New major

[blakeembrey]

Plan for version 3:

• Remove variadic argument behavior
• Provide native TypeScript types
• Investigate expansion of languages into multiple types to check
• Support parameter matching as part of the API (e.g. ; version=1)
• Publish as ESM
• Remove mime dependency for a smaller bundle

[blakeembrey]

A couple of other thoughts on the API:

1. Make it a TypeIs class so we can optimize normalize to only run once (most applications want to match the same thing over and over).
2. Expose a contentType method that accepts the parsed Content-Type header from content-type directly (downstream packages may be able to optimize the times they parse)

[blakeembrey]

We also need to change the "match" to be the match type, currently it's either the content-type or the match which is pretty confusing. Even the express.js documentation for this is wrong: https://expressjs.com/en/5x/api.html#req.is

Based on the behavior here:

type-is/index.js

Lines 72 to 73 in 2554c0a

	return type[0] === '+' || type.indexOf('*') !== -1
	? val

and tests here:

type-is/test/test.js

Line 85 in 2554c0a

assert.strictEqual(typeis(req, ['text/*', 'image/*']), 'image/png')

, it's actually return the MIME type. This makes it difficult to use a switch based behavior with the wildcards. An expected workflow would be something like:

const renderers = {
  'text/html': () => '<>',
  'text/*': () => '...',
};

const typeIs = new TypeIs(Object.keys(renderers));

const handler = (req, res, next) => {
  const renderer = renderers[typeIs.request(req)];
  if (renderer) return res.send(renderer);
  return next();
}

Everything that could be done in the current API would still be available, but the optimized path here would be to allow that switch (instead of multiple if (is(()) checks) and normalize the input types once allowing for the faster match against the input content type.

[blakeembrey]

May also want to consider removing or switching mime-types to mime. Mime types requires path, which means we can't support browsers (#45) but we also want getAllExtensions included in mime a getAllTypes, not extensions (not included in mime).

Alternatively, I think we could just drop the mime behavior in favor of some explicit hardcoded shortcuts.