@nikovirtala/typesafe-dynamodb
v0.0.64
Published
[](https://badge.fury.io/js/@nikovirtala%2Ftypesafe-dynamodb)
Readme
@nikovirtala/typesafe-dynamodb
@nikovirtala/typesafe-dynamodb is a fork of typesafe-dynamodb (a type-only library which replaces the type signatures of the AWS SDK v3's DynamoDB client) with schema validation based on zod.
It substitutes getItem, putItem, deleteItem and query API methods with type-safe and schema validated alternatives that are aware of the data in your tables and also adaptive to the semantics of the API request, e.g. by validating ExpressionAttributeNames and ExpressionAttributeValues contain all the values used in a ConditionExpression string, or by understanding the effect of a ProjectionExpression on the returned data type.
The end goal is to provide types and validation that have total understanding of the AWS DynamoDB API and enable full utilization of the TypeScript type system for modeling complex DynamoDB Tables, such as the application of union types and template string literals for single-table designs without forgetting runtime safety.

Installation
npm install --save-dev @nikovirtala/typesafe-dynamodbUsage
This library contains type definitions for AWS SDK v3 (@aws-sdk/client-dynamodb).
AWS SDK v3
Option 1 - DynamoDB
DynamoDB is a convenient way of using the DynamoDB API, except it is not optimized for tree-shaking (for that, see Option 2).
Start by declaring a standard TypeScript interface which describes the structure of data in your DynamoDB Table:
interface Record {
key: string;
sort: number;
attribute: string;
// all types are allowed, such as recursive nested types
records?: Record[];
}Then, cast the DynamoDB client instance to TypeSafeDynamoDBv3:
import { DynamoDB } from "@aws-sdk/client-dynamodb";
import { TypeSafeDynamoDBv3 } from "typesafe-dynamodb/lib/client-v3";
const client = new DynamoDB({..});
const typesafeClient: TypeSafeDynamoDBv3<Record, "key", "sort"> = client;"key" is the name of the Hash Key attribute, and "sort" is the name of the Range Key attribute.
Finally, use the client as you normally would, except now with intelligent type hints and validations.
Option 2 - DynamoDBClient (a Command-Response interface optimized for tree-shaking)
DynamoDBClient is a generic interface with a single method, send. To invoke an API, call send with an instance of the API's corresponding Command.
This option is designed for optimal tree-shaking when bundling code, ensuring that the bundle only includes code for the APIs your application uses.
For this option, type-safety is achieved by declaring your own commands and then calling the standard DynamoDBClient:
interface MyType {
key: string;
sort: number;
list: string[];
}
const client = new DynamoDBClient({});
const GetMyTypeCommand = TypeSafeGetItemCommand<MyType, "key", "sort">();
await client.send(
new GetMyTypeCommand({
..
})
);Document Client
The AWS SDK v3 provides a javascript-friendly interface called the DocumentClient. Instead of using the AttributeValue format, such as { S: "hello" } or { N: "123" }, the DocumentClient enables you to use native javascript types, e.g. "hello" or 123.
When defining your Command types, use the corresponding TypeSafe*DocumentCommand type, for example TypeSafeGetDocumentCommand instead of TypeSafeGetItemCommand:
- GetItem -
TypeSafeGetDocumentCommand - PutItem -
TypeSafePutDocumentCommand - DeleteItem -
TypeSafeDeleteDocumentCommand - UpdateItem -
TypeSafeUpdateDocumentCommand - Query -
TypeSafeQueryDocumentCommand - Scan -
TypeSafeScanDocumentCommand
See: https://docs.aws.amazon.com/AWSJavaScriptSDK/v3/latest/modules/_aws_sdk_lib_dynamodb.html
import { JsonFormat } from "typesafe-dynamodb";
import { TypeSafeGetDocumentCommand } from "typesafe-dynamodb/lib/get-document-command";
const MyGetItemCommand = TypeSafeGetDocumentCommand<MyType, "key", "sort">();For the SDK V3 client, cast it to TypeSafeDocumentClientV3:
import { DynamoDBClient } from "@aws-sdk/client-dynamodb";
import { DynamoDBDocumentClient, PutCommand } from "@aws-sdk/lib-dynamodb";
import { TypeSafeDocumentClientV3 } from "typesafe-dynamodb/lib/document-client-v3";
const client = new DynamoDBClient({});
const docClient = DynamoDBDocumentClient.from(
client,
) as TypeSafeDocumentClientV3<MyType, "key", "sort">;And then call .send with an instance of your TypeSafe Command:
docClient.send(new MyGetItemCommand({
..
}));Features
Type-aware Input and Output
The type of the Key is derived from the Record type.

