@kentcdodds/tmp-remix-utils

Remix Utils

This package contains simple utility functions to use with Remix.run.

Installation

npm install remix-utils

Additional optional dependencies may be needed, all peer dependencies are:

• zod
• react
• @remix-run/react (also @remix-run/router but you sohuld be using the React one)
• @remix-run/node or @remix-run/cloudflare or @remix-run/deno (actually it's @remix-run/server-runtime but you should use one of the others)

The utils that require an extra optional dependency mention it in their documentation.

API Reference

promiseHash

The promiseHash function is not directly related to Remix but it's a useful function when working with loaders and actions.

This function is an object version of Promise.all which lets you pass an object with promises and get an object with the same keys with the resolved values.

import { promiseHash } from "remix-utils/promise";

export async function loader({ request }: LoaderArgs) {
  return json(
    await promiseHash({
      user: getUser(request),
      posts: getPosts(request),
    }),
  );
}

You can use nested promiseHash to get a nested object with resolved values.

import { promiseHash } from "remix-utils/promise";

export async function loader({ request }: LoaderArgs) {
  return json(
    await promiseHash({
      user: getUser(request),
      posts: promiseHash({
        list: getPosts(request),
        comments: promiseHash({
          list: getComments(request),
          likes: getLikes(request),
        }),
      }),
    }),
  );
}

timeout

The timeout function lets you attach a timeout to any promise, if the promise doesn't resolve or reject before the timeout, it will reject with a TimeoutError.

import { timeout } from "remix-utils/promise";

try {
  let result = await timeout(fetch("https://example.com"), { ms: 100 });
} catch (error) {
  if (error instanceof TimeoutError) {
    // Handle timeout
  }
}

Here the fetch needs to happen in less than 100ms, otherwise it will throw a TimeoutError.

If the promise is cancellable with an AbortSignal you can pass the AbortController to the timeout function.

import { timeout } from "remix-utils/promise";

try {
  let controller = new AbortController();
  let result = await timeout(
    fetch("https://example.com", { signal: controller.signal }),
    { ms: 100, controller },
  );
} catch (error) {
  if (error instanceof TimeoutError) {
    // Handle timeout
  }
}

Here after 100ms, timeout will call controller.abort() which will mark the controller.signal as aborted.

cacheAssets

Note This can only be run inside entry.client.

This function lets you easily cache inside the browser's Cache Storage every JS file built by Remix.

To use it, open your entry.client file and add this:

import { cacheAssets } from "remix-utils/cache-assets";

cacheAssets().catch((error) => {
  // do something with the error, or not
});

The function receives an optional options object with two options:

• cacheName is the name of the Cache object to use, the default value is assets.
• buildPath is the pathname prefix for all Remix built assets, the default value is /build/ which is the default build path of Remix itself.

It's important that if you changed your build path in remix.config.js you pass the same value to cacheAssets or it will not find your JS files.

The cacheName can be left as is unless you're adding a Service Worker to your app and want to share the cache.

import { cacheAssets } from "remix-utils/cache-assets";

cacheAssests({ cacheName: "assets", buildPath: "/build/" }).catch((error) => {
  // do something with the error, or not
});

ClientOnly

Note This depends on react.

The ClientOnly component lets you render the children element only on the client-side, avoiding rendering it the server-side.

You can provide a fallback component to be used on SSR, and while optional, it's highly recommended to provide one to avoid content layout shift issues.

import { ClientOnly } from "remix-utils/client-only";

export default function Component() {
  return (
    <ClientOnly fallback={<SimplerStaticVersion />}>
      {() => <ComplexComponentNeedingBrowserEnvironment />}
    </ClientOnly>
  );
}

This component is handy when you have some complex component that needs a browser environment to work, like a chart or a map. This way, you can avoid rendering it server-side and instead use a simpler static version like an SVG or even a loading UI.

The rendering flow will be:

• SSR: Always render the fallback.
• CSR First Render: Always render the fallback.
• CSR Update: Update to render the actual component.
• CSR Future Renders: Always render the actual component, don't bother to render the fallback.

This component uses the useHydrated hook internally.

ServerOnly

Note This depends on react.

The ServerOnly component is the opposite of the ClientOnly component, it lets you render the children element only on the server-side, avoiding rendering it the client-side.

You can provide a fallback component to be used on CSR, and while optional, it's highly recommended to provide one to avoid content layout shift issues, unless you only render visually hidden elements.

import { ServerOnly } from "remix-utils/server-only";

export default function Component() {
  return (
    <ServerOnly fallback={<ComplexComponentNeedingBrowserEnvironment />}>
      {() => <SimplerStaticVersion />}
    </ServerOnly>
  );
}

This component is handy to render some content only on the server-side, like a hidden input you can later use to know if JS has loaded.

Consider it like the <noscript> HTML tag but it can work even if JS failed to load but it's enabled on the browser.

The rendering flow will be:

• SSR: Always render the children.
• CSR First Render: Always render the children.
• CSR Update: Update to render the fallback component (if defined).
• CSR Future Renders: Always render the fallback component, don't bother to render the children.

This component uses the useHydrated hook internally.

CORS

The CORS function let you implement CORS headers on your loaders and actions so you can use them as an API for other client-side applications.

There are two main ways to use the cors function.

1. Use it on each loader/action where you want to enable it.
2. Use it globally on entry.server handleRequest and handleDataRequest export.

