When building AWS Lambda functions, managing concerns like input validation, error handling, and logging can quickly become repetitive and cumbersome. A middleware-based workflow simplifies this by promoting code reuse and separation of concerns, making Lambdas more maintainable and scalable.

Using TypeScript for AWS Lambda provides type safety, reducing runtime errors and improving developer productivity with better tooling and auto-completion. It ensures that your code is more robust, self-documenting, and easier to refactor as your application grows.

One crucial aspect of a middleware-based workflow is schema validation. By validating incoming requests before processing them, we can prevent unexpected errors, enforce data integrity, and improve security. Leveraging @raminyavari/ajv-ts-schema with @middy, we can seamlessly integrate schema validation into our Lambda middleware, ensuring type safety and runtime validation work hand in hand.

In this article, we’ll explore how to implement a middleware-based AWS Lambda in TypeScript, combining @raminyavari/ajv-ts-schema for schema validation and @middy for middleware management. Let’s dive in!

We'll explore three approaches to building AWS Lambda functions in TypeScript:

• A plain Lambda function with no middleware.
• A Lambda function enhanced with @middy for middleware support.
• A fully optimized Lambda function using @middy with @raminyavari/ajv-ts-schema for robust schema validation.

This progression will highlight the benefits of middleware and schema validation in making Lambdas more maintainable, scalable, and secure.

Plain Lambda Function (No Middleware)

In our sample Lambda function, the input is a user object containing basic user information and a list of orders. Since the event payload can be unpredictable, we must carefully validate it and handle errors appropriately—returning a 400 Bad Request for invalid input or a 500 Internal Server Error for unexpected failures.

Currently, both validation and error handling are tightly coupled within the Lambda function itself, making the code more complex and harder to maintain.

Our goal is to separate these concerns using middleware, allowing us to shrink the handler to just a few lines—focused solely on business logic—while offloading validation and error handling to reusable components.

Let’s start with the first scenario. Below is the Lambda function:

export const handler = async (event: APIGatewayEvent): Promise<APIGatewayProxyResult> => {
  try {
    // Parse and validate the event body
    const body = event.body ? JSON.parse(event.body) : {};
    const validatedEvent = validateEvent(body);
    
    const result = await processRequest(validatedEvent);

    return {
      statusCode: 200,
      body: result,
    };
  } catch (error) {
    if (error instanceof ValidationError) {
      return {
        statusCode: 400,
        body: JSON.stringify({ error: error.message }),
      };
    }

    return {
      statusCode: 500,
      body: JSON.stringify({ error: "Internal server error" }),
    };
  }
};

Next, we define the types:

interface Order {
  productId: string;
  quantity: number;
}

interface Address {
  street: string;
  city: string;
  postalCode: string;
}

interface UserEvent {
  userId: string;
  email: string; // Should verify it is an email
  age: number; // Should be an integer between 18 and 100
  address: Address;
  orders: Order[];
}

Finally, we define a function along with a helper class to handle the validation:

// Custom error class for validation errors
class ValidationError extends Error {
  constructor(message: string) {
    super(message);
    this.name = "ValidationError";
  }
}

// Schema validation function
function validateEvent(event: any): UserEvent {
  if (typeof event !== "object" || event === null) {
    throw new ValidationError("Event must be a valid JSON object.");
  }

  if (typeof event.userId !== "string" || !event.userId.trim()) {
    throw new ValidationError("userId must be a non-empty string.");
  }

  if (typeof event.email !== "string" || !/^[^\s@]+@[^\s@]+\.[^\s@]+$/.test(event.email)) {
    throw new ValidationError("email must be a valid email address.");
  }

  if (typeof event.age !== "number" || event.age < 18 || event.age > 100) {
    throw new ValidationError("age must be a number between 18 and 100.");
  }

  if (
    typeof event.address !== "object" ||
    typeof event.address.street !== "string" ||
    typeof event.address.city !== "string" ||
    typeof event.address.postalCode !== "string"
  ) {
    throw new ValidationError("address must contain valid street, city, and postalCode.");
  }

  if (!Array.isArray(event.orders) || event.orders.length === 0) {
    throw new ValidationError("orders must be a non-empty array.");
  }

  for (const order of event.orders) {
    if (typeof order.productId !== "string" || !order.productId.trim()) {
      throw new ValidationError("Each order must have a valid productId.");
    }
    if (typeof order.quantity !== "number" || order.quantity <= 0) {
      throw new ValidationError("Each order must have a quantity greater than 0.");
    }
  }

  return event as UserEvent;
}

The drawbacks of this approach are:

Manual Schema Validation Is Tedious and Error-Prone

• Writing and maintaining custom validation logic is cumbersome.
• Each validation rule is hard-coded, making it difficult to scale and extend.
• If the schema changes (e.g., adding new fields), the validation logic must be manually updated.

Error Messages Are Not Standardized

• The validation errors are simple strings, which makes debugging harder.
• There's no structured way to return multiple validation errors at once.

