Proposal: z.not(parser) or equivalent

[rossipedia]

Purpose:

Returns the inverse of the parser that is passed. For example:

// accepts anything _except_ a `string`.
z.not(z.string()) 

// accepts any string that _isn't_ an email
z.string().not(z.string().email());

Would this be useful? Is there a way of doing this in standard zod that I'm missing?

I know I can use refine(), but the ergonomics there are a bit awkward:

z.string().refine((val) => !z.string().email().safeParse(val).success);

[JGJP]

I think this could be useful. The typescript equivalent would be doing a T extends string ? never : T I imagine. I think your second example would look better as:

z.not(z.string().email())

There should be a smart way to write your own utility function that replicates the behavior of z.not in this case

[colinhacks]

It's a nice idea, but there's no way to represent these "NOT" types in TypeScript so it isn't likely. It's important that Zod's inferred types accurately reflect what will pass validation. The closest thing that I'd be willing to introduce is an .exclude() method to mirror Exclude.

Then you could do something like z.unknown().exclude(z.string()) and the inferred type would be Exclude<any, string> (which is just unknown again).

[JGJP]

@colinhacks I think the .exclude() that you suggest is in line with the proposal, but if the input is typed then it should be generic and operate on that type.

[Kumar06Lav]

I've submitted a PR to add support for type negation using z.not() and the .not() method in schema validation.