How to escape special characters that that are not gathered in a special shape

Question

I have this string:

$var = "Foo (.* ) bar blla (.* ) b*l?a$ bla bla ";

I want to escape the * and ? and all special characters that are not gathered in this shape

"(.*)"

I wanted to use preg_quote($var, '\') But it escapes all the special characters, and I only need the single special characters to be escaped. I want this result:

$var = "Foo (.* ) bar bla(.*) b\*l\?a\$ bla bla ";

I want to use the final $var (the result) in a preg_match that matches all (.*) in an other string, and the special characters which are in my case theses :

., \, ^, $, |, ?, *, +, (, ), [, ], {, }, and /

should be parsed as a normal text so they should be escaped. while the (.*) one shouldn't be escaped. Only the special characters above should be escaped, because I will have to use $var in preg_match. The other special characters, no need to escape them.

preg_match("/" . $var . "/s", $anotherstring, $match);

Answer 1

Edit3

It appears as if it didn't work for you, so here's another attempt. And since mickmack seems to be worried about performance, he'll be glad that it's down to 146 steps ;)

Replace

([\w\s]*(?:\([^)]*\)[\w\s]*)*)([*?$&])

with

$1\\$2

Here at regex101.

It captures an optional range of non special characters. It goes on capturing an optional parenthesized group, followed by an optional range of non special characters. This last part can repeat any number of times. Finally it captures the special character.

So we have to capture groups - one with the text leading up to the special character (if any), and one with the special character.

Replacing this with the content of them with a \ in between, does the trick.

This is also more flexible with the parentheses part (happy mick? ;). It allows more complex regex'es inside the brackets (just not nested parentheses).

If the new requirement of handling \'s isn't a must, and a negated word class is OK \W, we're down to a blazing 76 steps :) Here at regex101.

--Original answer--

This is one way of doing it - replace

(?<!\(|\(.|\(..)([^\w\s])(?![^(]*\))

with

\$1

Note! You have to escape the \in the php string - i.e. "\\$1".

Since php only allows fixed with look-behinds, it tests that there isn't an opening parentheses before the special character in four steps with the (?<!\(|\(.|\(..|\(...)construct. Then it matches, and captures, the special character (not a word character, nor a space). Lastly it uses a negative look-ahead to make sure it isn't followed by a closing parentheses. Checking the parentheses both before and after may be redundant though.

Replacing the matched, and captured, character by itself - $1 - preceded by the wanted escape character \ will do the trick.

See it here at regex101.

Edit

Here's an alternative way if the special characters are limited to the one in your example - use

(?<!\(\.)([*?$&])(?!\))

as the search string and replace with \$1.

It matches your special characters as long as they're not preceded by (., nor followed by ).

Here at regex101.

(Neither of the ways are waterproof since they would fail to escape the & in (.& ).)

Edit2

Updated since OP changed escape character in question from / to \. And removed the space inside the capturing group as it was not wanted by OP.

Answer 2

Here are a few patterns that outperform ClasG's answer:

Input: Foo (.* ) bar blla (.* ) b*l?a$ && bla bla

Pattern: /\([^)]*\)(*SKIP)(*FAIL)|([^a-z\d ])/i Replace with: \\\1

Output: Foo (.* ) bar blla (.* ) b\*l\?a\$ \&\& bla bla

Pattern Demo (just 122 steps)

Basically it just omits the "protected" parenthetical portion and matches any non-alphebetic & non-space characters.

---

If you want to specifically list the symbols, you can just change the negated character class to the character class in the OP like this: (still 122 steps)

/\([^)]*\)(*SKIP)(*FAIL)|([-\/~`!@#$%^&*()_+={}[\]|;:'"<>,.?\\])/

---

or you can use only the symbols in your sample, here's the full pattern (still 122 steps):

/\([^)]*\)(*SKIP)(*FAIL)|([*?$&])/

---

All of ClasG's patterns are slower than my 3 patterns above:

ClasG's written pattern: (?<!\(|\(.|\(..)([^\w\s])(?![^(]*\)) fails and takes 418 steps - demo

ClasG's linked demo pattern: (?<!\(|\(.|\(..)([^\w\s])(?![^(]*\)) is correct but takes 367 steps - demo

ClasG's third pattern: (?<!\(\.)([*?$&])(?!\)) is correct but has a strict requirement for the parenthetical portion. It is the best pattern in that answer taking 186 steps - demo.

Answer 3

use preg_replace_callback, you can look regexp https://regex101.com/r/52qQwv/1

$s = 'Foo (.*) bar blla (.*) b*l?a$&& bla bla';
$regexp = '/([\.\*\?\&\$])[\w\s\&]/iu';
$f = function ($matches) {
    return '/' . $matches[1];
};
$a = preg_replace_callback($regexp, $f, $s);
var_dump($a);

string(39) "Foo (.) bar blla (.) b/*/?/$/&bla bla"