UK Postcode Regex Tester

The Pattern

``regex ^[A-Z]{1,2}\d[A-Z\d]? ?\d[A-Z]{2}$ `

Flags: gmi — multiline, so ^ and $ anchor to each line rather than the whole input.

What It Accepts and Rejects

Outward and Inward

A UK postcode has two halves. The outward code is one or two letters, a digit, and an optional further letter or digit; the inward code is always a digit followed by two letters. The space between them is conventional and frequently missing in stored data, so the pattern makes it optional with ? .

The i flag accepts lowercase input, which is how postcodes arrive from forms. Normalise to uppercase with a single space before storing, so that sw1a1aa and SW1A 1AA compare equal.

The Limits of This Pattern

This is a shape check, not a validity check. ZZ9 9ZZ matches the pattern and is not a real postcode. The full specification also excludes certain letters in certain positions — the inward code never uses C, I, K, M, O or V — and there are genuine special cases like GIR 0AA for Girobank.

For anything that affects delivery or eligibility, check against Royal Mail's PAF or the free Ordnance Survey open data rather than trusting the format.

Testing Before Shipping

A regex that has only been tried against inputs you expect to match is untested. Every pattern needs three kinds of case:

1. Valid inputs that should match, including the awkward-but-legal ones. 2. Invalid inputs that should not, especially near-misses that differ by one character. 3. Adversarial inputs — very long strings, unusual Unicode, and nesting that could trigger catastrophic backtracking.

Paste your own examples into the tester above and watch which lines highlight. A pattern that matches everything you throw at it is usually too permissive rather than correct.

Catastrophic Backtracking

Nested quantifiers over overlapping character classes — (a+)+, (\w+\s?)* — can take exponential time on a non-matching input. On a server that is a denial-of-service bug, not a performance issue. If a pattern is applied to user input, bound the input length first and prefer explicit alternation over nested repetition.

Anchors, Greediness and Backtracking

Three behaviours account for most regex surprises, and this pattern shows all three.

Anchors. ^ and $ pin the match to the start and end of the input. Without them, \d{3} matches the 123 inside abc123def. With m in the flags — as here — they pin to each *line* instead, which is what lets one pattern be tested against a list.

Greediness. .* takes as much as it can and gives back only when forced; .*? takes as little as possible. On , the pattern <.*> matches the whole string and <.*?> matches just .

Backtracking. When a match fails, the engine reverses and tries other splits. Nested quantifiers like (a+)+ make that exponential, and a 30-character input can hang a server — a class of denial of service known as ReDoS. Avoid nesting quantifiers, and prefer explicit character classes over . wherever you can.

Testing It Properly

`javascript const pattern = /^[A-Z]{1,2}\d[A-Z\d]? ?\d[A-Z]{2}$/gmi;

// A global regex keeps lastIndex between calls, so reusing one across // test() calls returns alternating true/false on the same input. pattern.lastIndex = 0;

// Named groups make the result readable const named = /(?\d{4})-(?\d{2})/; const { groups } = '2026-08'.match(named); `

Write the failing cases first. A pattern that accepts everything valid is easy; one that also rejects everything invalid is the hard half, and it is where the bugs are.

When Not to Use a Regex

Structured formats have parsers, and the parser is always more correct: new URL() for URLs, DOMParser for HTML, JSON.parse` for JSON, a date library for dates. Reach for a regex to *find* things in unstructured text, not to validate something a parser understands.