Regex Tester

Matching runs in a Web Worker with a deadline on it. That is not a performance decision — it is the only way this page can survive a pattern that goes exponential.

Why a timeout needs a separate thread

JavaScript regular-expression matching is synchronous and cannot be interrupted. There is no timeout argument, no abort signal, and no way to ask the engine how far along it is. A pattern such as (a+)+$ tested against a run of as ending in one character that spoils the match forces the engine to try every way of dividing that run between the two quantifiers — a number that doubles with each character added. On the main thread, the tab stops responding and stays that way.

So the work happens on a Worker, the page starts a clock, and if the Worker has not answered in time it is terminated and a fresh one takes its place. The pattern is reported as having exceeded the deadline rather than being left to finish.

The pattern review is advisory

Before running anything, the pattern is scanned for shapes known to backtrack badly: unbounded repetition nested inside unbounded repetition, a repeating group whose branches can match the same text, very large bounded repetition, and backreferences. Each is reported with the reason it matters.

A clean review means none of those shapes are present. It does not mean the pattern is fast — the general question is not one a scanner settles, and a backreference takes the pattern outside the class of expressions that can be matched in linear time at all. The deadline is the guarantee; the review is a warning you get earlier.

Flags behave as they will in your code

Without g, one match is returned, because that is what your own call will do. Capture groups are listed in source order, and a group that did not participate is shown as unset rather than empty — the difference matters when you branch on it.

The second action button loads (a+)+$ against a string built to defeat it. It is there to be run: the deadline fires, the Worker is replaced, and the page stays interactive throughout. That behaviour is the point of the tool, and it is easier to trust having watched it happen than having read that it would.