I'm curious -- why is this desirable? It breaks encapsulation, so if a more complex scheme were ever used to validate URLs (as in #18) there would be backwards compatibility issues.

@davisjam It may not be widely desirable. Although, i'm not quite sure i'm on the same page on compatibility, i'd have thought the situation would be the same given a regex change; it'd be a semvar major. My use case was, I was looking for a well tested regex packaged into a tidy module; but i don't want to directly call test on the regex. Instead, i wanted to use it with Joi's validation library.

const isUrl = require('is-url')

joi.object().keys({
  url: joi.string().regex(isUrl.matcher).required()
})

If the regex is not exported, then you could make a non-breaking change in how the URL is tested without changing the semver major.

That's what I've attempted (I think successfully?) in #18.

If the regex is exported, then any change to how the URL test is performed would be semver-major.

@davisjam Ah i got you. I think you can treat semvar exactly the same exposed or not; in any case the it's the behaviour of the regex that's depended on right and not it's exact pattern since two patterns can be equivocal, but different, for instance. In any case, this PR is not too important to me, just thought it would be a nice addition. So, happy to close for now. If this ever gets traction, would be happy to see it reopened and make any amends.

In #18 multiple regexes are used, and in futures perhaps still wilder changes might be afoot!

@davisjam Fair enough :) Indeed, multiple regexes would make this untenable in any case.