Introduction
Prisma uses a dedicated, intermediate format for importing and exporting data: The Normalized Data Format (NDF).
+--------------+ +----------------+ +------------+
| +--------------+ | | | |
| | | | | | | |
| | SQL | | (1) transform | NDF | (2) chunked upload | prisma |
| | MongoDB | | +--------------> | | +-------------------> | |
| | JSON | | | | | |
| | | | | | | |
+--------------+ | +----------------+ +------------+
+--------------+
In this tutorial, you'll perform the following steps:
- Create a Prisma service
- Seed some initial data for the service
- Export the data in NDF
- Deploy the service to a new stage
- Import the data in NDF
Create a Prisma service
In your terminal, navigate to a folder of your choice and run the following command:
prisma init import-example
When prompted what kind of template to use, choose the Minimal setup: database-only one.
This created a new directory called import-example with the root configuration file prisma.yml as well as the definition of the service's data model in datamodel.graphql.
Next, you'll update the data model to also include a relation.
Open datamodel.graphql and change the contents to looks as follows:
type User {
id: ID! @id
name: String!
posts: [Post!]!
}
type Post {
id: ID! @id
title: String!
author: User!
}
Seed initial data
Next, you're going to seed some initial data for the service.
Create a new file called seed.graphql inside the import-example directory and add the following mutation to it:
mutation {
createUser(
data: {
name: "Sarah"
posts: {
create: [
{ title: "GraphQL is awesome" }
{ title: "It really is" }
{ title: "How to GraphQL is the best GraphQL tutorial" }
]
}
}
) {
id
}
}
Now you need to tell the CLI that you created this file. You can do so by setting the seed property in prisma.yml.
Open prisma.yml and update its contents to look as follows:
service: import-example
stage: dev
datamodel: datamodel.graphql
# to enable auth, provide
# secret: my-secret
disableAuth: true
seed:
import: seed.graphql
When deploying the service, the CLI will now send the mutation defined in seed.graphql to your service's API.
Deploy the Prisma service by running the following command:
prisma deploy
When prompted where (i.e. to which cluster) to deploy your Prisma service, choose one of the public cluster options: prisma-eu1 or prisma-us1. (Note that seeding also works when deploying with Docker)
The CLI now deploys the service and executes the mutation in seed.graphql. To convince yourself the seeding actually worked, you can open up a GraphQL Playground and send the following query:
{
users {
name
posts {
title
}
}
}
The Prisma API will respond with the following data:
{
"data": {
"users": [
{
"name": "Sarah",
"posts": [
{
"title": "GraphQL is awesome"
},
{
"title": "It really is"
},
{
"title": "How to GraphQL is the best GraphQL tutorial"
}
]
}
]
}
}
Export the data in NDF
It's time to export the in the Normalized Data Format.
In the import-example in your terminal, execute the following command:
prisma export
This creates a new file called export-__TIMESTAMP__.zip where __TIMESTAMP__ represents the exact time of the export. The files in the zip directory are in NDF. To learn more about the structure, check out the reference documentation for the NDF.
Deploy the service to a new stage
Next, you'll create "clone" of the service by deploying it to a new stage.
Open prisma.yml and set the stage property to a new value. Also remove the seed and cluster properties!
service: import-example
stage: test
datamodel: datamodel.graphql
# to enable auth, provide
# secret: my-secret
disableAuth: true
Run prisma deploy again to deploy the service the new test stage.
Like before, when prompted where to deploy your Prisma service, either choose prisma-eu1 or prisma-us1.
Import the data in NDF
Now that the service is running, you can import the data from the zip directory!
Run the following command in your terminal. Note that you need to replace the __DATA__ placeholder with the path to the exported zip directory (e.g. export-2018-01-13T19:28:25.921Z.zip):
prisma import --data __DATA__
That's it! To convince yourself the import actually worked, you can open up a GraphQL Playground for the current test stage and send the above query again.