Understand Zero-length Regex Matches

Understand zero-length matches and why engines list them differently.

Checking engine…

Not saved: closing this tab loses your work. Save keeps the regex, flags, engine, texts and tests in this browser only.
Workspace actions
Regular expression
Reference

Regex quick reference

^ · $
Start/end; m = lines.
[abc] · \d
Listed character / digit.
* · + · ?
Zero or more, one or more, optional.
(...)
Capture a numbered group.
(?<name>...)
Named capture.
(?=...)
Ahead; lookbehind varies.
i · m · s
Ignore case / line anchors / dot-newline.
Browse the full reference

Tab inserts a tab · press Esc, then Tab, to leave the editor

No regex yet.

⋯

the RegExp of this browser (version not pinned)

Not available in JavaScript · This browser: x · ignore pattern whitespace; a · ASCII-only classes; U · lazy by default; J · duplicate names. u and v cannot be combined.

To run: enter a regex; enter a test text.

Test text

Tab inserts a tab · press Esc, then Tab, to leave the editor

No text yet.

No result yet

Results appear here after a run.

Verified Fixture FX-G04

What it does

A zero-length match is a real match of empty text, such as a* where no a follows. Engines list them differently.

How to use it

  1. Load the example: a* on "ba".
  2. Compare the empty matches in each engine.
  3. Use + instead of * if you do not want empty matches.

Verified example

As written

Text "ba"

JavaScript · This browser
a*

Result: "", "a", ""

Python re · Local WASM
a*

Result: "", "a", ""

PCRE2-8 · Local WASM
a*

Result: "", "a", ""

Go regexp · Local WASM
a*

Result: "", "a"

Recorded by running fixture FX-G04 on : rp.js@1 on Node v24.21.0 (V8), v24.21.0, CPython 3.14.2 (Pyodide 314.0.7), PCRE2 10.48, go1.27.1. JavaScript results come from the rp.js@1 core on Node’s V8; your browser runs the same core.

Engines

Limits