Regex & pattern matching
Enter the pattern body without / delimiters. Spaces and line breaks are preserved; an empty pattern is valid.
Use the pattern body, such as (cat), rather than /cat/g. Put g in the separate flags field. The tool does not trim your text; editor or clipboard handling can normalize line endings, so it is not a byte-preserving file test.
Ready for input or a manual test.
All matching runs locally in a browser Worker; no pattern or test text is uploaded.
Use supported flags once each. u and v cannot be combined. Without g, match, replace and extract stop at the first match.
d — Include capture indices.g — Find all matches; split is independent of this flag.i — Ignore case using JavaScript rules.m — Let ^ and $ match line boundaries.s — Let . include line terminators.u — Unicode-aware syntax and code-point matching.v — Unicode sets mode; cannot be combined with u.y — Sticky: match at lastIndex, starting at 0 for each run.Browser identity: Detecting browser…. This is the reported user agent, not an exact engine version.
Limits in UTF-16 units: text 100000, pattern 2000, replacement 10000, replacement output 1000000, full report 4000000. At most 10000 result rows; exceeding a limit fails the whole run. Timeout: 2 seconds.
Ready for input or a manual test.
Only a current replacement with valid Unicode and at most 100000 UTF-16 units can become input. Preview and apply first; undo restores the previous input, options and result.
A passing validation can hide a wrongly captured field. The Regex tester places the matched text, its numbered captures and named groups together so you can check exactly which value a later replacement will use.
Use this regular expression tester as a regex debugger for a small, realistic sample before changing application code. The regex capture groups and regex replace previews expose differences between a complete match and the fragment you intend to keep.
Keep syntax and output aligned
The Regex tester shows a complete match around its captured parts. Inspect names, numbers and UTF-16 positions instead of assuming that every parenthesis returns the field you wanted.
Replacing the first occurrence and replacing every occurrence are different changes. Toggle g to compare them on the same input; the original text changes only after a separate preview and apply action.
Use the pattern body, such as (cat), rather than /cat/g. Put g in the separate flags field. The tool does not trim your text; editor or clipboard handling can normalize line endings, so it is not a byte-preserving file test.
Test cat against cat cat and switch g off: the match list falls from two entries to one. Automatic mode waits 350 ms after edits; turn it off to use Ctrl/Cmd+Enter. split does not follow the first-match rule.
Compare null with "" in group details before exporting. The former means a group did not participate; the latter is a captured empty string. The list grows by 50, but the JSON report contains every accepted result.
For (Ada) (Lovelace), the full match is Ada Lovelace while groups 1 and 2 hold separate fields. With d, their ranges are included; a lookbehind capture can sit outside the full-match range.
With (?<year>\d{4})-(?<month>\d{2}), $<month>/$<year> turns 2026-09 into 09/2026. Compare g with the first-only behavior before applying the preview to your input.
For ID:(\d+), select group 1 to extract 42 from ID:42; select 0 to keep ID:42. A nonexistent selection fails when a match is present. With no matches, there is no group-existence check.
Splitting a,b with (,) yields ["a", ",", "b"]. Captured separators are retained. Alternatives such as (,)|(;), including their unmatched captures, can add null entries; empty strings are not filtered out.
Enter a pattern body and test text. The Regex tester starts in match mode with g enabled, then runs 350 ms after edits. Disable automatic mode to use the test button or Ctrl/Cmd+Enter; even an empty pattern can be tested.
g searches all occurrences for match, replace and extract; without it, only the first is used. m changes ^ and $ to line anchors, while s lets the dot include line breaks. split processes the whole input independently of g.
A greedy quantifier first tries more text; a lazy form such as +? tries less while allowing the rest of the pattern to succeed. Compare both on realistic samples: the limited token notes do not prove what a complex expression means.
Put the desired text in parentheses, then reference it with $1 or $<name> in replacement mode. Extraction also accepts 0 for the full match or a group number/name. Groups that did not participate remain distinct from empty captures.
Backtracking can keep an apparently short expression busy. Every operation runs in a separate Worker that is terminated after 2 seconds; a stopped or timed-out run cannot be exported. A successful test does not certify performance on other inputs.
Still have a question?
Try the Regex tester with a date or Unicode sample, then adjust the pattern to your own text.
Check the full JSON report before taking a replacement back into your code.
Edits run after 350 ms by default. Disable automatic mode for the button or Ctrl/Cmd+Enter. Each change immediately disables stale exports.