If you want to use it on every loader/action, you can do it like this:

import { cors } from "remix-utils/cors";

export async function loader({ request }: LoaderArgs) {
  let data = await getData(request);
  let response = json<LoaderData>(data);
  return await cors(request, response);
}

You could also do the json and cors call in one line.

import { cors } from "remix-utils/cors";

export async function loader({ request }: LoaderArgs) {
  let data = await getData(request);
  return await cors(request, json<LoaderData>(data));
}

And because cors mutates the response, you can also call it and later return.

import { cors } from "remix-utils/cors";

export async function loader({ request }: LoaderArgs) {
  let data = await getData(request);
  let response = json<LoaderData>(data);
  await cors(request, response); // this mutates the Response object
  return response; // so you can return it here
}

If you want to setup it globally once, you can do it like this in entry.server

import { cors } from "remix-utils/cors";

const ABORT_DELAY = 5000;

export default function handleRequest(
  request: Request,
  responseStatusCode: number,
  responseHeaders: Headers,
  remixContext: EntryContext,
) {
  let callbackName = isbot(request.headers.get("user-agent"))
    ? "onAllReady"
    : "onShellReady";

  return new Promise((resolve, reject) => {
    let didError = false;

    let { pipe, abort } = renderToPipeableStream(
      <RemixServer context={remixContext} url={request.url} />,
      {
        [callbackName]: () => {
          let body = new PassThrough();

          responseHeaders.set("Content-Type", "text/html");

          cors(
            request,
            new Response(body, {
              headers: responseHeaders,
              status: didError ? 500 : responseStatusCode,
            }),
          ).then((response) => {
            resolve(response);
          });

          pipe(body);
        },
        onShellError: (err: unknown) => {
          reject(err);
        },
        onError: (error: unknown) => {
          didError = true;

          console.error(error);
        },
      },
    );

    setTimeout(abort, ABORT_DELAY);
  });
}

export let handleDataRequest: HandleDataRequestFunction = async (
  response,
  { request },
) => {
  return await cors(request, response);
};

Options

Additionally, the cors function accepts a options object as a third optional argument. These are the options.

• origin: Configures the Access-Control-Allow-Origin CORS header. Possible values are:
  • true: Enable CORS for any origin (same as "*")
  • false: Don't setup CORS
  • string: Set to a specific origin, if set to "*" it will allow any origin
  • RegExp: Set to a RegExp to match against the origin
  • Array<string | RegExp>: Set to an array of origins to match against the string or RegExp
  • Function: Set to a function that will be called with the request origin and should return a boolean indicating if the origin is allowed or not. The default value is true.
• methods: Configures the Access-Control-Allow-Methods CORS header. The default value is ["GET", "HEAD", "PUT", "PATCH", "POST", "DELETE"].
• allowedHeaders: Configures the Access-Control-Allow-Headers CORS header.
• exposedHeaders: Configures the Access-Control-Expose-Headers CORS header.
• credentials: Configures the Access-Control-Allow-Credentials CORS header.
• maxAge: Configures the Access-Control-Max-Age CORS header.

CSRF

Note: This depends on react, and a Remix server runtime.

The CSRF related functions let you implement CSRF protection on your application.

This part of Remix Utils needs React and server-side code.

First create a new CSRF instance.

// app/utils/csrf.server.ts
import { CSRF } from "remix-utils/csrf";
import { createCookie } from "@remix-run/node"; // or /cloudflare

export const cookie = createCookie("csrf", {
  path: "/",
  httpOnly: true,
  secure: process.env.NODE_ENV === "production",
  sameSite: "lax",
  secrets: ["s3cr3t"],
});

export const csrf = new CSRF({
  cookie,
  // what key in FormData objects will be used for the token, defaults to `csrf`
  formDataKey: "csrf",
  // an optional secret used to sign the token, recommended for extra safety
  secret: "s3cr3t",
});

Then you can use csrf to generate a new token.

import { csrf } from "~/utils/csrf.server";

export async function loader({ request }: LoaderArgs) {
  let token = csrf.generate();
}

You can customize the token size by passing the byte size, the default one is 32 bytes which will give you a string with a length of 43 after encoding.

let token = csrf.generate(64); // customize token length

You will need to save this token in a cookie and also return it from the loader. For convenience, you can use the CSRF#commitToken helper.

import { csrf } from "~/utils/csrf.server";

export async function loader({ request }: LoaderArgs) {
  let [token, cookieHeader] = await csrf.commitToken();
  return json({ token }, { headers: { "set-cookie": cookieHeader } });
}

Note: You could do this on any route, but I recommend you to do it on the root loader.

Now that you returned the token and set it in a cookie, you can use the AuthenticityTokenProvider component to provide the token to your React components.

import { AuthenticityTokenProvider } from "remix-utils/authenticity-token";

let { csrf } = useLoaderData<LoaderData>();
return (
  <AuthenticityTokenProvider token={csrf}>
    <Outlet />
  </AuthenticityTokenProvider>
);

Render it in your root component and wrap the Outlet with it.

When you create a form in some route, you can use the AuthenticityTokenInput component to add the authenticity token to the form.

import { Form } from "@remix-run/react";
import { AuthenticityTokenInput } from "remix-utils/authenticity-token";

export default function Component() {
  return (
    <Form method="post">
      <AuthenticityTokenInput />
      <input type="text" name="something" />
    </Form>
  );
}

Note that the authenticity token is only really needed for a form that mutates the data somehow. If you have a search form making a GET request, you don't need to add the authenticity token there.

This AuthenticityTokenInput will get the authenticity token from the AuthenticityTokenProvider component and add it to the form as the value of a hidden input with the name csrf. You can customize the field name using the name prop.

<AuthenticityTokenInput name="customName" />

You should only customize the name if you also changed it on createAuthenticityToken.

