Failed to rename tag which contains Asian characters

[cuixiping]

ENV:

• PHP version: 8.3
• Database: MariaDB 10.4

Steps

1. Create a new post with Asian characters tag, for example 文字.
2. Go to tags manager, rename the tag.
3. There is no change when back to tags list.

My fix

I solved it by doubling the backslashes in "tags.php". I am not very sure it's correct or not.

            $results = SQL::current()->select(
                tables:"post_attributes",
                fields:"post_id",
                conds:array(
                    "name" => "tags",
---                 "value LIKE" => $this->tags_name_match($_POST['original'])
+++                 "value LIKE" => str_replace("\\","\\\\",$this->tags_name_match($_POST['original']))
                )
            )->fetchAll();

[xenocrat]

Thanks for the report. I see the problem: the tag names are stored as JSON, which will encode any non-ASCII character as a Unicode codepoint using the hex notation "\uXXXXXX". We need to tell the RDBMS we are searching for a literal backslash there.

You're solution is fine for a quick fix, but I've added a commit to do this in a more correct way. It's still not working right on PostgreSQL - I need to investigate that further.

[xenocrat]

Better fix committed just now. :-) The problem is specifically related to LIKE... ESCAPE. Thanks for reporting this!

[xenocrat]

I've decided to specify | as the default escape char for LIKE statements. QueryBuilder.php and tags.php have both been updated accordingly.

[cuixiping]

Great! Thanks for your quickly fix.

And I think it's not very necessary to escape unicode in json.

json_encode($a, JSON_UNESCAPED_UNICODE)

[xenocrat]

Yes indeed, it isn't needed now that Chyrp Lite enforces utf8mb4 character set in MySQL. But it's better not to change it, or it will cause a problem with discovering existing tags. With this escaping issue resolved, all is well. :-)

[cuixiping]

We can upgrade all existing tags to unescaped. With unescaped tags, program SQL is simpler, and database content is more readable for human who use database tools.