Random Data from a Regex

They constrain a match against its surroundings rather than describing characters. A lookahead says what must follow, and a backreference says a later part must equal an earlier one — neither adds anything to build from, so they are reported instead of being quietly dropped.

A star or plus has no upper limit, so a random repeat count in a small range is used. That keeps the output a usable length instead of occasionally producing something enormous.

Every batch is tested against the real regular expression engine, and any mismatch is reported. That check is what catches the cases where a pattern uses a construct the generator cannot fully honour.

Some patterns describe only a handful of strings — a pattern like (yes|no) has exactly two. When the pattern cannot express as many distinct values as you asked for, the shortfall is reported rather than the tool looping.

For anything with a fixed shape, yes: reference numbers, product codes, postcodes, version strings. For names, addresses and emails that need to look plausible rather than merely well-shaped, the sample data generator is the better fit.