed and pushed bffaae1 to main and 000c505 to 11.x. Thanks!

Consulting other committers on whether to backport to 11.4.

was about to it but let me know!

Consulted with @longwave and we're thinking not to backport. It's an edge-case bug that's been around awhile, and it's possible some sites are relying on the current behavior.