Skip to main content

Write to the Database

We are ready to add our new comments feature.

Scaffold business logicโ€‹

We'll start by scaffolding the domain code first. As mentioned in the last chapter, we'll add this to our core package.

Open up packages/core/src/article.ts and add the following two functions to the bottom of the file.

export function addComment(articleID: string, text: string) {
// code for adding a comment to an article

export function comments(articleID: string) {
// code for getting a list of comments of an article

Before we can implement them, we'll need to create a new table to store the comments.

Create a migrationโ€‹

Let's create a new migration for this.

Run this in the root of the project to create a new migration

npm run gen migration new

It'll ask you to name your migration. Type in comment.

? Migration name โ€บ comment

Once the migration is created, you should see the following in your terminal.

โœ” Migration name ยท comment

Loaded templates: _templates
added: packages/core/migrations/1661988563371_comment.mjs

Open up the new migration script and replace its content with:

import { Kysely } from "kysely";

* @param db {Kysely<any>}
export async function up(db) {
await db.schema
.addColumn("commentID", "text", (col) => col.primaryKey())
.addColumn("articleID", "text", (col) => col.notNull())
.addColumn("text", "text", (col) => col.notNull())

* @param db {Kysely<any>}
export async function down(db) {
await db.schema.dropTable("comment").execute();

This migration will create a new table called comment. While undoing the migration will drop the table.

Run a migrationโ€‹

Let's go ahead and run the migration.

Go to the RDS tab in SST Console and click Apply on our comment migration.

Console run migration

To verify that the table has been created; enter the following in the query editor, and hit Execute.

SELECT * FROM comment

Console query comments table

You should see 0 rows being returned.

Query the tableโ€‹

We are now ready to implement the addComment and comments functions.

Replace the two placeholder functions in packages/core/src/article.ts with:

export function addComment(articleID: string, text: string) {
return SQL.DB.insertInto("comment")
commentID: ulid(),

export function comments(articleID: string) {
return SQL.DB.selectFrom("comment")
.where("articleID", "=", articleID)

We are using Kysely to run typesafe queries against our database.

Behind the scenes

There are a couple of interesting details here, let's dig in:

  1. SQL.DB is the Kysely instance imported from packages/core/src/sql.ts.

    export const DB = new Kysely<Database>({
    dialect: new DataApiDialect({
    mode: "postgres",
    driver: {
    secretArn: RDS.db.secretArn,
    resourceArn: RDS.db.clusterArn,
    database: RDS.db.defaultDatabaseName,
    client: new RDSDataService(),
  2. RDS is coming from the SST Node client package.

    import { RDS } from "sst/node/rds";

    It has access to the config of our database, thanks to Resource Binding. You might recall us binding our database to the functions in our API back in the Project Structure chapter.

    function: {
    bind: [rds],

    By binding the rds cluster to our API in stacks/Api.ts, our API can access the database ARN (an ARN is an AWS identifier), database name, and ARN of the secret to access the database in our functions.

  3. The Kysely instance also needs a Database type. This is coming from packages/core/src/sql.generated.ts.

    export interface Database {
    article: Article;
    comment: Comment;

    The keys of this interface are the table names in our database. And they in turn point to other interfaces that list the column types of the respective tables. For example, here's the new Comment table we just created:

    export interface Comment {
    articleID: string;
    commentID: string;
    text: string;
  4. The sql.generated.ts types file, as you might've guessed in auto-generated. Our infrastructure code generates this when a new migration is run!

    It's defined in stacks/Database.ts.

    const rds = new RDS(stack, "rds", {
    engine: "postgresql11.13",
    migrations: "packages/core/migrations",
    types: "packages/core/src/sql.generated.ts",
    defaultDatabaseName: "main",

    Even though this file is auto-generated, you should check it into Git. We'll be relying on it later on in this tutorial.

Now with our business logic and database queries implemented, we are ready to hook up our API.