Regex generator from examples

The candidate table is finite, and the tool only does one thing: it verifies which patterns really satisfy your positives and negatives. When none do, it says so plainly instead of producing a regex that looks right but matches the wrong things. To check a pattern you wrote yourself, the regex tester is the better tool.

No — the built-in candidates use search semantics, so \d+ matches the 123 inside abc123. For whole-string validation (say “6–12 lowercase letters or digits”) use the custom character classes block: it emits an anchored pattern such as ^[a-z0-9]{6,12}$.

Because they all satisfy your constraints — they just differ in strictness. The email pattern is far more specific than the identifier pattern, so it is ranked first. Add a few negative samples to weed out the loose ones; that is usually more effective than adding more positives.

When no built-in pattern covers “fixed length plus a specific character set”, pick the sets (upper, lower, digits, underscore, Chinese, whitespace, hyphen) and the length range; the resulting anchored pattern joins the candidate list. The hyphen is always placed last so a class like [a-z-0-9] never turns into an accidental range.

At most 100 samples, each up to 200 characters; blank lines are ignored. Going over the limit is reported rather than silently truncated, because dropping a few samples can completely change which pattern fits.