Regex Generator from Sample

How the pattern is built

The tool reads the sample from left to right. Whenever it meets a run of the same kind of character, it emits a character class sized to that run:

• a run of digits becomes \d{n}, where n is how many digits there are
• a run of uppercase A-Z becomes [A-Z]{n}
• a run of lowercase a-z becomes [a-z]{n}
• a single space becomes \s, and several spaces become \s{n}
• anything else (punctuation, symbols, accented or non-Latin letters) is kept as an escaped literal character

The finished pattern is wrapped in ^ and $ so it matches the whole string from start to end, not just a fragment.

Example

For the sample Order #12345 placed on 2026-04-16 the tool produces:

/^[A-Z]{1}[a-z]{4}\s\#\d{5}\s[a-z]{6}\s[a-z]{2}\s\d{4}\-\d{2}\-\d{2}$/

Order is one uppercase letter followed by four lowercase letters; 12345 becomes \d{5}; and the date 2026-04-16 becomes \d{4}\-\d{2}\-\d{2}.

What it does and does not do

• The lengths are fixed. 12345 becomes \d{5}, which matches exactly five digits, not “one or more”. If the count varies in your data, change {5} to + or a range such as {3,8} by hand.
• It works from a single sample. It mirrors the structure of the one string you give it, so it cannot generalise across several examples on its own. Feed it the most representative string.
• Only A-Z and a-z count as letters. Accented and non-Latin characters are kept as literals, so replace them with a class such as \w or \p{L} if you want them to match a wider set.
• A regex cannot express everything. Things like balanced brackets are beyond any regex, no matter the sample. For nested structures you need a parser, not a pattern.

A good workflow

1. Generate a pattern from one clear, representative sample.
2. Relax the fixed counts ({5} becomes + or {3,8}) wherever the length varies.
3. Paste the pattern into the Regex Tester with a larger set of real strings.
4. Adjust until it matches everything it should and nothing it should not.