search results: bug of feature?

colak · 2025-01-08 10:32:21

I added some gibberish text for tests I am doing and described on forum.textpattern.com/viewtopic.php?pid=338662#p338662. As it takes longer than I expected to solve the issue, I have earlier today wrapped the text in <txp:if_logged_in>aaaab</txp:if_logged_in>, but txp actually still returns one result. Initially, I thought that this was because of my cache but I also checked it in Tor Browser and the result is there.

I would expect that anything wrapped in if_logged_in would not be visible. Up to now, I only used the tag in pages not articles, and I can understand why the result is returned, but is this actually a correct functionality? The same issue occurs when something is wrapped in txp:hide.

etc · 2025-01-08 15:57:56

colak wrote #338717:

The same issue occurs when something is wrapped in txp:hide.

Yiannis, are you sure to edit the right page/form?

colak · 2025-01-08 17:34:27

etc wrote #338721:

Yiannis, are you sure to edit the right page/form?

My if_logged_in and hide tags are inside an article, and the results are returned, admittedly without showing the ‘hidden words.’ Tested it also in the textpattern demo site by adding <txp:hide>aaaaaa</txp:hide> in the one default article that comes with it and then searched for aaaaaa but also for hide. Easy to test.

As I reported before, I can understand why, but is this something we can programme against? ie do not return any articles in the front end search results for words enclosed in hide tags and only return articles with keywords enclosed in if_logged_in tags for those who are logged in.

etc · 2025-01-08 18:46:58

colak wrote #338723:

do not return any articles in the front end search results for words enclosed in hide tags and only return articles with keywords enclosed in if_logged_in tags for those who are logged in.

Ah. This would be too expensive, since we’d need to parse each body/excerpt before scanning it by search. The best you can presently do is including all ‘hidden’ content via forms or other tags, I think.

kuopassa · 2025-01-09 09:36:01

How about a regex that checks if search result contains <txp:hide>something</txp:hide> or <txp:if_logged_in>something</txp:if_logged_in> (and their short-tag variants), then just hide those tags from search results and what’s between them?

etc · 2025-01-09 11:19:17

That looks a bit complicated, since

presently, <txp:hide /> does not only serve to hide things, but also to postpone processing to the next parser pass
<txp:if_logged_in /> has also group and name attributes to respect
other tags (plugins?) might also contain some ‘hidden’ content, so the problem remains
more important, hiding does not prevent the match detection

A plugin to replace <txp:search_result_excerpt />, maybe?

etc · 2025-01-09 13:30:52

We could, though, parse article body before highlighting search terms. This would effectively hide all ‘hidden’ content, but still not prevent it from matching search queries. Better than nothing?

Bloke · 2025-01-09 14:44:21

Happy to try that if it doesn’t break too many existing site searches.

etc · 2025-01-09 15:17:19

Bloke wrote #338730:

Happy to try that if it doesn’t break too many existing site searches.

It wouldn’t alter search results, just hide ‘hidden’ content in <txp:search_result_excerpt /> output. Easy to do, but possible ramifications make me hesitate. And it wouldn’t solve the OP problem anyway, as I get it.

Textpattern CMS

Textpattern CMS support forum

#1 2025-01-08 10:32:21