Inconsistent Type Safety

• The function relies on runtime checks, but there's no strong enforcement of the UserEvent type.
• The validation function casts event as UserEvent, which is unsafe if validation misses something.

No Re-usability of Validation Logic

• If other Lambdas need similar validation, the same logic must be duplicated.
• There's no way to reuse the validation in other contexts (e.g., event-driven Lambda functions, API gateways).

Error Handling Could Be Improved

• The ValidationError class is custom but doesn’t provide detailed context.
• All validation failures result in a 400 response, but finer-grained error responses (e.g., missing vs. invalid fields) could be helpful.

Lambda Function Enhanced with @middy

Middy is a lightweight, middleware engine for AWS Lambda functions. It allows you to easily add reusable logic to Lambda functions without modifying the core business logic. You can apply middleware to handle things like input validation, error handling, logging, or request transformation. Middy follows the middleware pattern, making it simple to compose multiple functionalities in a clean and modular way, enhancing Lambda functions' flexibility and maintainability.

Middy is particularly useful for improving Lambda performance by separating concerns and ensuring consistent behavior across multiple functions. It is designed for Lambda with API Gateway, assuming middleware compliance with API calls that include a body. First, we'll examine it in this context before exploring its use with AppSync and GraphQL.

Now, let's see how the code looks when we use @middy.

As shown in the snippet below, the Lambda function becomes much simpler, focusing solely on the core business logic while offloading other concerns like validation and error handling to middleware.

const baseHandler = async async (event: UserEvent) => {
  return await processRequest(event);
};

export const handler = middify({ handler: baseHandler, schema: userSchema});

With this approach, each Lambda function becomes as simple as shown above. We only need to implement a middify function once, which takes a Lambda function and its corresponding schema, and returns an enhanced version of the Lambda function with validation and error handling handled automatically.

Let’s take a look at the implementation:

import middy from '@middy/core';
import jsonBodyParser from '@middy/http-json-body-parser';
import httpErrorHandler from '@middy/http-error-handler';
import validator from '@middy/validator';
import { transpileSchema } from '@middy/validator/transpile';

export const middify = ({ handler, schema }: { handler: unknown; schema: any }) => {
  const handlerWrapper = async (event: unknown, context: unknown) => {
    const result = await handler(event, context);

    return {
      statusCode: 200,
      body: typeof result === "string" ? result : JSON.stringify(result),
    };
  };

  return middy()
    .use(jsonBodyParser())
    .use(validator({ eventSchema: transpileSchema(schema) }))
    .use(httpErrorHandler())
    .handler(handlerWrapper);
};

When using @middy for schema validation, the schema must conform to the AJV standard (Learn more about AJV).

Based on our example and the AJV specifications, the schema would look like this:

const userSchema = {
  type: "object",
  required: ["userId", "email", "age", "address", "orders"],
  properties: {
    userId: { type: "string", minLength: 1 },
    email: { type: "string", format: "email" },
    age: { type: "integer", minimum: 18, maximum: 100 },
    address: {
      type: "object",
      required: ["street", "city", "postalCode"],
      properties: {
        street: { type: "string" },
        city: { type: "string" },
        postalCode: { type: "string" },
      },
    },
    orders: {
      type: "array",
      minItems: 1,
      items: {
        type: "object",
        required: ["productId", "quantity"],
        properties: {
          productId: { type: "string", minLength: 1 },
          quantity: { type: "number", minimum: 1 },
        },
      },
    },
  },
};

With this approach, we still need to define our types separately, as the AJV schema is simply a JSON object, not a type definition.

Therefore, we will define the types as follows:

interface UserEvent {
  // Properties are exactly the same as the first scenario
  // So, not repeating them here
}

This approach also has some drawbacks:

• We need to define both schema and types for each Lambda function.
• We must ensure that the schema and type definitions are always in sync. When one changes, the other must be updated, which can be cumbersome and error-prone.
• Defining AJV schema manually is not type-safe, and there’s a risk of introducing issues. While some packages allow you to define the schema in TypeScript using decorators, they can still be error-prone and require recompilation of the schema after every change.

To address these issues, we will use the @raminyavari/ajv-ts-schema package, which provides a more efficient solution.

Optimized Lambda Function with @raminyavari/ajv-ts-schema

By using @raminyavari/ajv-ts-schema, the schema and type definition are combined. This approach offers several advantages:

• Combines schema and type definition into one source of truth.
• Fully typed and error-proof.
• No need for compilation; the schema can be defined and used immediately.

