README

@nrk/nrkno-sanity-typesafe-schemas

This library underpins the principles of nrkno-sanity and option driven design.

Sanity Typesafe Schemas

Utility functions to ease Sanity schema development:

schema(typeName, schemaDefinition): used to define schemas that is consumed by createSchema.
field(typeName, fieldDefinition): used to define fields for document, object, file and image schemas.
arrayOf(typeName, arrayMemberDefinition): used to define of-entries in array schemas.
typed<T>(anything): passthrough function to provide inline types in json objects.
checkSchema<T>(anything): safeguard function that ensures an interface contains all field-names used in a schema
forType<T>(anything): utility function meant to be used with checkSchema

API design goals

Typesafe Sanity schemas (and with that, autocompletion)
Feature discovery of built-in Sanity features (types, options, validations)
Breadcrumb-trails to relevant Sanity docs
Feature discovery & documentation of extension types & features (custom input-resolver options, reusable schema-types etc.)
Decrease time from idea to finished schema for developers, by embracing the above.
Opt-in. Ie, this can be sprinkled into an existing codebase to gradually add type safety to schemas.
Minimal runtime footprint. Should not mess with the schema provided by the developer.

See the-gist-of-it-all.ts for a scaled down version of how the helper functions work.

Type discovery

Feature discovery

Typesafety

Documentation

Installation

npm install --save @nrk/nrkno-sanity-typesafe-schemas

yarn add @nrk/nrkno-sanity-typesafe-schemas

Usage

import { field, schema } from '@nrk/nrkno-sanity-typesafe-schemas';

export const typesafeDocumentSchema = schema('document', {
  name: 'some-doc',
  title: 'Some document',
  fields: [
    field('string', {
      name: 'someField',
      title: 'Some title',
      initialValue: 'a',
      options: {
        list: [
          { value: 'a', title: 'A' },
          { value: 'b', title: 'B' },
        ],
        layout: 'radio',
      },
    }),
  ],
});

//@ts-expect-error schema is not correctly defined
const withErrorSchema = schema('string', {});

See code examples for more.

Extending Sanity schemas

Use TypeScript declaration merging to extend the types under @nrk/nrkno-sanity-typesafe-schemas:

// document-extension.ts
declare module '@nrk/nrkno-sanity-typesafe-schemas' {
  // adds field 'document' schemas
  interface DocumentSchema {
    custom?: boolean;
  }
}

// number-extension.ts
import '@nrk/nrkno-sanity-typesafe-schemas';
declare module '@nrk/nrkno-sanity-typesafe-schemas' {
    // adds the custom option to 'number' schemas (Number already has options, so we extend the NumberOption)
    interface NumberOptions {
       custom?: boolean;
    }
}

// somewhere using typesafe helpers
export const typesafeDocumentSchema = schema('document', {
    /* omitted */
   custom: true, // custom is now a valid option for all document schemas
});

export const typesafeNumberSchema = schema('document', {
    /* omitted */
    options: {
        custom: true, // custom is now a valid option for all number schemas
    },
});

A list of all extendable types can be found in index.ts.

Adding reusable schema types

Define your schema type, and use TypeScript declaration merging to extend SchemaDirectory in @snorreeb/sanity-typesafe-schemas:

// special-string-schema.ts
import { StringSchema } from '@nrk/nrkno-sanity-typesafe-schemas';
export type SpecialStringSchema = Omit<StringSchema, 'type'> & {
  type: 'special-string';
  options: {
    special: string;
  };
};

// schema-directory-extension.ts
import '@nrk/nrkno-sanity-typesafe-schemas';
declare module '@nrk/nrkno-sanity-typesafe-schemas' {
  interface SchemaDirectory {
    'special-string': SpecialStringSchema;
  }
}

// somewhere using typesafe helpers
export const specialStringSchema = schema(
  'special-string', // special-string is now a valid, autocompletable type
  {
    /* omitted */
  }
);

Escape hatches

Defining ad-hoc types without extending anything

The '' string is a general escape hatch that still provides some safety. undefined and nullwill also work.

