Please use paths instead of array indices for nested schema objects

const test = zod
  .object({
    name: zod
      .object({
        firstName: zod.string().nonempty({ message: "First name can not be empty." }),
        middleName: zod.string(),
        lastName: zod.string().nonempty(),
      })
      .refine((name) => name.firstName === name.middleName, {
        message: "First and Middle Name must be the same", // refinement test.
      }),
  })
  .safeParse({
    name: {
      firstName: "",
      middleName: "estrada",
      lastName: "",
    },
  }).error.flatten();

console.log:

  fieldErrors: {
    name: (3) ["First Name can not be empty.", "Should be at least 1 characters", "First and Middle Name must be the same"]
  }

I have a component for each field and the lack of path prevents me from passing the error message to their respective components (ex: <FirstNameInput error={ {error.fieldErrors?.name?} }) />). We can always use object.keys to achieve the current behavior as well.

Another issue is the error message becomes very unclear since they don’t reference any paths in their message (ex: Should be at least 1 characters). The workaround here is to manually type the name inside the custom error message but that just isn’t elegant especially if we’ll be doing that per field.

[colinhacks]

@flybayer I don’t think your proposal is what the OP is asking for. @cybervaldez wants a full path array associated with each issue (correct me if I’m wrong). If you need that granularity of information, then you should use the raw ZodError.issues array.

The idea behind flatten is to easily collapse everything down to an object of depth one. I see now that this isn’t particularly useful. Your proposal is essentially the opposite of a flattening 😛 It takes the .issues array and converts it into a nested keyed object. Seems like a really useful utility, but it should be a separate method. Maybe treeify or denormalize — any thoughts?

BTW you’re using the .formErrors getter, which just delegates to .flatten(). I removed .formErrors from the documentation because .flatten() is more explicit.

[cybervaldez]

@flybayer I don’t think your proposal is what the OP is asking for. @cybervaldez wants a full path array associated with each issue (correct me if I’m wrong). If you need that granularity of information, then you should use the raw ZodError.issues array.

The idea behind flatten is to easily collapse everything down to an object of depth one. I see now that this isn’t particularly useful. Your proposal is essentially the opposite of a flattening 😛 It takes the .issues array and converts it into a nested keyed object. Seems like a really useful utility, but it should be a separate method. Maybe treeify or denormalize — any thoughts?

BTW you’re using the .formErrors getter, which just delegates to .flatten(). I removed .formErrors from the documentation because .flatten() is more explicit.

I was referring to the index and @flybayer 's solution is exactly what I’m looking for since it tells us what fields is causing the error. Without that the error becomes very unclear forcing us to set a custom message for each. Just look at the error produced by the middleName which doesn’t have any custom message: “Should be at least 1 characters”"