Creating a constantly updated Read Replica with dgraph

Users
,
3

Hi,

I would model your data like this: Account <-- Transaction

Schema:

type Account {
  account.Id
  account.Balance
}

type Transaction {
  transaction.Id
  transaction.Amount
  transaction.Account
}

transaction.Account: uid @reverse .

then I wouldn’t need to care about transaction upserts from the account perspective, as it would be solved through the revese edge.

Upserting Account just like in your post above.
Upserting Transaction with 2 query blocks, one for the Account of the transaction, the other one for the Transaction itself:

upsert {
  query {
    Account as var(func: eq(<account.Id>, "1"))
    Transaction as var(func: eq(<transaction.Id>, "1"))
  }
  mutation {
      set {
        uid(Transaction) <transaction.Id> "1" .
        uid(Transaction) <transaction.Amount> "75" .
        uid(Transaction) <dgraph.type> "Transaction" .
        uid(Transaction) <transaction.Account> uid(Account) .
	# if an Account can not exist, when adding the transaction, add:
	uid(Account) <account.Id> "1" .
        uid(Account) <dgraph.type> "Account" .
      }
  }
}

So, answering questions:

  1. You don’t have to add data bi-directional in a graph. It will work correctly with @reverse directive. Just 2 query blocks for each transactions are ok, as shown above.
    To query Account transactions:
  query {
    Account as var(func: eq(<account.Id>, "1")) {
      account.Id
      account.Balance
      transactions: ~transaction.Account {
           transaction.Id
           transaction.Amount
      }
  }
  1. You can just process transactions row by row with this schema.

In summary, you will not need the “Transactions” column from the Account CSV at all.

Hope that helps!