@swastikbhattacharyya/next-actions - v0.3.2
    Preparing search index...

    @swastikbhattacharyya/next-actions - v0.3.2

    Next Actions

    Next Actions is a utility library for defining type-safe server actions and data access layers.

    Read Full Documentation

    Next Actions is a lightweight and simple TypeScript utility library for building type-safe, validated server actions and data access layers using zod along with clearly defined and structured error responses.

    Define an input schema with Zod, add layered validation functions that return typed error codes with payloads, or context to be used later in the action execution. The action itself can also return typed errors or a success payload. Execution gives you a strongly typed result—success or failure, with full type safety.

    export const action = new Action()
    .setInputSchema(
    z.object({
    id: z.string(),
    name: z.string().min(2),
    }),
    )
    .addValidator(
    async ({ context: { inputSchema }, params }) =>
    zodValidator(inputSchema, params), // Validator to safely parse arguments by client/form
    )
    .setActionFn(async () => {
    // Your Server Action
    // ...
    return { success: true, message: "Action successful" };
    });

    The following example demonstrates how to create a server action with input validation with zod, user session validation and user role check. Assume the action inserts a blog post into the database.

    This schema defines the parameters of the server action.

    Note: Schema validation is not automatic. You must use the zodValidator utility from the library, or define your own custom Zod Validator. This example will use zodValidator for input validation.

    const postSchema = z.object({
    title: z.string(),
    description: z.string(),
    content: z.string(),
    });

    Before we define the validator, let us know what a Validator actually is. Validators are just functions used to enforce rules before the action executes. The function must return a Promise of a ValidationResult. It can accept their own parameters and perform any logic needed.

    The ValidationResult type accepts two type arguments:

    1. Injected Context: Data that will be made available to all subsequent validators and the action function if the validation succeeds. For example: the user session validator will inject the user id into the execution so that the action can later pass the user id to the query to insert into the database.
    2. Possible Errors: An object containing all the error codes the validator might return, along with their associated payloads. It helps to distinguish which validator actually failed and what error occured. You can handle each failure case accordingly.

    Note: You can use shortcut types like ValidationResultWithoutContext, ValidationResultWithoutError, or ValidationResultAny if your validator doesn't inject context, doesn't return errors, or neither.

    This validator will inject the user id into the context and can return an error code userSessionValidator_noSession if not session was found and a string containing a message as payload for that error code (as an example, you can return any payload that might help with error handling).

    async function userSessionValidator(): Promise<
    ValidationResult<
    { userId: string },
    { userSessionValidator_noSession: { message: string } }
    >
    > {
    const cookieStore = await cookies();

    if (!cookieStore.has("userSession"))
    return {
    ok: false,
    errorCode: "userSessionValidator_noSession",
    payload: { message: "User session not found in cookies" },
    };

    const userId = cookieStore.get("userSession")?.value as string;
    return { ok: true, context: { userId } };
    }

    This validator will inject the user role into the context and can return two errors: userRoleValidator_roleNotFound and userRoleValidator_insufficientPermissions, but with no payload.

    async function userRoleValidator(userId: string): Promise<
    ValidationResult<
    { role: "admin" | "editor" },
    {
    userRoleValidator_roleNotFound: null;
    userRoleValidator_insufficientPermissions: null;
    }
    >
    > {
    const role = await getUserRole(userId); // Assume this fetches from DB or session

    if (!role) {
    return {
    ok: false,
    errorCode: "userRoleValidator_roleNotFound",
    payload: null,
    };
    }

    if (role !== "admin") {
    return {
    ok: false,
    errorCode: "userRoleValidator_insufficientPermissions",
    payload: null,
    };
    }

    return {
    ok: true,
    context: { role },
    };
    }

    The Action class uses the builder pattern to compose a server action step by step using the chainable methods .setInputSchema(), .addValidator(), and .setActionFn().

