.strict() and .and() combined is not working

[fabien0102]

Bug description

The method .strict() and .and() combined results in an unexpected parsing error.

Zod version

3.0.0-alpha.33

Playground

const a = z.object({ a: z.string() }).strict();
const b = z.object({ b: z.string() }).strict();
const c = a.and(b);

c.parse({ a: "a", b: "b" });

Output

error: ZodError: [
    {
      "code": "unrecognized_keys",
      "keys": [
        "b"
      ],
      "path": [],
      "message": "Unrecognized key(s) in object: 'b'"
    },
    {
      "code": "unrecognized_keys",
      "keys": [
        "a"
      ],
      "path": [],
      "message": "Unrecognized key(s) in object: 'a'"
    }
  ]

[colinhacks]

This is expected. The .and method creates a "dumb" intersection. It passes the input into both A and B. If they both pass, validation is considered successful. Because A and B both individually don't allow unrecognized keys, they both fail. Get rid of the .strict() calls if you want this to work.

If you're just looking to merge two objects, use .merge which more like you expect. It also returns a new ZodObject instance instead of a ZodIntersection.

const a = z.object({ a: z.string() });
const b = z.object({ b: z.string() });
const c = a.merge(b).strict();

Small side-note, when doing A.merge(B), the resulting object inherits its strictness setting from A.

[fabien0102]

Thanks for the explanations.

Sadly, this .merge doesn't really solved my issue, and I think I found another issue 🙄

const a = z.object({ a: z.string(), b: z.number() });
const b = z.object({ b: z.string() });
const c = a.merge(b).strict(); // { a: string; b: never }

// Passed
c.parse({ a: "test", b: "42" });

// Failed
c.parse({ a: "test", b: 42 });

The z.infer<typeof c>["b"] is never instead of string (regarding the actual validation behaviour)

Anyway, I can't use this merge to translate a typescript intersection in the context of ts-to-zod.

[colinhacks]

Good catch! I relatively recently changed the behavior of "merge" to behave more like "extend". I forgot to update the inferred type. Fixed in alpha.38.

Like you say, merge isn't an intersection, it just merges the sets of properties from A and B into a new object. In the case of a clash, the properties of B override the properties of A. I guess there isn't a way do what you're trying to do. You'll have to get rid of the .strict() calls if you want the intersection to work properly. In fact, part of the reason the default behavior was changed from "strict" to "strip" was so intersections would work on "plain" ZodObjects.

Is there a reason you need strict checking? I'd recommend against having ts-to-zod generate strict ZodObjects by default, but it's totally your call.

[colinhacks]

By the way, I just added an Ecosystem section to the README to boost projects like yours: https://github.com/colinhacks/zod/tree/v3#ecosystem

[fabien0102]

Is there a reason you need strict checking?

No reason at all, was just to be "feature complete", but I will probably drop this idea (and my plan was to put this an option, not the default)

[colinhacks]

Gotcha. Loving your work on this so far! Sorry it took so long to respond here, just started a new job. Should be more responsive in the future. Feel free to DM me on Twitter @colinhacks as well for a speedier response.

[fabien0102]

Don't worry, I also have a lot of work and I totally understand than side-projects are not always a priority (and can be sometime hard to manage on top of all other stuff going on 😅 )