Prompts / Regex from examples
Regex from examples
Build a regex from concrete match/non-match examples so the pattern is tested against your real edge cases, not just the obvious case.
Copying runs entirely in your browser - nothing here is ever sent anywhere.
Fill in the variables
I need a regular expression. Here's what it should match:
{{should_match}}
Here's what it should NOT match (these are the tricky edge cases - pay
close attention to these):
{{should_not_match}}
Context on where this will run (language/regex flavor, e.g. PCRE,
JavaScript, Python `re`): {{context}}
Give me the pattern, then walk through it piece by piece explaining what
each part does. Then explicitly check the pattern against every example
above (both the should-match and should-not-match lists) one by one, and
flag any example it doesn't handle correctly instead of silently leaving
it broken.
When to use
Anytime you’re tempted to write a regex from memory and hope it’s right - which is most of the time. Especially valuable for the edge cases (a trailing dot, an escaped character, an empty string) that are easy to forget when writing a pattern by hand.
Why it works
Regex mistakes are almost always in the edge cases, not the obvious case - a pattern that matches the “should match” example on the first try often still fails on a should-not-match case nobody thought to test. Giving both lists up front, and asking the model to explicitly check the pattern against every one of them, catches that instead of producing a pattern that merely looks right.
Variations
- Ask for the pattern with named capture groups if you need to extract parts of the match, not just detect it.
- Add “Also give me 3 more tricky edge cases I might have missed” before finalizing - self-generated adversarial examples often catch real gaps.
- For a catastrophic-backtracking concern (a pattern that could hang on pathological input), ask explicitly whether the proposed pattern is safe against that, and why.