If you need to use useFetcher (or useSubmit) instead of Form you can also get the authenticity token with the useAuthenticityToken hook.

import { useFetcher } from "@remix-run/react";
import { useAuthenticityToken } from "remix-utils/authenticity-token";

export function useMarkAsRead() {
  let fetcher = useFetcher();
  let csrf = useAuthenticityToken();
  return function submit(data) {
    fetcher.submit(
      { csrf, ...data },
      { action: "/api/mark-as-read", method: "post" },
    );
  };
}

Finally, you need to validate the authenticity token in the action that received the request.

import { CSRFError } from "remix-utils/csrf";
import { redirectBack } from "remix-utils/redirect-back";
import { csrf } from "~/utils/csrf.server";

export async function action({ request }: ActionArgs) {
  try {
    await csrf.validate(request);
  } catch (error) {
    if (error instanceof CSRFError) {
      // handle CSRF errors
    }
    // handle other possible errors
  }

  // here you know the request is valid
  return redirectBack(request, { fallback: "/fallback" });
}

If you need to parse the body as FormData yourself (e.g. to support file uploads) you can also call CSRF#validate with the FormData and Headers objects.

let formData = await parseMultiPartFormData(request);
try {
  await csrf.validate(formData, request.headers);
} catch (error) {
  // handle errors
}

Warning: If you call CSRF#validate with the request instance, but you already read its body, it will throw an error.

In case the CSRF validation fails, it will throw a CSRFError which can be used to correctly identify it against other possible errors that may get thrown.

The list of possible error messages are:

• missing_token_in_cookie: The request is missing the CSRF token in the cookie.
• invalid_token_in_cookie: The CSRF token is not valid (is not a string).
• tampered_token_in_cookie: The CSRF token doesn't match the signature.
• missing_token_in_body: The request is missing the CSRF token in the body (FormData).
• mismatched_token: The CSRF token in the cookie and the body don't match.

You can use error.code to check one of the error codes above, and error.message to get a human friendly description.

Warning: Don't send those error messages to the end-user, they are meant to be used for debugging purposes only.

ExternalScripts

Note: This depends on react, @remix-run/react, and a Remix server runtime.

If you need to load different external scripts on certain routes, you can use the ExternalScripts component together with the ExternalScriptsFunction type.

In the route you want to load the script add a handle export with a scripts method, this method should implement the ExternalScriptsFunction type.

import { ExternalScriptsFunction } from "remix-utils/external-scripts";

type LoaderData = SerializeFrom<typeof loader>

// create the scripts function with the correct type (loader type is optional)
// and export it through the handle
export let handle = {
  scripts({ id, data, params, matches, location, parentsData }) {
    return [
      {
        src: "https://code.jquery.com/jquery-3.6.0.min.js",
        integrity: "sha256-/xUj+3OJU5yExlq6GSYGSHk7tPXikynS7ogEvDej/m4=",
        crossOrigin: "anonymous",
      },
    ];
  },
} satifies { scripts: ExternalScriptsFunction<LoaderData> }

Then, in the root route, add the ExternalScripts component together with the Remix's Scripts component, usually inside a Document component.

import { Links, LiveReload, Meta, Scripts, ScrollRestoration } from "remix";
import { ExternalScripts } from "remix-utils/external-scripts";

type Props = { children: React.ReactNode; title?: string };

export function Document({ children, title }: Props) {
  return (
    <html lang="en">
      <head>
        <meta charSet="utf-8" />
        <meta name="viewport" content="width=device-width,initial-scale=1" />
        {title ? <title>{title}</title> : null}
        <Meta />
        <Links />
      </head>
      <body>
        {children}
        <ScrollRestoration />
        <ExternalScripts />
        <Scripts />
        <LiveReload />
      </body>
    </html>
  );
}

Now, any script you defined in the ScriptsFunction will be added to the HTML together with a <link rel="preload"> before it.

Tip: You could use it together with useShouldHydrate to disable Remix scripts in certain routes but still load scripts for analytics or small features that need JS but don't need the full app JS to be enabled.

useGlobalNavigationState

Note: This depends on react, and @remix-run/react.

This hook allows you to read the value of transition.state, every fetcher.state in the app, and revalidator.state.

import { useGlobalNavigationState } from "remix-utils/use-global-navigation-state";

export function GlobalPendingUI() {
  let states = useGlobalNavigationState();

  if (state.includes("loading")) {
    // The app is loading.
  }

  if (state.includes("submitting")) {
    // The app is submitting.
  }

  // The app is idle
}

The return value of useGlobalNavigationState can be "idle", "loading" or "submitting"

Note This is used by the hooks below to determine if the app is loading, submitting or both (pending).

useGlobalPendingState

Note: This depends on react, and @remix-run/react.

This hook lets you know if the global navigation, if one of any active fetchers is either loading or submitting, or if the revalidator is running.

import { useGlobalPendingState } from "remix-utils/use-global-navigation-state";

export function GlobalPendingUI() {
  let globalState = useGlobalPendingState();

  if (globalState === "idle") return null;
  return <Spinner />;
}

The return value of useGlobalPendingState is either "idle" or "pending".

Note: This hook combines the useGlobalSubmittingState and useGlobalLoadingState hooks to determine if the app is pending.

Note: The pending state is a combination of the loading and submitting states introduced by this hook.

useGlobalSubmittingState

Note: This depends on react, and @remix-run/react.

This hook lets you know if the global transition or if one of any active fetchers is submitting.

import { useGlobalSubmittingState } from "remix-utils/use-global-navigation-state";

export function GlobalPendingUI() {
  let globalState = useGlobalSubmittingState();

  if (globalState === "idle") return null;
  return <Spinner />;
}

The return value of useGlobalSubmittingState is either "idle" or "submitting".

useGlobalLoadingState

Note: This depends on react, and @remix-run/react.

This hook lets you know if the global transition, if one of any active fetchers is loading, or if the revalidator is running

import { useGlobalLoadingState } from "remix-utils/use-global-navigation-state";

export function GlobalPendingUI() {
  let globalState = useGlobalLoadingState();

  if (globalState === "idle") return null;
  return <Spinner />;
}

The return value of useGlobalLoadingState is either "idle" or "loading".

useHydrated

Note: This depends on react.

This hook lets you detect if your component is already hydrated. This means the JS for the element loaded client-side and React is running.

