Regex Escape — Escape Special Characters for a Literal Match

Pick the flavor of the engine that will compile the pattern, paste the literal text you want matched exactly, and press Escape. Drop the result into a larger pattern or a character class — it is a pattern fragment, with no surrounding delimiters added.

For the common JavaScript case: a dot, parentheses ( ), brackets [ ], braces { }, a backslash, and a forward slash all get a backslash prefix under the classic flavor — . becomes \., ( becomes \(, and so on — because each one means something to the engine outside a character class. A literal dash - only needs escaping inside a character class, and only when it isn't first or last (a-z is a range, but [-az] or [az-] is a literal dash); outside a class, a bare - is never special and this tool leaves it alone. The full table below covers all five flavors side by side, since PHP and Python draw slightly different lines.

Because the ES2025 RegExp.escape function is designed to be safe to CONCATENATE, not just safe to compile on its own — if the escaped text is spliced directly after something like \p{...} or a backreference number, an unescaped leading letter or digit could be read as part of that adjacent token instead of as the start of your literal. Hex-escaping the first character when it's an ASCII letter or digit removes that ambiguity entirely, at the cost of one two-character escape you'll only ever notice at the very start of the output. The classic escaping style most existing JavaScript code uses doesn't do this, which is why the two flavors can disagree on the very first character of otherwise-identical output.

()[]{}?*+-|^$\.&~# and the five whitespace characters space, \t, \n, \r, \v, \f. Before Python 3.7, re.escape backslash-prefixed every non-alphanumeric character, which produced needlessly noisy patterns for ordinary punctuation like a comma or an apostrophe; 3.7 trimmed the set down to exactly the characters that are actually special in Python's re syntax (plus whitespace, since re.VERBOSE mode treats unescaped whitespace as insignificant). Letters, digits and underscore were never escaped in either version.

\Q begins a literal run and \E ends it — every character in between, including ones that would otherwise be metacharacters, is matched exactly as written, with no per-character escaping needed. Java's Pattern.quote(s) is literally implemented as \Q + s + \E, and PCRE (used by PHP's preg_match_all, when you write the delimiters yourself, and many other engines) supports the same \Q...\E pair inside a pattern you write by hand. The one trap: if the literal text itself contains \E, it would prematurely end the quoted run, so a correct implementation (this tool included) splices it as \E\\E\Q — close the quote, match a literal \E the hard way, reopen the quote — rather than searching-and-replacing it away.

Yes, and this tool escapes for use OUTSIDE a character class (the far more common case — dropping literal text into the body of a pattern). Inside [...], the rules shrink: ] needs escaping (or must come first), a literal backslash still needs escaping, ^ needs escaping only if it's the very first character (otherwise it means negation), and - needs escaping unless it's first or last (otherwise it defines a range like a-z). Everything else that's special outside a class — . ( ) { } | ? * + — is just a literal character inside one and needs no escaping at all. If you're building a character class from a literal set of characters, escape only ] \ ^ - by hand rather than pasting this tool's output verbatim.