A more efficient string cleaning Regex in PHP - php

Question

Okay, I was hoping someone could help me with a little regex-fu.
I am trying to clean up a string.
Basically, I am:
Replacing all characters except A-Za-z0-9 with a replacement.
Replacing consecutive duplicates of the replacement with a single instance of the replacement.
Trimming the replacement from the beginning and end of the string.
Example Input:
(&&(%()$()#&#&%&%%(%$+-_The dog jumped over the log*(&)$%&)#)##%&)&^)##)
Required Output:
The+dog+jumped+over+the+log
I am currently using this very discombobulated code and just know there is a much more elegant way to accomplish this....
function clean($string, $replace){
$ok = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
$ok .= $replace;
$pattern = "/[^".preg_quote($ok, "/")."]/";
return trim(preg_replace('/'.preg_quote($replace.$replace).'+/', $replace, preg_replace($pattern, $replace, $string)),$replace);
}
Could a Regex-Fu Master please grace me with a simpler/more efficient solution?
A much better solution suggested and explained by Botond Balázs and hakre:
function clean($string, $replace, $skip=""){
// Escape $skip
$escaped = preg_quote($replace.$skip, "/");
// Regex pattern
// Replace all consecutive occurrences of "Not OK"
// characters with the replacement
$pattern = '/[^A-Za-z0-9'.$escaped.']+/';
// Execute the regex
$result = preg_replace($pattern, $replace, $string);
// Trim and return the result
return trim($result, $replace);
}

Answer 1

I'm not a "regex ninja" but here's how I would do it.
function clean($string, $replace){
/// Remove all "not OK" characters from the beginning and the end:
$result = preg_replace('/^[^A-Za-z0-9]+/', '', $string);
$result = preg_replace('/[^A-Za-z0-9]+$/', '', $result);
// Replace all consecutive occurrences of "not OK"
// characters with the replacement:
$result = preg_replace('/[^A-Za-z0-9]+/', $replace, $result);
return $result;
}
I guess this could be simplified more but when dealing with regexes, clarity and readability is often more important than being clever or writing super-optimal code.
Let's see how it works:
/^[^A-Za-z0-9]+/:
^ matches the beginning of the string.
[^A-Za-z0-9] matches all non-alphanumeric characters
+ means "match one or more of the previous thing"
/[^A-Za-z0-9]+$/:
same thing as above, except $ matches the end of the string
/[^A-Za-z0-9]+/:
same thing as above, except it matches mid-string too
EDIT: OP is right that the first two can be replaced with a call to trim():
function clean($string, $replace){
// Replace all consecutive occurrences of "not OK"
// characters with the replacement:
$result = preg_replace('/[^A-Za-z0-9]+/', $replace, $result);
return trim($result, $replace);
}

Answer 2

I don't want to sound super-clever, but I would not call it regex-foo.
What you do is actually pretty much in the right direction because you use preg_quote, many others are not even aware of that function.
However probably at the wrong place. Wrong place because you quote for characters inside a character class and that has (similar but) different rules for quoting in a regex.
Additionally, regular expressions have been designed with a case like yours in mind. That is probably the part where you look for a wizard, let's see some options how to make your negative character class more compact (I keep the generation out to make this more visible):
[^0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz]
There are constructs like 0-9, A-Z and a-z that can represent exactly that. As you can see - is a special character inside a character class, it is not meant literal but as having some characters from-to:
[^0-9A-Za-z]
So that is already more compact and represents the same. There are also notations like \d and \w which might be handy in your case. But I take the first variant for a moment, because I think it's already pretty visible what it does.
The other part is the repetition. Let's see, there is + which means one or more. So you want to replace one or more of the non-matching characters. You use it by adding it at the end of the part that should match one or more times (and by default it's greedy, so if there are 5 characters, those 5 will be taken, not 4):
[^0-9A-Za-z]+
I hope this is helpful. Another step would be to also just drop the non-matching characters at the beginning and end, but it's early in the morning and I'm not that fluent with that.