With useHydrated, you can render different things on the server and client while ensuring the hydration will not have a mismatched HTML.

import { useHydrated } from "remix-utils/use-hydrated";

export function Component() {
  let isHydrated = useHydrated();

  if (isHydrated) {
    return <ClientOnlyComponent />;
  }

  return <ServerFallback />;
}

When doing SSR, the value of isHydrated will always be false. The first client-side render isHydrated will still be false, and then it will change to true.

After the first client-side render, future components rendered calling this hook will receive true as the value of isHydrated. This way, your server fallback UI will never be rendered on a route transition.

useLocales

Note: This depends on react.

This hooks lets you get the locales returned by the root loader. It follows a simple convention, your root loader return value should be an objet with the key locales.

You can combine it with getClientLocal to get the locales on the root loader and return that. The return value of useLocales is a Locales type which is string | string[] | undefined.

import { useLocales } from "remix-utils/use-locales";
import { getClientLocales } from "remix-utils/get-client-locales";

// in the root loader
export async function loader({ request }: LoaderArgs) {
  let locales = getClientLocales(request);
  return json({ locales });
}

// in any route (including root!)
export default function Component() {
  let locales = useLocales();
  let date = new Date();
  let dateTime = date.toISOString;
  let formattedDate = date.toLocaleDateString(locales, options);
  return <time dateTime={dateTime}>{formattedDate}</time>;
}

The return type of useLocales is ready to be used with the Intl API.

useShouldHydrate

Note: This depends on @remix-run/react.

If you are building a Remix application where most routes are static, and you want to avoid loading client-side JS, you can use this hook, plus some conventions, to detect if one or more active routes needs JS and only render the Scripts component in that case.

In your document component, you can call this hook to dynamically render the Scripts component if needed.

import type { ReactNode } from "react";
import { Links, LiveReload, Meta, Scripts } from "@remix-run/react";
import { useShouldHydrate } from "remix-utils/use-should-hydrate";

interface DocumentProps {
  children: ReactNode;
  title?: string;
}

export function Document({ children, title }: DocumentProps) {
  let shouldHydrate = useShouldHydrate();
  return (
    <html lang="en">
      <head>
        <meta charSet="utf-8" />
        <link rel="icon" href="/favicon.png" type="image/png" />
        {title ? <title>{title}</title> : null}
        <Meta />
        <Links />
      </head>
      <body>
        {children}
        {shouldHydrate && <Scripts />}
        <LiveReload />
      </body>
    </html>
  );
}

Now, you can export a handle object with the hydrate property as true in any route module.

export let handle = { hydrate: true };

This will mark the route as requiring JS hydration.

In some cases, a route may need JS based on the data the loader returned. For example, if you have a component to purchase a product, but only authenticated users can see it, you don't need JS until the user is authenticated. In that case, you can make hydrate be a function receiving your loader data.

export let handle = {
  hydrate(data: LoaderData) {
    return data.user.isAuthenticated;
  },
};

The useShouldHydrate hook will detect hydrate as a function and call it using the route data.

getClientIPAddress

This function receives a Request or Headers objects and will try to get the IP address of the client (the user) who originated the request.

import { getClientIPAddress } from "remix-utils/get-client-ip-address";

export async function loader({ request }: LoaderArgs) {
  // using the request
  let ipAddress = getClientIPAddress(request);
  // or using the headers
  let ipAddress = getClientIPAddress(request.headers);
}

If it can't find he IP address the return value will be null. Remember to check if it was able to find it before using it.

The function uses the following list of headers, in order of preference:

• X-Client-IP
• X-Forwarded-For
• Fly-Client-IP
• CF-Connecting-IP
• Fastly-Client-Ip
• True-Client-Ip
• X-Real-IP
• X-Cluster-Client-IP
• X-Forwarded
• Forwarded-For
• Forwarded
• DO-Connecting-IP
• oxygen-buyer-ip

When a header is found that contains a valid IP address, it will return without checking the other headers.

getClientLocales

This function let you get the locales of the client (the user) who originated the request.

import { getClientLocales } from "remix-utils/get-client-locales";

export async function loader({ request }: LoaderArgs) {
  // using the request
  let locales = getClientLocales(request);
  // or using the headers
  let locales = getClientLocales(request.headers);
}

The return value is a Locales type, which is string | string[] | undefined.

The returned locales can be directly used on the Intl API when formatting dates, numbers, etc.

import { getClientLocales } from "remix-utils/get-client-locales";
export async function loader({ request }: LoaderArgs) {
  let locales = getClientLocales(request);
  let nowDate = new Date();
  let formatter = new Intl.DateTimeFormat(locales, {
    year: "numeric",
    month: "long",
    day: "numeric",
  });
  return json({ now: formatter.format(nowDate) });
}

The value could also be returned by the loader and used on the UI to ensure the user's locales is used on both server and client formatted dates.

isPrefetch

This function let you identify if a request was created because of a prefetch triggered by using <Link prefetch="intent"> or <Link prefetch="render">.

This will let you implement a short cache only for prefetch requests so you avoid the double data request.

import { isPrefetch } from "remix-utils/is-prefetch";

export async function loader({ request }: LoaderArgs) {
  let data = await getData(request);
  let headers = new Headers();

  if (isPrefetch(request)) {
    headers.set("Cache-Control", "private, max-age=5, smax-age=0");
  }

  return json(data, { headers });
}

Responses

Redirect Back

This function is a wrapper of the redirect helper from Remix. Unlike Remix's version, this one receives the whole request object as the first value and an object with the response init and a fallback URL.

The response created with this function will have the Location header pointing to the Referer header from the request, or if not available, the fallback URL provided in the second argument.

import { redirectBack } from "remix-utils/redirect-back";

export async function action({ request }: ActionArgs) {
  return redirectBack(request, { fallback: "/" });
}

This helper is most useful when used in a generic action to send the user to the same URL it was before.

Not Modified

Helper function to create a Not Modified (304) response without a body and any header.

import { notModified } from "remix-utils/responses";

export async function loader({ request }: LoaderArgs) {
  return notModified();
}

JavaScript

Helper function to create a JavaScript file response with any header.