Use alongside the typed generic helper function, to define inline types.

import { schema, typed } from '@snorreeb/sanity-typesafe-schemas';

export const stringSchemaUsedByNameSomewhereElse = schema('', {
  type: 'string', // when using '', type must be provided here
  name: 'this-name-is-used-in-another-schema',
  title: 'Custom stringtype for whatever reason',
  inputComponent: () => 'Render me softly',
  options: typed<{ customOption: boolean }>({
    customOption: true,
  }),
});

// ...

export const schemaUsingTheAbove = schema('', {
  type: stringSchemaUsedByNameSomewhereElse.name,
  title: 'New title',
  name: 'type-pointing-to-another-type',
});

Force `undefined` title

Title is optional in Sanity Studio, but will result in console.log warnings if omitted. Therefore, it is required by the schema-helpers.

To set an undefined title anyway, hack-cast it:

const mySchema = schema('object', {
  title: undefined as unknown as string,
});

Runtime footprint

Compiled to javascript, the helper functions look like this:

function schema(type, schema) {
  return { type, ...schema };
}

function field(type, schema) {
  return { type, ...schema };
}

function arrayOf(type, schema) {
  return { type, ...schema };
}

function typed(anything) {
  return anything;
}

Ie, they are basically an identity functions with no overhead.

Bonus feature: checkSchema

It is often useful to have the stored type of a Sanity document available in the studio, so it can be used in custom components, hidden-functions, validation-context or similar.

A common problem then, is that it is easy to forget to update the interface after adding a field to schema. To rectify this slightly, the checkSchema helper-function can be used.

Example

import { schema, field, checkSchema, forType } from '@snorreeb/sanity-typesafe-schemas';

// as const narrows string to 'my-type' literal
const type = 'my-type' as const; 

interface SchemaType {
    _type: typeof type;
    field1: string;
    field2: string;
}

export const mySchema = schema('object', {
    name: type,
    fields: [
        field('datetime', {
            name: 'field1', // field automatically narrows name from string to 'field1' literal type
            // intellisense for datetime-properties
        }),
        {
            name: 'field2' as const, // need const here, to narrow type from string to 'field1' literal
            type: 'string',
        },
    ],
});

// this will fail to compile if a new field is added to mySchema without also adding it in SchemaType 
// Note: it does NOT check that schema-field types correspond with interface types
checkSchema(forType<SchemaType>(), mySchema);

See src/examples/check-schema.ts for details.

FAQ

Why are Sanity schema types partially reimplemented?

Types in @sanity/types are the runtime definition of schemas, obtained after passing schema-definitions to createSchema.

The types we want when defining schemas are subtly different and has more optionality and must allow for extension. Some of the @sanity/types ae also to wide to be useful for autocompletion.

Therefore, types are reimplemented to resemble the Sanity schema documentation, in most places.

Why not use Schema types directly?

It is tempting to think this "just work":

const docSchema: DocumentSchema = {
  type: 'document',
  name: 'my-schema',
  title: 'My document',
  fields: [
    {
      // autocomplete all the things :fingers-crossed:
    },
  ],
};

This works fine for schemas. In fact, feel free to use the exported Schema-types in this repository like this if you like.

However, for fields, this becomes tricky, when a union type is used to describe all possible field-types: fields: UnionOfSchemas[].

First, until the discriminator type-field is provided, the autocomplete-listing is overwhelming, showing all possible fields for all possible schemas.

Second, even with the discriminator type-field, autocomplete in IDE's becomes unreliable for options and other type-fields that are common across schemas (due to unintended, partial type merging). The helper functions circumvent this by using generics to do exact type-lookup for each schema.

Third, types cannot be used in js-codebases. The helper-functions can though, and will still provide intellisense in IDE's.

Fourth, a downside of using schema-types directly is that developers need to know the name of the schema. With the helper functions, simply starting an open string for type will list all available schemas with a registered type-definition.

Usage no npm install needed!