Using Zod Codec instead of transform.

[mwalkersigma]

So question,
I we are using Zod Schemas, and doing mapping based on input / output shape ( hence the transform in the documentation ), Wouldn't it make more sense to use codecs, since that is inherently a bi-directional spec ?

for example, date fields the official docs currently recommend,

z.union([z.string(), z.date()]).transform(val => typeof val === 'string' ? new Date(val) : val)

But Zod supports Implicitly specifying both types while maintaining JSON Schema compatibility.

const isoDatetimeToDate = z.codec(z.iso.datetime(), z.date(), {
  decode: (isoString) => new Date(isoString),
  encode: (date) => date.toISOString(),
});
 
isoDatetimeToDate.decode("2024-01-15T10:30:00.000Z");  // => Date object
isoDatetimeToDate.encode(new Date("2024-01-15")); 

This makes the types more easily re-used and, removes in my opinion some of the ambiguity from the first example

Zod Codecs

[KyleAMathews]

yeah sure, that'd be a better way of doing it. Happy to merge a PR with the examples change!

[mwalkersigma]

Well, I'd happily submit a PR changing the sample, but would my above code even work in current state?
Unless I am doing something wrong, the schema is happily accepted but as soon as I preform an update on the data, my live query that filters based on the date field stops working.

If I misunderstood, and you were suggesting that I implement the feature, and update the docs my apologies !

[KyleAMathews]

Oh what doesn't work? We only support the subset of features in standard schema. I don't know how codec works.

[mwalkersigma]

@KyleAMathews I was just now able to get it to work by doing this, again I should have read the docs better.

export const dateStringCodec = z.codec(
    z.union([z.string(), z.date()]), // Input Schema
    z.string(),                      // Output Schema
    {
        decode: (val) => (typeof val === 'string' ? val : val.toISOString()),
        encode: (val) => new Date(val).toISOString()
    }
);

Output must be a subset of input for tanstack DB

[mwalkersigma]

@KyleAMathews
#1080

I gave it a shot updating the docs, if it isn't up to Tanstack standards feel free to rewrite, or not use any of it !

[mwalkersigma]

Also Also, Fantastic job on the library!

[KazimirPodolski]

I highly support this. I'm currently using PowerSync + Kysely + Zod + my custom de-normalization and it works well-ish, but looking at TanStack DB as a client-side ORM replacing Kysely and custom stuff. However, there are many missing spots in DB which my previous setup nails, including IMO quite non-ergonomical schemas support.

@KyleAMathews please take a look at Zod's codecs feature, especially the .decode()/.encode() stuff. I believe DB can have much simpler process with schemas if those are utilized.