CODEWITHSUNDEEP
amazon-dynamodbnosql
Diego Sanabria
Diego Sanabria

Reputation: 63

DynamoDB Collection(s) Model

I am learning how to model a NoSQL databases and trying to understand how to make this works properly.

For my use case I have spin up a local instance of aws dynamo db and connected successfully.

Now, let's say I have two contexts "Users" and "Groups", where one user can belong to any group, and one group can have multiple users. I am trying to avoid multiple collections but I cannot see the way to.

What I have in mind is something like:

{    
     name: "Bob Smith",    
     dob: "1980-01-01     
     groups: [
          {
               name: "Sports"
               CreatedAt: "2022-01-05"
          }.
          {
               name: "Arts"
               CreatedAt: "2022-01-08"
          }
     ]
}

What I am not sure is what should we do if I need to create or update a Group? if I follow the previous pattern, how would be the query for the groups?

Shall I have 2 collections?

Thanks guys for your advice.

Upvotes: 0

Views: 574

Answers (1)

Leeroy Hannigan
Leeroy Hannigan

Reputation: 19893

You would typically only hold a pointer to the groups within a users item, if you need to update the group details then you would do so in the item for the group.

pk sk groups other
user123 USER#user123 groups[{pk:group1, sk: GROUP#group1}, {pk:group3, sk: GROUP#group3}] user data
user009 USER#user009 groups[{pk:group5, sk: GROUP#group5}, {pk:group7, sk: GROUP#group7}] user data
group1 GROUP#group1 group-info thats editable
group3 GROUP#group3 group-info thats editable
group5 GROUP#group5 group-info thats editable
group7 GROUP#group7 group-info thats editable

Of course this is not the only way to do this, but with all things NoSQL it totally depends on your access patterns. For example, if you have a requirement to get all users for a given group then you would need to change the above schema:

pk sk other other
user123 USER#user123 user-data
user123 GROUP#group1 some data
user123 GROUP#group3 some data
group1 GROUP#group1 group data
group3 GROUP#group3 group data

Now you can get all the groups a user belongs to by issuing a Query stating:

pk=user123 AND sk BEGINS_WITH(GROUP#).

Then use a Global Secondary Index to get all the users which belong to a given group, with sk being your GSI's partition key:

gsi_pk = GROUP#group1

Upvotes: 1

Related Questions