Let’s take a look at the handler:

import { type AjvJsonSchema } from '@raminyavari/ajv-ts-schema';

const baseHandler: LambdaHandler<AjvJsonSchema<UserSchema>, unknown> = async (event) => {
  return await processRequest(event);
};

export const handler = middify({ handler: addTenant, schema: UserSchema });

To improve type safety, we've define a new type as follows:

type LambdaHandler<Event = unknown, Context = unknown> = (
  event: Event,
  context: Context
) => Promise<string | object | unknown>;

The middify function remains largely the same, with a small change: the type of the schema has been updated from any to typeof AjvSchema. All schemas must now extend the AjvSchema class. With this update, the middify function looks like this:

import middy from '@middy/core';
import jsonBodyParser from '@middy/http-json-body-parser';
import httpErrorHandler from '@middy/http-error-handler';
import validator from '@middy/validator';
import { transpileSchema } from '@middy/validator/transpile';
import { AjvSchema } from '@raminyavari/ajv-ts-schema';

export const middify = ({ handler, schema }: { 
	handler: unknown; 
	schema: typeof AjvSchema 
}) => {
	const ajvSchema = schema.getSchema();

  const handlerWrapper = async (event: unknown, context: unknown) => {
    const result = await (handler as LambdaHandler)(event, context);

    return {
      statusCode: 200,
      body: typeof result === "string" ? result : JSON.stringify(result),
    };
  };

  return middy()
    .use(jsonBodyParser())
    .use(validator({ eventSchema: transpileSchema(ajvSchema) }))
    .use(httpErrorHandler())
    .handler(handlerWrapper);
};

The most powerful part of this approach is that we can define both the schema and the event type simultaneously.

import {
	AjvSchema,
	AjvObject,
	AjvProperty,
} from '@raminyavari/ajv-ts-schema';

@AjvObject()
class Order extends AjvSchema {
  @AjvProperty({ type: "string", minLength: 1, required: true })
  productId!: string;

  @AjvProperty({ type: "number", minimum: 1, required: true })
  quantity!: number;
}

@AjvObject()
class Address extends AjvSchema {
  @AjvProperty({ type: "string", required: true })
  street!: string;

  @AjvProperty({ type: "string", required: true })
  city!: string;

  @AjvProperty({ type: "string", required: true })
  postalCode!: string;
}

@AjvObject()
class UserEvent extends AjvSchema {
  @AjvProperty({ type: "string", minLength: 1, required: true })
  userId!: string;

  @AjvProperty({ type: "formatted-string", format: "email", required: true })
  email!: string;

  @AjvProperty({ type: "integer", minimum: 18, maximum: 100, required: true })
  age!: number;

  @AjvProperty(Address)
  address!: Address;

  @AjvProperty({ type: "array", items: Order, minItems: 1 })
  orders!: Order[];
}

With this approach, we gain the following benefits:

• The entire code is type-safe.
• Validation is seamless; defining the schema automatically ensures event validation. Validation errors and exceptions are handled by @middy within our middify function.
• We have a single source of truth for both schema and types.
• The type-safe decorators eliminate the risk of mistakes when defining the schema, while also providing the added benefit of autocomplete.
• There's no need for schema compilation—changes are immediately reflected and can be used right away.

Considerations for AppSync and GraphQL

When our Lambda is invoked by GraphQL through AppSync, everything remains the same except for the middify function.

In this case, two adjustments are needed:

• The jsonBodyParser middleware assumes the event contains a body, which isn’t the case with AppSync. We need to create a custom middleware.
• The httpErrorHandler middleware returns HTTP errors, whereas GraphQL expects a different error format.

With these changes, our function will look like this:

import middy from '@middy/core';
import jsonBodyParser from '@middy/http-json-body-parser';
import httpErrorHandler from '@middy/http-error-handler';
import validator from '@middy/validator';
import { transpileSchema } from '@middy/validator/transpile';
import { AjvSchema } from '@raminyavari/ajv-ts-schema';

export const middify = ({ handler, schema }: { 
	handler: unknown; 
	schema: typeof AjvSchema 
}) => {
	const ajvSchema = schema.getSchema();

  const handlerWrapper = async (event: unknown, context: unknown) => {
    const result = await (handler as LambdaHandler)(event, context);

    return {
      statusCode: 200,
      body: typeof result === "string" ? result : JSON.stringify(result),
    };
  };

  return middy()
    .use(eventParser())
    .use(validator({ eventSchema: transpileSchema(ajvSchema) }))
    .use(errorHandler())
    .handler(handlerWrapper);
};

A basic implementation of eventParser:

const eventParser = () => {
  return {
    before: async (request: any) => {
      request.event = parseJson(request.event);
    },
  };
};

const parseJson = (value: string | object) => {
  try {
    return typeof value === "string" ? JSON.parse(value) : value;
  } catch {
    return value;
  }
};

Finally, the errorHandler middleware:

import { normalizeHttpResponse } from "@middy/util";

const errorHandler = () => ({
  onError: async (request: any) => {
    if (request.response !== undefined) return;

    normalizeHttpResponse(request);
    const { statusCode, name } = request.error;
    const message = (request.error.message as string) || "Internal Server Error";

    const error = {
      message,
      extensions: {
        exception: { code: name },
        code: name,
      },
    };

    request.response = {
      ...request.response,
      statusCode,
      body: JSON.stringify({
        errors: [error],
      }),
    };

    request.error = {
      ...request.error,
      message: undefined,
      expose: true,
    };
  },
});

Learn More About @raminyavari/ajv-ts-schema

• 📖 Documentation
• 💻 GitHub Repository
• 📦 NPM Package