Use the PCRE-escaped non-breaking space so the /x modifier doesn’t strip it out

[wvell]

We encountered weird behaviour on our mac development stations where the space-collapse-fix would add a space between every character.

The reason seems to be that the /x modifier strips the U::NO_BREAK_SPACE. This only seems to happen on mac when using setlocale before running this function. On ubuntu with all the versions the same(php/pcre) it works normally.

I have attached a small test script that shows the error on my system.

There are multiple issues touching this issue:

• Strange   display issues in heading tags nystudio107/craft-typogrify#33
• PHP 7.3 #109

PHP version: 8.4.8
PCRE library: 10.45 2025-02-05

<?php

$nbsp  = "\xC2\xA0";
$nbspEscaped  = "\\x{00A0}";
$input = "a{$nbsp}b";

// The result should become: a-b
// First without setlocale (the default locale is C.UTF-8 on my mac).
echo preg_replace("/{$nbsp}/xu", '-', $input) . "\n"; // This works and outputs a-b

// Set the locale to the same that i currently have.
setlocale(LC_CTYPE, "C.UTF-8");
echo preg_replace("/{$nbsp}/xu", '-', $input) . "\n"; // Wrong output -a-b-
echo preg_replace("/{$nbspEscaped}/xu", '-', $input) . "\n"; // Correct output a-b

// It looks like any locale i set triggers the error.
setlocale(LC_CTYPE, "en_US.UTF-8");
echo preg_replace("/{$nbsp}/xu", '-', $input) . "\n"; // Wrong output -a-b-
echo preg_replace("/{$nbspEscaped}/xu", '-', $input) . "\n"; // Correct output a-b

[mundschenk-at]

Thank you for the report, I'll have a closer look in the next few days. (I'd prefer a solution without a magic constant if possible.)