Schema Validation

It doesn't take long before you are really tired of writing code that looks like this:
if (title === '') {
	errors.fieldErrors.title.push('Title is required')
}
if (title.length > titleMaxLength) {
	errors.fieldErrors.title.push('Title must be at most 100 characters')
}
if (content === '') {
	errors.fieldErrors.content.push('Content is required')
}
if (content.length > contentMaxLength) {
	errors.fieldErrors.content.push('Content must be at most 10000 characters')
}
I'm afraid to say it gets worse.
If you haven't felt the draw to write a utility to improve this yet, you will. For example, we could do something like this:
function validate(value: string, max: number) {
	const errors = []
	if (value === '') {
		errors.push('Required')
	}
	if (value.length > max) {
		errors.push(`Must be at most ${max} characters`)
	}
	return errors
}

errors.fieldErrors.title.push(...validate(title, titleMaxLength))
errors.fieldErrors.content.push(...validate(content, contentMaxLength))
Luckily, there are already libraries that do this for us so we don't have to trouble ourselves with writing our own validation. On top of this, they allow you to be much more declarative with your validation. The library we'll use looks like this:
z.object({
	title: z.string().max(titleMaxLength),
	content: z.string().max(contentMaxLength),
})
This is how you create your schema. And then you use that to perform the validation (parsing):
const result = schema.safeParse({ title, content })

if (result.success) {
	// we're good, check result.data
} else {
	// we're not good, check result.error
}
Schema validation is a great way to validate data due to its declarative nature.

Zod

The library we're going to use is Zod. Zod is a "TypeScript-first schema validation with static type inference" that allows you to define schemas for your data. A lot of the web ecosystem already uses it and it has great integrations with other libraries we'll want to use.
And because it has fantastic TypeScript support, it allows us to have a lot more confidence in our code as the types flow through our application. It's a great way to manage I/O boundaries to your application like reading from the filesystem, making HTTP requests, and interacting with databases.
Zod also allows for a great deal of customization and refinement in your validation and error messages. Here are some examples from the Zod docs:
z.string().max(5, { message: 'Must be 5 or fewer characters long' })

// even type inference based on the schema:
const A = z.string()
type A = z.infer<typeof A> // string

const name = z.string({
	required_error: 'Name is required',
	invalid_type_error: 'Name must be a string',
})

const user = z.object({
	username: z.string().optional(),
})

const nonEmptyStrings = z.array(z.string()).nonempty({
	message: "Can't be empty!",
})
Zod is extremely powerful. I recommend you have the documentation open during these exercises for reference.

Conform

One of the benefits to using a declarative interface for validation is that you can use that same schema to assist in the generation of that data–including for the forms users will fill out!
While you most definitely can generate the entire form based on the schema if you want to (I've done this before), we're going to keep our UI flexibility and just use the schema to give us type safe data to create the props for our forms.
You may not have thought about it, but often we duplicate our validation logic between our client and our server. We do it on the server because we have to and on the client because we want to give the user feedback as they're filling out the form (for a better UX).
With Zod, we can use the same schema to validate the data on the client and the server. But then, that's just the JavaScript portion. What about the HTML attributes?
If you have a schema like this:
const FormSchema = z.object({
	email: z.string().email(),
})
Then you want the form to look like this:
<form method="post">
	<label for="email-input">Email</label>
	<input id="email-input" type="email" required />
</form>
And when there's an error, we also want to have the right aria attributes:
<form method="post">
	<label for="email-input">Email</label>
	<input
		id="email-input"
		type="email"
		aria-invalid="true"
		aria-describedby="email-errors"
		required
	/>
	<ul id="email-errors">
		<li>Must be a valid email address</li>
	</ul>
</form>
Those attributes are very important (and very intentional). Without those attributes, the user doesn't get any client-side validation until the JavaScript finishes loading. Additionally, screen readers will reference these attributes in their assistance to the user. So these attributes are very important.
Luckily, we can use the schema to generate these attributes for us. We'll use a library called conform to do this. Conform is "a progressive enhancement first form validation library for Remix and React Router."
Progressive Enhancement is the idea that your application starts with a baseline functionality, and then you layer on additional functionality as the user's device and browser support it. Practically speaking, this means that our form will work before the JavaScript finishes loading (or even if it never does), and then we use the JavaScript to enhance the experience with better pending UI and faster transitions.
Conform has an adapter for Zod schemas with utilities that are perfect for what we're looking for. Here's an example:
import { conform, useForm } from '@conform-to/react'
import { getFieldsetConstraint, parse } from '@conform-to/zod'
import { Form } from '@remix-run/react'
import { json, redirect } from '@remix-run/node'
import { z } from 'zod'

const LoginSchema = z.object({
	email: z
		.string({ required_error: 'Email is required' })
		.email('Email is invalid'),
	password: z.string({ required_error: 'Password is required' }),
})

export async function action({ request }: ActionFunctionArgs) {
	const formData = await request.formData()
	const submission = parse(formData, {
		schema: LoginSchema,
	})

	if (submission.intent !== 'submit') {
		// the user hasn't submitted the form yet
		// this will happen if Conform is validating the form before submission
		// (like if we configure Conform to validate onBlur)
		return json({ status: 'idle', submission } as const)
	}

	if (!submission.value) {
		// there's no value because there is an error in the form
		return json({ status: 'error', submission } as const, {
			status: 400,
		})
	}

	const { email, password } = submission.value

	const isAuthenticated = await authenticate({ email, password })
	if (!isAuthenticated) {
		// set the form error:
		submission.error[''] = ['Invalid email or password']
		return json(
			{
				status: 'error',
				submission,
			} as const,
			{
				status: 401,
			},
		)
	}

	return redirect('/dashboard')
}

export default function LoginForm() {
	const actionData = useActionData<typeof action>()
	const [form, fields] = useForm({
		id: 'login-form',
		constraint: getFieldsetConstraint(LoginSchema),
		lastSubmission: actionData?.submission,
		onValidate({ formData }) {
			return parse(formData, { schema: LoginSchema })
		},
	})

	return (
		<Form method="post" {...form.props}>
			<div>
				<label htmlFor={fields.email.id}>Email</label>
				<input {...conform.input(fields.email)} />
				<ErrorList id={fields.email.errorId} errors={fields.email.errors} />
			</div>
			<div>
				<label htmlFor={fields.password.id}>Password</label>
				<input
					{...conform.input(fields.password, {
						type: 'password',
					})}
				/>
				<ErrorList
					id={fields.password.errorId}
					errors={fields.password.errors}
				/>
			</div>
			<ErrorList id={form.errorId} errors={form.errors} />
			<button type="submit">Login</button>
		</Form>
	)
}

function ErrorList({
	id,
	errors,
}: {
	id?: string
	errors?: Array<string> | null
}) {
	if (!errors) return null
	errors = Array.isArray(errors) ? errors : [errors]

	return errors.length ? (
		<ul id={id} className="flex flex-col gap-1">
			{errors.map((error, i) => (
				<li key={i} className="text-foreground-destructive text-[10px]">
					{error}
				</li>
			))}
		</ul>
	) : null
}
There's quite a bit in there, but we'll look through all of it piece by piece in the exercise.
Conform v1 was released after these workshops were created and while they will eventually receive an update, the breaking changes are minimal and the concepts are the same. Until these workshops are updated, feel free to watch the video below to see the changes (maybe bookmark it as something to watch after you finish the exercises).