This is useful to create JS files based on data inside a Resource Route.

import { javascript } from "remix-utils/responses";

export async function loader({ request }: LoaderArgs) {
  return javascript("console.log('Hello World')");
}

Stylesheet

Helper function to create a CSS file response with any header.

This is useful to create CSS files based on data inside a Resource Route.

import { stylesheet } from "remix-utils/responses";

export async function loader({ request }: LoaderArgs) {
  return stylesheet("body { color: red; }");
}

PDF

Helper function to create a PDF file response with any header.

This is useful to create PDF files based on data inside a Resource Route.

import { pdf } from "remix-utils/responses";

export async function loader({ request }: LoaderArgs) {
  return pdf(await generatePDF(request.formData()));
}

HTML

Helper function to create a HTML file response with any header.

This is useful to create HTML files based on data inside a Resource Route.

import { html } from "remix-utils/responses";

export async function loader({ request }: LoaderArgs) {
  return html("<h1>Hello World</h1>");
}

XML

Helper function to create a XML file response with any header.

This is useful to create XML files based on data inside a Resource Route.

import { xml } from "remix-utils/responses";

export async function loader({ request }: LoaderArgs) {
  return xml("<?xml version='1.0'?><catalog></catalog>");
}

TXT

Helper function to create a TXT file response with any header.

This is useful to create TXT files based on data inside a Resource Route.

import { txt } from "remix-utils/responses";

export async function loader({ request }: LoaderArgs) {
  return txt(`
    User-agent: *
    Allow: /
  `);
}

Typed Cookies

Note: This depends on zod, and a Remix server runtime.

Cookie objects in Remix allows any type, the typed cookies from Remix Utils lets you use Zod to parse the cookie values and ensure they conform to a schema.

import { createCookie } from "@remix-run/node";
import { createTypedCookie } from "remix-utils/typed-cookie";
import { z } from "zod";

let cookie = createCookie("returnTo", cookieOptions);
// I recommend you to always add `nullable` to your schema, if a cookie didn't
// come with the request Cookie header Remix will return null, and it can be
// useful to remove it later when clearing the cookie
let schema = z.string().url().nullable();

// pass the cookie and the schema
let typedCookie = createTypedCookie({ cookie, schema });

// this will be a string and also a URL
let returnTo = await typedCookie.parse(request.headers.get("Cookie"));

// this will not pass the schema validation and throw a ZodError
await typedCookie.serialize("a random string that's not a URL");
// this will make TS yell because it's not a string, if you ignore it it will
// throw a ZodError
await typedCookie.serialize(123);

You could also use typed cookies with any sessionStorage mechanism from Remix.

let cookie = createCookie("session", cookieOptions);
let schema = z.object({ token: z.string() }).nullable();

let sessionStorage = createCookieSessionStorage({
  cookie: createTypedCookie({ cookie, schema }),
});

// if this works then the correct data is stored in the session
let session = sessionStorage.getSession(request.headers.get("Cookie"));

session.unset("token"); // remove a required key from the session

// this will throw a ZodError because the session is missing the required key
await sessionStorage.commitSession(session);

Now Zod will ensure the data you try to save to the session is valid removing any extra field and throwing if you don't set the correct data in the session.

Note The session object is not really typed so doing session.get will not return the correct type, you can do schema.parse(session.data) to get the typed version of the session data.

You can also use async refinements in your schemas because typed cookies uses parseAsync method from Zod.

let cookie = createCookie("session", cookieOptions);

let schema = z
  .object({
    token: z.string().refine(async (token) => {
      let user = await getUserByToken(token);
      return user !== null;
    }, "INVALID_TOKEN"),
  })
  .nullable();

let sessionTypedCookie = createTypedCookie({ cookie, schema });

// this will throw if the token stored in the cookie is not valid anymore
sessionTypedCookie.parse(request.headers.get("Cookie"));

Finally, to be able to delete a cookie, you can add .nullable() to your schema and serialize it with null as value.

// Set the value as null and expires as current date - 1 second so the browser expires the cookie
await typedCookie.serialize(null, { expires: new Date(Date.now() - 1) });

If you didn't add .nullable() to your schema, you will need to provide a mock value and set the expires date to the past.

let cookie = createCookie("returnTo", cookieOptions);
let schema = z.string().url().nullable();

let typedCookie = createTypedCookie({ cookie, schema });

await typedCookie.serialize("some fake url to pass schema validation", {
  expires: new Date(Date.now() - 1),
});

Typed Sessions

Note: This depends on zod, and a Remix server runtime.

Session objects in Remix allows any type, the typed sessions from Remix Utils lets you use Zod to parse the session data and ensure they conform to a schema.

import { createCookieSessionStorage } from "@remix-run/node";
import { createTypedSessionStorage } from "remix-utils/typed-session";
import { z } from "zod";

let schema = z.object({
  token: z.string().optional(),
  count: z.number().default(1),
});

// you can use a Remix's Cookie container or a Remix Utils' Typed Cookie container
let sessionStorage = createCookieSessionStorage({ cookie });

// pass the session storage and the schema
let typedSessionStorage = createTypedSessionStorage({ sessionStorage, schema });

Now you can use typedSessionStorage as a drop-in replacement for your normal sessionStorage.

let session = typedSessionStorage.getSession(request.headers.get("Cookie"));

session.get("token"); // this will be a string or undefined
session.get("count"); // this will be a number
session.get("random"); // this will make TS yell because it's not in the schema

session.has("token"); // this will be a boolean
session.has("count"); // this will be a boolean

// this will make TS yell because it's not a string, if you ignore it it will
// throw a ZodError
session.set("token", 123);

Now Zod will ensure the data you try to save to the session is valid by not allowing you to get, set or unset data.

Note Remember that you either need to mark fields as optional or set a default value in the schema, otherwise it will be impossible to call getSession to get a new session object.

