How to use PHP preg_replace on a formatted number based on search query while ignoring comma and decimal?

Given how long is takes to get an answer, I think we can conclude that this isn't a trivial problem.

I assume, although you don't explicitly say so, that you actually only want to highlight formatted numbers, and that these numbers aren't embedded in a bigger text. I also assume you only want to highlight one part of the number, not multiple parts. So in 12,341,234.00 only the first 1234 will be highlighted.

My solution is a real algorithm, and not based on complex build-in functions or regular expressions. That would be possible of course, but I avoided them.

function highlight(string $search, string $number): string
{
    $searchBegin = NULL;
    $searchIndex = 0;
    $search      = str_replace([',', '.'], '', $search);
    foreach (str_split($number) as $numberIndex => $character) {
        if ($character == $search[$searchIndex]) {
            if (is_null($searchBegin)) {
                $searchBegin = $numberIndex;
            }    
            $searchIndex++;
            if ($searchIndex == strlen($search)) {
               return substr($number, 0, $searchBegin) .
                      '<span style="background: yellow">' .
                      substr($number, $searchBegin, $numberIndex - $searchBegin + 1) .
                      '</span>' .
                      substr($number, $numberIndex + 1);                
            }
        } elseif (!in_array($character, [',', '.']))  {
            $searchBegin = NULL;
            $searchIndex = 0;
        }
    }
    return $number;
}

Live demo: https://3v4l.org/StFn7

I think the code is quite self-explanatory. Just follow what it does and you'll understand it. There's nothing mysterious about it.

This code isn't optimized, and doesn't check for errors. It is intended as a starting point for further optimization. str_split() is handy, but not very fast, you can easily make this algorithm faster by using a for loop and a string index. Being able to highlight multiple sections would be the first thing I would do. You could do this recursively on the part of the $number after the match. Like so: https://3v4l.org/aRZtN

This code might be longer than you expect, but that doesn't make it bad or slow code. Sometimes a good old-fashioned algorithm wins the day.