:auto transaction doesn't run as described in the manual

Neo4j Graph Platform Cypher
, ,
25

@glilienfield @bennu_neo @Rcolinp Justr for your info, I loaded 11 million transactions of different kind, using the latest approach proposed by @glilienfield.

This approach using apoc turned out to be by far the most readable one (after a bit of a sprucing up of the query!)

This is the result:

LOAD CSV WITH HEADERS FROM 'file:///transactions.csv' AS row FIELDTERMINATOR ';'

CALL {

WITH row

WITH row
    WHERE row.transactionid IS NOT NULL

MERGE (transaction:Transaction { transactionid : row.transactionid })
        SET transaction = row,
            transaction.uuid = apoc.create.uuid(),
            transaction.consumer_id = NULL // remove the attribute no more used

    with transaction, row
    
    OPTIONAL MATCH (organisation:Organisation {id: toInteger(row.organisation_id)})

    CALL apoc.do.when(organisation IS NOT NULL, 
    'WITH $tran as tran, $org as org 
        MERGE (tran)-[h:TRANSACTION_BELONGS_TO_ORGANISATION]->(org) 
        RETURN 1',
    'WITH $tran as tran 
        MERGE (error:Error) 
        MERGE (tran)-[h:TRANSACTION_HAS_NO_ORGANISATION]->(error) 
        RETURN 1',
    {tran: transaction, org: organisation}) yield value AS X

    WITH transaction, row

    OPTIONAL MATCH (bankAccount:BankAccount {id: toInteger(row.bank_account_id)})

    CALL apoc.do.when(bankAccount IS NOT NULL, 
    'WITH $tran as tran, $bankAcct as bankAcct 
        MERGE (tran)-[h:TRANSACTION_HAS_BANK_ACCOUNT]->(bankAcct)
        RETURN 1', 
    'WITH $tran as tran  
        MERGE (error:Error) 
        MERGE (tran)-[h:TRANSACTION_HAS_NO_BANK_ACCOUNT]->(error) 
        RETURN 1',
    {tran: transaction, bankAcct: bankAccount}) yield value AS Y

return transaction

} IN TRANSACTIONS OF 5000 ROWS

return count(transaction)

The story will follow with a new discussion .... but this is an achieved result I'd like to share!

Thank you to everybody!