Database Seeds
Seeds are data that are required in order for your app to function. Think of the data a new developer would need to get up and running with your codebase, or data that needs to exist when a new instance of your application is deployed to a new environment.
Seed data are things like:
- An admin user so that you can log in to your new instance
- A list of categories that can be assigned to a Product
- Lists of roles and permissions
Seed data is not meant for:
- Sample data to be used in development
- Data to run tests against
- Randomized data
Best Practices
Ideally seed data should be idempotent: you can execute the seed script against your database at any time and end up with the seed data properly populated in the database. It should not result in wiping out existing records or creating duplicates of any seeded data that is already present.
Making your seeds idempotent requires more code than just a straight
createMany()
. The code examples below use the safest idempotent strategy
by having an upsert
check if a record with the same unique identifier
already exists, and if so just update it, if not then create it. But, this
technique requires a separate SQL statement for each member of your data array
and is less performant than createMany()
.
You could also do a check if any data exists in the database first, and if
not, create the records with createMany()
. However, this means that any
existing seed data that may have been modified will remain, and would not be
updated to match what you expect in your seed.
When in doubt, upsert
!
When seeds run
Seeds are automatically run the first time you migrate your database:
yarn rw prisma migrate dev
They are run every time you reset your database:
yarn rw prisma migrate reset
You can manually run seeds at any time with the following command:
yarn rw prisma db seed
You generally don't need to keep invoking your seeds over and over again, so it
makes sense that Prisma only does it on a complete database reset, or when the
database is created with the first prisma migrate dev
execution. But as your
schema evolves you may add a new model that requires some seeded data and so
you can add it to your seed file and then manually run it to create those
records.
Performance
Prisma is faster at execting a createMany()
instead of many create
or
upsert
functions. Unfortunately, you lose the ability to easily make your seed
idempotent with a single function call.
One solution to simulate an upsert
will still using createMany()
could be
to start with the full array of data and first check to see whether each of
those records already exist in the database. If they do, create two
arrays: one for records that don't exist and run createMany()
with them, and
the second list for records that do exist, and run updateMany()
on those.
Unfortunately this relies on a select query for each record, which may negate
the performance benefits of createMany()
. Since you are running seeds
realitively rarely, it's our recommendation that you focus less on absolute
performance and worry more about making them easy to maintain.
Types
If you're using Typescript you'll probably want to type your seeds as well. Getting the right types for Prisma models can be tricky, but here's the formula:
import { db } from 'api/src/lib/db'
import type { Prisma } from '@prisma/client'
export default async () => {
try {
const users: Prisma.UserCreateArgs['data'][] = [
{ name: 'Alice', email: 'alice@redwoodjs.com },
{ name: 'Bob', email: 'bob@redwoodjs.com },
]
await db.user.createMany({ data: users })
} catch (error) {
console.error(error)
}
}
Creating seed data
Take a look at scripts/seed.js
(or .ts
if you're working on a Typescript
project):
import { db } from 'api/src/lib/db'
export default async () => {
try {
// Create your database records here! For example, seed some users:
//
// const users = [
// { name: 'Alice', email: 'alice@redwoodjs.com },
// { name: 'Bob', email: 'bob@redwoodjs.com },
// ]
//
// await db.user.createMany({ data: users })
console.info(
'\n No seed data, skipping. See scripts/seed.ts to start seeding your database!\n'
)
} catch (error) {
console.error(error)
}
}
Let's create some categories for a bookstore. For this example, assume the
Category
model has a unique constraint on the name
field. Remove the
commented example and add your code:
export default async () => {
try {
const data = [
{ name: 'Art', bisacCode: 'ART000000' },
{ name: 'Biography', bisacCode: 'BIO000000' },
{ name: 'Fiction', bisacCode: 'FIC000000' },
{ name: 'Nature', bisacCode: 'NAT000000' },
{ name: 'Travel', bisacCode: 'TRV000000' },
{ name: 'World History', bisacCode: 'HIS037000' },
]
for (const item of data) {
await db.category.upsert({
where: { name: item.name },
update: { code: item.code },
create: { name: item.name, code: item.code },
})
}
} catch (error) {
console.error(error)
}
}
You can now execute this seed as many times as you want and you'll end up with that exact list in the database each time. And, any additional categories you've created in the meantime will remain. Remember: seeds are meant to be the minimum amount of data you need for your app to run, not necessarily all the data that will ever be present in those tables.
Seeding users for dbAuth
If using dbAuth and seeding users, you'll need to add a hashedPassword
and
salt
using the same algorithm that dbAuth uses internally. Here's an easy way
do that:
import { hashPassword } from '@redwoodjs/auth-dbauth-api'
export default async () => {
const users = [
{ name: 'John', email: 'john@example.com', password: 'secret1' },
{ name: 'Jane', email: 'jane@example.com', password: 'secret2' },
]
for (const user of users) {
const [hashedPassword, salt] = hashPassword(user.password)
await db.user.upsert({
where: {
email: user.email,
},
create: {
name: user.name,
email: user.email,
hashedPassword,
salt,
},
update: {
name: user.name,
hashedPassword,
salt,
},
})
}
}
What if I don't need seeds?
In order to stop automatically executing seeds with the prisma migrate
commands you can remove the following lines from package.json
in the root of
your app:
"prisma": {
"seed": "yarn rw exec seed"
},
You can then delete the scripts/seed.js
file.