d1-batch

Improve the way you batch operations with Cloudflare D1.

The Batch class provides a way to assign keys to a batch of queries, run the queries, and subsequently fetch the results by referring to the keys. Queries are run in the same order as they are added to the Batch class.

For example, the following piece of code is a simplified invitiation flow. A user row is created if needed and a message is recorded when provided. Using Cloudflare D1's built-in batch() function:

var _insertUserResult, newUserResult, _insertMessageResult;
if (message) {
  [_insertUserResult, newUserResult, _insertMessageResult] = await d1.batch([
    d1.prepare("INSERT INTO users (name, email) VALUES (?, ?) ON CONFLICT (email) DO NOTHING").bind(name, email),
    d1.prepare("SELECT * FROM users WHERE email=?").bind(email),
    d1.prepare("INSERT INTO messages (sender, receiver, content) VALUES (?, LAST_INSERT_ROWID(), ?)").bind(from!.id, message),
  ]);
} else {
  [_insertUserResult, newUserResult, _insertMessageResult] = await d1.batch([
    d1.prepare("INSERT INTO users (name, email) VALUES (?, ?) ON CONFLICT (email) DO NOTHING").bind(name, email),
    d1.prepare("SELECT * FROM users WHERE email=?").bind(email),
  ]);
}
const newUser = (newUserResult as D1Result<UsersRow>).results[0];

The code is however cleaner if you use the Batch class. The code is more readable, and easier to change over time:

import { Batch } from "d1-batch";

const b = new Batch(d1);
b.enqueue("_insertUser", d1.prepare("INSERT INTO users (name, email) VALUES (?, ?) ON CONFLICT (email) DO NOTHING").bind(name, email));
b.enqueue("newUser", d1.prepare("SELECT * FROM users WHERE email=?").bind(email));
if (message) {
  b.enqueue("_insertMessage", d1.prepare("INSERT INTO messages (sender, receiver, content) VALUES (?, LAST_INSERT_ROWID(), ?)").bind(from!.id, message));
}
await b.query();
const newUser = b.first<UsersRow>("newUser")!;

Installation

npm install d1-batch

Some additional context

In general, it is preferrable to use a lightweight query builder or ORM to perform CRUD operations: it's quicker to implement and less brittle. However, while working on a latency sensitive Cloudflare Workers based project, we needed to carefuly control the batching of queries. Our endpoints where making anywhere from a coulpe to a dozen queries to D1, but with never more than 3 round-trips -- resulting in backend endpoints that would run in about 300ms. We initially used the default batch() function, but eventually implemented the Batch class for better ergonomics.

The concept of a Batch class makes it easier to hook up automatic error handling, query logging, or performance monitoring when needed.