You can also use async refinements in your schemas because typed sesions uses parseAsync method from Zod.

let schema = z.object({
  token: z
    .string()
    .optional()
    .refine(async (token) => {
      if (!token) return true; // handle optionallity
      let user = await getUserByToken(token);
      return user !== null;
    }, "INVALID_TOKEN"),
});

let typedSessionStorage = createTypedSessionStorage({ sessionStorage, schema });

// this will throw if the token stored in the session is not valid anymore
typedSessionStorage.getSession(request.headers.get("Cookie"));

Server-Sent Events

Note: This depends on react.

Server-Sent Events are a way to send data from the server to the client without the need for the client to request it. This is useful for things like chat applications, live updates, and more.

There are two utils provided to help with the usage inside Remix:

• eventStream
• useEventSource

The eventStream function is used to create a new event stream response needed to send events to the client. This must live in a Resource Route.

// app/routes/sse.time.ts
import { eventStream } from "remix-utils/event-stream";

export async function loader({ request }: LoaderArgs) {
  return eventStream(request.signal, function setup(send) {
    let timer = setInterval(() => {
      send({ event: "time", data: new Date().toISOString() });
    }, 1000);

    return function clear() {
      clearInterval(timer);
    };
  });
}

Then, inside any component, you can use the useEventSource hook to connect to the event stream.

// app/components/counter.ts
import { useEventSource } from "remix-utils/use-event-source";

function Counter() {
  // Here `/sse/time` is the resource route returning an eventStream response
  let time = useEventSource("/sse/time", { event: "time" });

  if (!time) return null;

  return (
    <time dateTime={time}>
      {new Date(time).toLocaleTimeString("en", {
        minute: "2-digit",
        second: "2-digit",
        hour: "2-digit",
      })}
    </time>
  );
}

The event name in both the event stream and the hook is optional, in which case it will default to message, if defined you must use the same event name in both sides, this also allows you to emit different events from the same event stream.

For Server-Sent Events to work, your server must support HTTP streaming. If you don't get SSE to work check if your deployment platform has support for it.

Because SSE count towards the limit of HTTP connections per domain, the useEventSource hook keeps a global map of connections based on the provided URL and options. As long as they are the same, the hook will open a single SSE connection and share it between instances of the hook.

Once there are no more instances of the hook re-using a connection, it will be closed and removed from the map.

You can use the <EventSourceProvider /> component to control the map.

let map: EventSourceMap = new Map();
return (
  <EventSourceProvider value={map}>
    <YourAppOrPartOfIt />
  </EventSourceProvider>
);

This way, you can overwrite the map with a new one for a specific part of your app. Note that this provider is optional and a default map will be used if you don't provide one.

Rolling Cookies

Note: This depends on zod, and a Remix server runtime.

Rolling cookies allows you to prolong the expiration of a cookie by updating the expiration date of every cookie.

The rollingCookie function is prepared to be used in entry.server exported function to update the expiration date of a cookie if no loader set it.

For document request you can use it on the handleRequest function:

import { rollingCookie } from "remix-utils/rolling-cookie";

import { sessionCookie } from "~/session.server";

export default function handleRequest(
  request: Request,
  responseStatusCode: number,
  responseHeaders: Headers,
  remixContext: EntryContext,
) {
  await rollingCookie(sessionCookie, request, responseHeaders);

  return isbot(request.headers.get("user-agent"))
    ? handleBotRequest(
        request,
        responseStatusCode,
        responseHeaders,
        remixContext,
      )
    : handleBrowserRequest(
        request,
        responseStatusCode,
        responseHeaders,
        remixContext,
      );
}

And for data request you can do it on the handleDataRequest function:

import { rollingCookie } from "remix-utils/rolling-cookie";

export let handleDataRequest: HandleDataRequestFunction = async (
  response: Response,
  { request },
) => {
  let cookieValue = await sessionCookie.parse(
    responseHeaders.get("set-cookie"),
  );
  if (!cookieValue) {
    cookieValue = await sessionCookie.parse(request.headers.get("cookie"));
    responseHeaders.append(
      "Set-Cookie",
      await sessionCookie.serialize(cookieValue),
    );
  }

  return response;
};

Note: Read more about rolling cookies in Remix.

Named actions

Note: This depends on a Remix server runtime.

It's common to need to handle more than one action in the same route, there are many options here like sending the form to a resource route or using an action reducer, the namedAction function uses some conventions to implement the action reducer pattern.

import { namedAction } from "remix-utils/named-action";

export async function action({ request }: ActionArgs) {
  return namedAction(request, {
    async create() {
      // do create
    },
    async update() {
      // do update
    },
    async delete() {
      // do delete
    },
  });
}

export default function Component() {
  return (
    <>
      <Form method="post" action="?/create">
        ...
      </Form>

      <Form method="post" action="?/update">
        ...
      </Form>

      <Form method="post" action="?/delete">
        ...
      </Form>
    </>
  );
}

This function can follow many conventions

You can pass a FormData object to the namedAction, then it will try to

• Find a field named /something and use it as the action name removing the /
• Find a field named intent and use the value as the action name
• Find a field named action and use the value as the action name
• Find a field named _action and use the value as the action name

You can pass an URLSearchParams object to the namedAction, then it will try to

• Find a query parameter named /something and use it as the action name removing the /
• Find a query parameter named intent and use the value as the action name
• Find a query parameter named action and use the value as the action name
• Find a query parameter named _action and use the value as the action name

You can pass an URL object to the namedAction, it will behave as with a URLSearchParams object.

You can pass a Request object to the namedAction, then it will try to

• Call new URL(request.url) and use it as the URL object
• Call request.formData() and use it as the FormData object

If, in any case, the action name is not found, the actionName then the library will try to call an action named default, similar to a switch in JavaScript.

If the default is not defined it will throw a ReferenceError with the message Action "${name}" not found.

