The latest updates on your projects. Learn more about Vercel for Git ↗︎

Open in Stackblitz

npm i https://pkg.pr.new/valibot@1013

commit: d42d890

Seems mostly good, but optional defaults are broken:

> v.parse(v.object({ a: v.optional(v.string(), "foo") }), {})
{}

(expected { a: 'foo' })

I think this is fixed now. I also added support for the nullish schema. This PR increases the bundle size of object by 100 bytes. If those 100 bytes fix all our problems, it should be fine.

The new behavior (these are breaking changes) of this PR is as follows:

• optional no longer allows undefined as a value, it only allows missing object entries
• Users must explicitly allow undefined by using undefinedable or union inside optional
• object now distinguishes between missing and undefined values and only allows missing entries if the outer schema of an entry is optional or nullish
• Any other schema will return an issue for missing object keys

Example:

const Schema = v.object({
  // Disallow missing keys
  key0: v.number(),                               // number
  key1: v.undefinedable(v.number()),              // number | undefined
  key2: v.nullable(v.number()),                   // number | null
  key3: v.unknown(),                              // unknown

  // Allow missing keys
  key4: v.optional(v.number()),                   // number
  key5: v.optional(v.undefinedable(v.number())),  // number | undefined
  key6: v.nullish(v.number()),                    // number | undefined | null
  key7: v.optional(v.unknown())                   // unknown
});

If we think this is the right solution for Valibot, I would also update these changes in looseObject, strictObject and objectWithRest. I would also update the tests to match the new behavior.

I have implemented your feedback and removed support for missing keys when using the nullish schema. The idea behind this change was to make our API more consistent. The only way to make an object entry optional now is to explicitly use optional as the outermost schema of the entry.

Furthermore, I refactored the implementation of the object schema and optimized its bundle size. These changes lowered the addition required bytes from about 100 to 77.

I think this PR is going in the right direction. My next steps are to apply these changes to the entire codebase and update our tests. Once this PR is done, I will probably create a GitHub discussion to explain these changes and reach out to the community and ask for feedback.

Please discuss with use in #1022.

I followed @xcfox's idea: #1022 (comment)

Valibot v1.0.0-beta.13 is available