Same for the Item in the response:

Single Table Design
Below are two interface declarations, representing two types of data stored in a single DynamoDB Table - User and Order. Single table design in DynamoDB is achieved by creating "composite keys", e.g. USER#${UserID}. In TypeScript, we use template literal types to encode this in the Type System.
interface User<UserID extends string = string> {
PK: `USER#${UserID}`;
SK: `#PROFILE#${UserID}`;
Username: string;
FullName: string;
Email: string;
CreatedAt: Date;
Address: string;
}
interface Order<
UserID extends string = string,
OrderID extends string = string,
> {
PK: `USER#${UserID}`;
SK: `ORDER#${OrderID}`;
Username: string;
OrderID: string;
Status: "PLACED" | "SHIPPED";
CreatedAt: Date;
Address: string;
}With these two types defined, you can now use a union type to declare a TypeSafeDynamoDB instance aware of the two types of data in your tables:
const client: TypeSafeDynamoDB<User | Order, "PK", "SK">;When making calls such as getItem, TypeScript will narrow the returned data type to the corresponding type based on the structure of the Key in your request:

Type-Safe DynamoDBStreamEvent
Leverage your data types in Lambda Functions attached to the DynamoDB Table Stream:
import { DynamoDBStreamEvent } from "typesafe-dynamodb/lib/stream-event";
export async function handle(
event: DynamoDBStreamEvent<User | Order, "PK", "SK", "KEYS_ONLY">
) {
..
}The event's type is derived from the data type and the the StreamViewType, e.g. "NEW_IMAGE" | "OLD_IMAGE" | "KEYS_ONLY" | "NEW_AND_OLD_IMAGES".
Schema-Validated DynamoDBStreamEvent
Validate DynamoDB stream events at runtime using Zod schemas:
import { z } from "zod";
import { validateStreamEvent } from "typesafe-dynamodb";
import type { DynamoDBStreamEvent } from "typesafe-dynamodb/lib/stream-event";
const UserSchema = z.object({
PK: z.string(),
SK: z.string(),
name: z.string(),
email: z.string().email(),
});
type User = z.infer<typeof UserSchema>;
export async function handle(
event: DynamoDBStreamEvent<User, "PK", "SK", "NEW_AND_OLD_IMAGES">,
) {
try {
const validatedEvent = validateStreamEvent(event, UserSchema);
// Process validated stream records
} catch (error) {
if (error instanceof z.ZodError) {
console.error("Schema validation failed:", error.issues);
}
throw error;
}
}The validateStreamEvent function validates both NewImage and OldImage data against your Zod schema, ensuring runtime type safety for your stream processing logic.
Filter result with ProjectionExpression
The ProjectionExpression field is parsed and applied to filter the returned type of getItem and query.

Filter with AttributesToGet
If you specify AttributesToGet, then the returned type only contains those properties.

Validate ExpressionAttributeNames and ExpressionAttributeValues
If you add a ConditionExpression in putItem, you will be prompted for any #name or :valu placeholders:

Same is true for a query:

Marshall a JS Object to an AttributeMap
A better type definition @aws-sdk/util-dynamodb's marshall function is provided which maintains the typing information of the value passed to it and also adapts the output type based on the input marshallOptions
Given an object literal:
const myObject = {
key: "key",
sort: 123,
binary: new Uint16Array([1]),
buffer: Buffer.from("buffer", "utf8"),
optional: 456,
list: ["hello", "world"],
record: {
key: "nested key",
sort: 789,
},
} as const;Call the marshall function to convert it to an AttributeMap which maintains the exact structure in the type system:
import { marshall } from "typesafe-dynamodb/lib/marshall";
// marshall the above JS object to its corresponding AttributeMap
const marshalled = marshall(myObject)
// typing information is carried across exactly, including literal types
const marshalled: {
readonly key: S<"key">;
readonly sort: N<123>;
readonly binary: B;
readonly buffer: B;
readonly optional: N<456>;
readonly list: L<readonly [S<"hello">, S<"world">]>;
readonly record: M<...>;
}Unmarshall an AttributeMap back to a JS Object
A better type definition @aws-sdk/util-dynamodb's unmarshall function is provided which maintains the typing information of the value passed to it and also adapts the output type based on the input unmarshallOptions.
import { unmarshall } from "typesafe-dynamodb/lib/marshall";
// unmarshall the AttributeMap back into the original object
const unmarshalled = unmarshall(marshalled);
// it maintains the literal typing information (as much as possible)
const unmarshalled: {
readonly key: "key";
readonly sort: 123;
readonly binary: NativeBinaryAttribute;
readonly buffer: NativeBinaryAttribute;
readonly optional: 456;
readonly list: readonly [...];
readonly record: Unmarshall<...>;
}If you specify {wrapNumbers: true}, then all number types will be wrapped as { value: string }:
const unmarshalled = unmarshall(marshalled, {
wrapNumbers: true,
});
// numbers are wrapped in { value: string } because of `wrappedNumbers: true`
unmarshalled.sort.value; // string
// it maintains the literal typing information (as much as possible)
const unmarshalled: {
readonly key: "key";
// notice how the number is wrapped in the `NumberValue` type?
// this is because `wrapNumbers: true`
readonly sort: NumberValue<123>;
readonly binary: NativeBinaryAttribute;
readonly buffer: NativeBinaryAttribute;
readonly optional: NumberValue<...>;
readonly list: readonly [...];
readonly record: Unmarshall<...>;
};