If the library couldn't found the name at all, it will throw a ReferenceError with the message Action name not found

Preload Route Assets

The Link header allows responses to push to the browser assets that are needed for the document, this is useful to improve the performance of the application by sending those assets earlier.

Once Early Hints is supported this will also allows you to send the assets even before the document is ready, but for now you can benefit to send assets to preload before the browser parse the HTML.

You can do this with the functions preloadRouteAssets, preloadLinkedAssets and preloadModuleAssets.

All functions follows the same signature:

// entry.server.tsx
export default function handleRequest(
  request: Request,
  statusCode: number,
  headers: Headers,
  context: EntryContext,
) {
  let markup = renderToString(
    <RemixServer context={context} url={request.url} />,
  );
  headers.set("Content-Type", "text/html");

  preloadRouteAssets(context, headers); // add this line
  // preloadLinkedAssets(context, headers);
  // preloadModuleAssets(context, headers);

  return new Response("<!DOCTYPE html>" + markup, {
    status: statusCode,
    headers: headers,
  });
}

The preloadRouteAssets is a combination of both preloadLinkedAssets and preloadModuleAssets so you can use it to preload all assets for a route, if you use this one you don't need the other two

The preloadLinkedAssets function will preload any link with rel: "preload" added with the Remix's LinkFunction, so you can configure assets to preload in your route and send them in the headers automatically. It will additionally preload any linked stylesheet file (with rel: "stylesheet") even if not preloaded so it will load faster.

The preloadModuleAssets function will preload all the JS files Remix adds to the page when hydrating it, Remix already renders a <link rel="modulepreload"> for each now before the <script type="module"> used to start the application, this will use Link headers to preload those assets.

Safe Redirects

When performing a redirect, if the URL is user provided we can't trust it, if you do you're opening a vulnerability to phishing scam by allowing bad actors to redirect the user to malicious websites.

https://remix.utills/?redirectTo=https://malicious.app

To help you prevent this Remix Utils gives you a safeRedirect function which can be used to check if the URL is "safe".

Note: In this context, safe means the URL starts with / but not //, this means the URL is a pathname inside the same app and not an external link.

export async function loader({ request }: LoaderArgs) {
  let { searchParams } = new URL(request.url);
  let redirectTo = searchParams.get("redirectTo");
  return redirect(safeRedirect(redirectTo, "/home"));
}

The second argumento of safeRedirect is the default redirect which by when not configured is /, this lets you tell safeRedirect where to redirect the user if the value is not safe.

JSON Hash Response

When returning a json from a loader function, you may need to get data from different DB queries or API requests, typically you would something like this

export async function loader({ params }: LoaderData) {
  let postId = z.string().parse(params.postId);
  let [post, comments] = await Promise.all([getPost(), getComments()]);
  return json({ post, comments });

  async function getPost() {
    /* … */
  }
  async function getComments() {
    /* … */
  }
}

The jsonHash function lets you define those functions directly in the json, reducing the need to create extra functions and variables.

export async function loader({ params }: LoaderData) {
  let postId = z.string().parse(params.postId);
  return jsonHash({
    async post() {
      // Implement me
    },
    async comments() {
      // Implement me
    },
  });
}

It also calls your functions using Promise.all so you can be sure the data is retrieved in parallel.

Additionally, you can pass non-async functions, values and promises.

export async function loader({ params }: LoaderData) {
  let postId = z.string().parse(params.postId);
  return jsonHash({
    postId, // value
    comments: getComments(), // Promise
    slug() {
      // Non-async function
      return postId.split("-").at(1); // get slug from postId param
    },
    async post() {
      // Async function
      // Implement me
    },
  });

  async function getComments() {
    /* … */
  }
}

The result of jsonHash is a TypedResponse and it's correctly typed so using it with typeof loader works flawlessly.

export default function Component() {
  // all correctly typed
  let { postId, comments, slug, post } = useLoaderData<typeof loader>();

  // more code…
}

Delegate Anchors to Remix

When using Remix, you can use the <Link> component to navigate between pages. However, if you have a <a href> that links to a page in your app, it will cause a full page refresh. This can be what you want, but sometimes you want to use client-side navigation here instead.

The useDelegatedAnchors hook lets you add client-side navigation to anchor tags in a portion of your app. This can be specially useful when working with dynamic content like HTML or Markdown from a CMS.

import { useDelegatedAnchors } from "remix-utils/use-delegated-anchors";

export async function loader() {
  let content = await fetchContentFromCMS();
  return json({ content });
}

export default function Component() {
  let { content } = useLoaderData<typeof loader>();

  let ref = useRef<HTMLDivElement>(null);
  useDelegatedAnchors(ref);

  return <article ref={ref} dangerouslySetInnerHTML={{ __html: content }} />;
}

Prefetch Anchors

If additionally you want to be able to prefetch your anchors you can use the PrefetchPageAnchors components.

This components wraps your content with anchors inside, it detects any hovered anchor to prefetch it, and it delegates them to Remix.

import { PrefetchPageAnchors } from "remix-utils/use-delegated-anchors";

export async function loader() {
  let content = await fetchContentFromCMS();
  return json({ content });
}

export default function Component() {
  let { content } = useLoaderData<typeof loader>();

  return (
    <PrefetchPageAnchors>
      <article ref={ref} dangerouslySetInnerHTML={{ __html: content }} />
    </PrefetchPageAnchors>
  );
}

Now you can see in your DevTools that when the user hovers an anchor it will prefetch it, and when the user clicks it will do a client-side navigation.

Debounced Fetcher

Note: This depends on react, and @remix-run/react.

The useDebounceFetcher is a wrapper of useFetcher that adds debounce support to fetcher.submit.

The hook is based on @JacobParis article.

