Updating indexes

[phawxby]

Following this discussion I had a plan to add a new index to my entity definition I could then index.

      primary: {
        pk: {
          field: 'pk',
          composite: ['client_id']
        },
        sk: {
          field: 'sk',
          composite: []
        }
      },
      byType: { // <-- This is the new one
        index: 'byType',
        pk: {
          field: 'byTypePk',
          composite: []
        },
        sk: {
          field: 'byTypeSk',
          composite: ['client_id']
        }
      }

This allows me to easily query for all rows of a type (although to be honest if electrodb allows you to define a a GSI on edb_e that would make this whole problem go away). Performing an update on the entity doesn't seem to update the secondary index and I'm not sure why.

[0] |  +482ms      UpdateExpression: 'SET #updated_at = :updated_at_u0, #client_id = :client_id_u0, #__edb_e__ = :__edb_e___u0, #__edb_v__ = :__edb_v___u0',
[0] |  +482ms      ExpressionAttributeNames: {
[0] |  +482ms        '#pk': 'pk',
[0] |  +482ms        '#sk': 'sk',
[0] |  +482ms        '#updated_at': 'updated_at',
[0] |  +482ms        '#client_id': 'client_id',
[0] |  +482ms        '#__edb_e__': '__edb_e__',
[0] |  +482ms        '#__edb_v__': '__edb_v__'
[0] |  +482ms      },
[0] |  +482ms      ExpressionAttributeValues: {
[0] |  +482ms        ':updated_at_u0': '2025-05-27T19:59:14.285Z',
[0] |  +482ms        ':client_id_u0': '01JW9MFBQMB74BX8HGFNSQHSPJ',
[0] |  +482ms        ':__edb_e___u0': 'client',
[0] |  +482ms        ':__edb_v___u0': '1'
[0] |  +482ms      },
[0] |  +482ms      TableName: 'paulhawxby-myenergi-auth-auth',
[0] |  +482ms      Key: {
[0] |  +482ms        pk: '$auth#client_id_01jw9mfbqmb74bx8hgfnsqhspj',
[0] |  +482ms        sk: '$client_1'
[0] |  +482ms      },
[0] |  +482ms      ConditionExpression: 'attribute_exists(#pk) AND attribute_exists(#sk)'
[0] |  +482ms    }

Aside from the above, what strategies have people used to add new indexes with new attributes to existing tables?

[phawxby]

Turns out I rectified this by updating electrodb and then doing an inverse GSI on SK/PK so I could then do a query on the SK