Anyone have any examples of adding a GSI1PK example in SST I Serverless Stack #help

Anyone have any examples of adding a GSI1PK exampl...

Devin

12/22/2021, 1:49 PM

Anyone have any examples of adding a GSI1PK example in SST. I’m trying to do the single table design thing. My table is created like

Copy code

const customersTable = new sst.Table(this, "Customers", {
      fields: {
        pk: sst.TableFieldType.STRING,
        sk: sst.TableFieldType.STRING,
      },
      primaryIndex: { partitionKey: "pk", sortKey: "sk" },
    });

What I want is to make,

createdAt

a secondary primary key so that I can fulfill the access pattern of

getCustomerByCreatedAt

It seems like this is the correct way to do this but I”m not sure.

Copy code

customersTable.addGlobalIndexes({
      indexName: "createdAt",
      partitionKey: { name: "gs1pk", type: sst.TableFieldType.STRING },
    });

thdxr

12/22/2021, 2:00 PM

The index name should probably be generic

thdxr

12/22/2021, 2:00 PM

instead of specific to

createdAt

Kujtim Hoxha

12/22/2021, 2:11 PM

since you use

gs1pk

you should probably call the index

gs1

Kujtim Hoxha

12/22/2021, 2:14 PM

Here is an example of how I have set this up

Copy code

// Organization service table
    this.organizationTable = new sst.Table(this, "OrganizationTable", {
      dynamodbTable: {
        tableName: `${this.stage}-devstride-organizations`,
        removalPolicy: process.env.IS_LOCAL
          ? RemovalPolicy.DESTROY
          : RemovalPolicy.RETAIN,
      },
      fields: {
        // Primary index
        PK: sst.TableFieldType.STRING,
        SK: sst.TableFieldType.STRING,

        // GSI1 
        PK1: sst.TableFieldType.STRING,
        SK1: sst.TableFieldType.STRING,
      },
      primaryIndex: { partitionKey: "PK", sortKey: "SK" },
      secondaryIndexes: {
        GSI1: {
          partitionKey: "PK1",
          sortKey: "SK1",
          indexProps: {
            projectionType: ProjectionType.ALL,
          },
        },
      },
    });

In your particular case I imagine that you have some sort of

storeId

which the consumer belongs to, in that case your GSI

PK

should be the

storeId

and your

SK

something like

${createdAt}:${customerId}

Kujtim Hoxha

12/22/2021, 2:15 PM

^ this is assuming that you want the customers sorted by the createAt date

Devin

12/22/2021, 2:17 PM

Yeah that’s very close to it. So for existing items that have a

createdAt

what will happen?

Kujtim Hoxha

12/22/2021, 2:26 PM

So if there is data already in the system you will have to: • Create the index • Run a migration script that adds the new PK/SK for all records What I usually do is that I keep the

GSI

columns just for the purpose of querying and still keep the columns that are used in the

PK

and

SK

in the table so in your case you would have

Copy code

| PK          	| SK 	| PK1        	| SK1                     	| storeId  	| customerId  	| createdAt  	|
|-------------	|----	|------------	|-------------------------	|----------	|-------------	|------------	|
| c:customer1 	| c: 	| c:storeId1 	| c:2020-01-01:customerId1 	| storeId1 	| customerId1 	| 2020-01-01 	|

Kujtim Hoxha

12/22/2021, 2:27 PM

^ you probably need to expand the thread panel to see the table correctly 😂

Devin

12/22/2021, 2:33 PM

Thanks kujtim! I need to learn how to do this anyway so I guess now's the time

Kujtim Hoxha

12/22/2021, 2:50 PM

I have started using https://github.com/sensedeep/dynamodb-onetable for my personal project if you are using JS or TS this is a pretty good library

Kujtim Hoxha

12/22/2021, 2:51 PM

It has some small issues here and there but overall I am happy with it

Devin

12/22/2021, 9:40 PM

I have Dynobase

Devin

12/22/2021, 9:42 PM

so I can use that should I need to. I’m not exactly sure if at this point I know enough about DynamoDB to use an abstraction. IME it’s better to fumble your way through and learn the why first. I’ll absolutely look at dynamo-onetable as this goes along

Open in Slack

Previous Next