How do I structure DynamoDB keys / indices

I am trying to set up what I think is a relative simple structure in DDB, but I am finding the docs very confusing and my experiments keep running into problems. I'm hoping someone can help point me in the right direction on how I should structure this

I have three different object classes. Each has a unique, static id number; a non-unique date identifier that can change; and a set of other attributes (non-unique, changeable). The set of other attributes is different for each class

I need to be able to retrieve objects by id; retrieve all objects of a specific class; retrieve all objects within a range of dates; and update an object's date and other attributes. I may need to search by other attributes later.

What is the best way to structure tables / keys / secondary indices for this situation?

So far I have tried:

1. A single table, hash key = object type, range key = id, GlobalSecondaryIndex = date as hash key.

   • GSI doesn't support 'between' operator for date lookups.
     (KeyConditionExpression=Key('date').between(d_min, d_max))

2. One table for each class, hash key = id, range key = date

   • when I update an item whose date has changed, I end up with a second record as the primary key is composed of both id and date

3. One table for each class, hash key = id, no range key. GSI = date as hash key

   • GSI doesn't support 'between' operator for date lookups.

4. One table for each class, hash key = id, no range key. GSI = date as range key

   • GSI requires a hash key

1. I need to be able to retrieve objects by id;
2. retrieve all objects of a specific class;
3. retrieve all objects within a range of dates;
4. update an object's date and other attributes.

I may need to search by other attributes later = add more GSI's.

Base Table

Fulfills use-case 1 + 2 + 4, assuming you know the class when searching by id.

GSI

Fulfills use-case 4. This makes the assumption your writes won't exceed 1000 WCU, if they will, you will need to write shard the GSIPK