The main difference with Jacob's version is that Remix Utils' version overwrites fetcher.submit instead of appending a fetcher.debounceSubmit method.

import { useDebounceFetcher } from "remix-utils/use-debounce-fetcher";

export function Component({ data }) {
  let fetcher = useDebounceFetcher<Type>();

  function handleClick() {
    fetcher.submit(data, { debounceTimeout: 1000 });
  }

  return (
    <button type="button" onClick={handleClick}>
      Do Something
    </button>
  );
}

Derive Fetcher init

Note: This depends on @remix-route/react.

Derive the value of the deprecated fetcher.type from the fetcher and navigation data.

import { getFetcherType } from "remix-utils/fetcher-type";

function Component() {
  let fetcher = useFetcher();
  let navigation = useNavigation();
  let fetcherType = getFetcherType(fetcher, navigation);
  useEffect(() => {
    if (fetcherType === "done") {
      // do something once the fetcher is done submitting the data
    }
  }, [fetcherType]);
}

You can also use the React Hook API which let's you avoid calling useNavigation.

import { useFetcherType } from "remix-utils/fetcher-type";

function Component() {
  let fetcher = useFetcher();
  let fetcherType = useFetcherType(fetcher);
  useEffect(() => {
    if (fetcherType === "done") {
      // do something once the fetcher is done submitting the data
    }
  }, [fetcherType]);
}

If you need to pass the fetcher type around, you can also import FetcherType type.

import { type FetcherType } from "remix-utils/fetcher-type";

function useCallbackOnDone(type: FetcherType, cb) {
  useEffect(() => {
    if (type === "done") cb();
  }, [type, cb]);
}

respondTo for Content Negotiation

If you're building a resource route and wants to send a different response based on what content type the client requested (e.g. send the same data as PDF or XML or JSON), you will need to implement content negotiation, this can be done with the respondTo header.

import { respondTo } from "remix-utils/respond-to";

export async function loader({ request }: LoaderArgs) {
  // do any work independent of the response type before respondTo
  let data = await getData(request);

  let headers = new Headers({ vary: "accept" });

  // Here we will decide how to respond to different content types
  return respondTo(request, {
    // The handler can be a subtype handler, in `text/html` html is the subtype
    html() {
      // We can call any function only really need to respond to this
      // content-type
      let body = ReactDOMServer.renderToString(<UI {...data} />);
      headers.append("content-type", "text/html");
      return new Response(body, { headers });
    },
    // It can also be a highly specific type
    async "application/rss+xml"() {
      // we can do more async work inside this code if needed
      let body = await generateRSSFeed(data);
      headers.append("content-type", "application/rss+xml");
      return new Response(body, { headers });
    },
    // Or a generic type
    async text() {
      // To respond to any text type, e.g. text/plain, text/csv, etc.
      let body = generatePlain(data);
      headers.append("content-type", "text/plain");
      return new Response(body, { headers });
    },
    // The default will be used if the accept header doesn't match any of the
    // other handlers
    default() {
      // Here we could have a default type of response, e.g. use json by
      // default, or we can return a 406 which means the server can't respond
      // with any of the requested content types
      return new Response("Not Acceptable", { status: 406 });
    },
  });
}

Now, the respondTo function will check the Accept header and call the correct handler, to know which one to call it will use the parseAcceptHeader function also exported from Remix Utils

import { parseAcceptHeader } from "remix-utils/parse-accept-header";

let parsed = parseAcceptHeader(
  "text/html, application/xhtml+xml, application/xml;q=0.9, image/webp, image/*, */*;q=0.8",
);

The result is an array with the type, subtype and extra params (e.g. the q value). The order will be the same encountered in the header, in the example aabove text/html will be the first, followed by application/xhtml+xml.

This means that the respondTo helper will prioritize any handler that match text/html, in our example above, that will be the html handler, but if we remove it then the text handler will be called instead.67

Form Honeypot

Note: This depends on react.

Honeypot is a simple technic to prevent spam bots from submitting forms, it works by adding a hidden field to the form that bots will fill, but humans won't.

There's a pair of utils in Remix Utils to help you implement this.

First, create a honeypot.server.ts where you will instantiate and configure your Honeypot.

import { Honeypot } from "remix-utils/honeypot";

// Create a new Honeypot instance, the values here are the defaults, you can
// customize them
export const honeypot = new Honeypot({
  randomizeNameFieldName: false,
  nameFieldName: "name__confirm",
  validFromFieldName: "from__confirm", // null to disable it
  encryptionSeed: undefined, // Ideally it should be unique even between processes
});

Then, in your app/root loader, call honeypot.getInputProps() and return it.

// app/root.tsx
import { honeypot } from "~/honeypot.server";

export async function loader() {
  // more code here
  return json({ honeypot: honeypot.getInputProps() });
}

And in the app/root component render the HoneypotProvider component wrapping the rest of the UI.

import { HoneypotProvider } from "remix-utils/honeypot-inputs";

export default function Component() {
  // more code here
  return (
    // some JSX
    <HoneypotProvider>
      <Outlet />
    </HoneypotProvider>
    // end that JSX
  );
}

Now, in every public form you want protect against spam (like a login form), render the HoneypotInputs component.

import { HoneypotInputs } from "remix-utils/honeypot-inputs";

function SomePublicForm() {
  return (
    <Form method="post">
      <HoneypotInputs label="Please leave this field blank" />
      {/* more inputs and some buttons */}
    </Form>
  );
}

Note: The label value above is the default one, use it to allow the label to be localized, or remove it if you don't want to change it.

Finally, in the action the form submits to, you can call honeypot.check.

import { SpamError } from "remix-utils/honeypot";
import { honeypot } from "~/honeypot.server";

export async function action({ request }) {
  let formData = await request.formData();
  try {
    honeypot.check(formData);
  } catch (error) {
    if (error instanceof SpamError) {
      // handle spam requests here
    }
    // handle any other possible error here, e.g. re-throw since nothing else
    // should be thrown
  }
  // the rest of your action
}

Author

• Sergio Xalambrí

License

• MIT License