    Note: If using Next.js, both .addValidator() and .setActionFn() must receive async functions, since server functions must always be asynchronous.

    export const action = new Action()
    .setInputSchema(postSchema) // injects schema instance into context
    .addValidator(async ({ context: { inputSchema }, params }) =>
    zodValidator(inputSchema, params),
    )
    .addValidator(async () => userSessionValidator()) // injects userId into context
    .addValidator(async () => userRoleValidator()) // injects role into context
    .setActionFn(async ({ context, params }) => {
    // You can now access context.userId, context.role and the validated input in params
    // context also contains the input schema instance if needed. It is injected by .setInputSchema()
    return { success: true, message: "Successfully executed" };
    });

    You can now define the actual server action to perform after the validation steps are successful.

    export const action = new Action()
    .setInputSchema(postSchema)
    .addValidator(async ({ context: { inputSchema }, params }) =>
    zodValidator(inputSchema, params),
    )
    .addValidator(async () => userSessionValidator())
    .addValidator(async () => userRoleValidator())
    .setActionFn(async ({ context, params }) => {
    const { userId } = context;
    const { title, description, content } = params;

    await db.query(
    `INSERT INTO posts (user_id, title, description, content)
    VALUES (?, ?, ?, ?)`,
    [userId, title, description, content],
    );

    return { success: true, message: "Post created successfully" };
    });

    An action is conceptually similar to a validator — it returns a type called ActionResult, which is structurally the same as ValidationResult. That means an action can either:

    • Succeed, and optionally return a success payload (e.g. { id: string })

    • Fail, and return a structured error with a code and optional payload You can define the types for success and error codes using .setActionFn<SuccessPayload, ErrorMap>().

      .setActionFn<{ postId: string }, { dbError: { reason: string } }>(
      async ({ context, params }) => {
      // If Error case
      return {
      success: false,
      message: "Failed",
      errorCode: "dbError",
      errorPayload: { reason: "Some Reason" },
      };
      // If success case
      return {
      success: true,
      message: "Successfully executed",
      payload: { postId: "some Id" },
      };
      }

    Just like validators, action logic can be extracted into reusable functions. Your function should return an ActionResult, and you can use any of the shortcut types:

    • ActionResultWithoutPayload – for actions with no success payload
    • ActionResultWithoutError – for actions that never fail
    • ActionResultAny – for actions with no payloads or errors

    For example:

    async function createPost(
    userId: string,
    input: PostInput
    ): ActionResult<{ postId: string }, { dbError: null }> {...}
    .setActionFn(async ({context, params}) => createPost(context.userId, params));
    

    Not always. You don’t need to extract validators into separate functions unless you want to reuse them across multiple actions. Validators can be written inline directly inside .addValidator(...), and still define their own context and error map via the return type:

    .addValidator<{ role: string }, { invalidRole: null }>(async (): => {...})
    

    You can invoke an action just like any async function by calling it with the required input object. The returned result is strongly typed and will indicate whether the action succeeded or failed.

    Here’s an example of using the toast.promise pattern (e.g. from react-hot-toast) to handle the result with loading, success, and error states while taking advantage of the typed error codes and payloads:

    const promise = action({
    title: "Post Title",
    description: "Post Description",
    content: "Post Content",
    });

    toast.promise(
    async () => {
    const res = await promise;

    if (!res.success) {
    switch (res.errorCode) {
    case "dbError":
    throw new Error(res.errorPayload.reason);
    case "zodValidator_invalidParams":
    throw new Error(res.errorPayload.issues[0].message);
    }
    }

    return res;
    },
    {
    loading: "Submitting",
    success: (res) => res.message,
    error: (res: Error) => res.message,
    },
    );
    Condition Structure
    Success (no payload) { success: true, message: string }
    Success (with payload) { success: true, message: string, payload: SuccessPayload }
    Failure (from error map) { success: false, message: string, errorCode: keyof ErrorMap, errorPayload: ErrorMap[errorCode] }
    ActionResult<SuccessPayload, ErrorMap>
    Condition Structure
    Success (no context) { ok: true }
    Success (with context) { ok: true, context: OutputContext }
    Failure (from error map) { ok: false, errorCode: keyof ErrorMap, payload: ErrorMap[errorCode] }
    ValidationResult<OutputContext, ErrorMap>