node package manager
Share your code. npm Orgs help your team discover, share, and reuse code. Create a free org »


Travis Build Status npm version


Typescript ORM of DynamoDB, written from scratch to fully support DynamoDB. Powering Vingle


  1. Serialize / Deserialize DynamoDB record -> TS class object based on annotations.
  2. Table Configurations
    • CreateTable
      • Create secondary indexes (Both local / global)
      • Configure TTL
    • DropTable
  3. PrimaryKey
    • FullPrimaryKey (Hash, Range)
    • HashPrimaryKey (Hash)
  4. Indexes
    • Local, both hash and range key
    • Global, both hash and range key
  5. Attribute
    • Type Support (Number / String / Boolean / Array / Object / Buffer)
    • TimeToLive
  6. DAX Support
  7. Optimized aws-sdk usage
    • aws-sdk has a serious problem of not reusing HTTP connection towards DynamoDB by default. check this issue this could cause unbearable latency sometimes with showing > 100ms. this is more of an issue of NodeJS HTTP module but nevertheless, this is optimized as a default. Code
  8. AWS X-Ray support   - X-Ray is serverless distributed tracing service. In order to log DynamoDB transaction into it, you also need to some sort of risk monkey-patching. Here you can turn it on by setting process.env.ENABLE_XRAY = "true"
  9. Testing
    • You can change the endpoint of DynamoDB by setting the environment variable or setting new connection, So you can install local-dynamo locally at setup endpoint to local. refer package.json for the detailed how-to

Also, dynamo-types let you overcome several limits that DynamoDB or the aws-sdk has.

  1. BatchWrite (batchDelete / batchPut) has a limit of a maximum of 25 items per request.
    • dynamo-typeorm automatically splits given items into chunks of 25 and sends requests in parallel
  2. BatchGet has a limit of a maximum of 100 items per requests
    • dynamo-typeorm automatically splits given keys to chunks of 100 and sends requests in parallel
  3. BatchGet doesn't keep the order of items as it is in input keys,
    • dynamo-typeorm sort return items based on input keys
  4. BatchGet doesn't handle "missing items".
    • dynamo-typeorm has "BatchGet" / "BatchGetFull"
      • BatchGet
        order items follow to keys, missing items are just missing. return type Promise<Array>
        so keys.legnth !== items.keys in this case
      • BatchGetFull
        order items follow to keys, fill missing items with "null". return type Promise<Array<Item | null>>
        so keys.length === items.keys always true

And most importantly, all of those queries regardless of whether it's from index or primary key, strongly typed. I mean what's the point of using typescript if not anyway?


  @Decorator.Table({ name: "prod-Card" })
  class Card extends Table {
    public id: number;
    public title: string;
    @Decorator.Attribute({ timeToLive: true })
    public expiresAt: number;
    @Decorator.FullPrimaryKey('id', 'title')
    static readonly primaryKey: Query.FullPrimaryKey<Card, number, string>;
    static readonly writer: Query.Writer<Card>;
  // Create Table At DynamoDB
  await Card.createTable();
  // Drop Table At DynamoDB
  await Card.dropTable();
  // Creating Record
  const card = new Card(); = 100;
  card.title = "Title";
  await Card.writer.put(card);
  // OR just
  // Batch Put
  await Card.writer.batchPut([
    new Card(),
    new Card()
  // Get Record
  await Card.primaryKey.get(100, "Title");
  // BatchGet
  // This array is strongly typed such as Array<[number, string]> so don't worry.
  await Card.primaryKey.batchGet([
    [100, "Title"],
    [200, "Title2"]
  // Query
  // Range key opreations are stringly typed. ([">=", T] | ["=", T] ...)
  await Card.primaryKey.query({
    hash: 100,
    range: [">=", "Title"]
  // Delete record
  await card.delete()
import {
} from "dynamo-types";
@Decorator.Table({ name: `table_name` })
export class CardStat extends Table {
  public static readonly primaryKey: Query.HashPrimaryKey<CardStat, number>;
  public static readonly writer: Query.Writer<CardStat>;
  @Decorator.Attribute({ name: "card_id" })
  public cardId: number;
  @Decorator.Attribute({ name: "impressions_count" })
  public impressionsCount: number = 0;
  @Decorator.Attribute({ name: "shares" })
  public shares: number = 0;


DynamoDB supports 2 different kinds of connections. Plain connections to DynamoDB through HTTP, or through DAX. dynamo-typeorm supports this by letting you create a separate connection for each table.

@Decorator.Table({ name: "prod-Card1", connection: new DAXConnection({ endpoints: ["dax-domain:8892"] }) })
class Card extends Table {
  public id: number;
  public title: string;
  @AttributeDecorator({ name: "complicated_field"})
  public complicatedField: string;
  @FullPrimaryKeyDecorator('id', 'title')
  static readonly primaryKey: Query.FullPrimaryKey<Card, number, string>;
  static readonly writer: Query.Writer<Card>;

Then any query that is sent to the Card table will be sent through DAXConnection.

If you don't specify any connection, it automatically uses default connection, which is DynamoDBConnection.