Is an update in Cassandra not an anti pattern?

Question

As per Datastax documentation a read before a write in Cassandra is an anti pattern.

Whenever we use UPDATE either in CQLSH or using the Datastax drivers to set a few columns (with IFs & collection updates), does it not do a read before write first? Is that not an anti pattern? Am I missing something?

P.S I am not talking about mere UPSERTS but UPDATES on specific columns.

TIA!

Answer 1

No, Update is not an anti-pattern.

In Cassandra update is an upsert operation similar to insert.

UPDATE writes one or more column values to a row in a Cassandra table. Like INSERT, UPDATE is an upsert operation: if the specified row does not exist, the command creates it. All UPDATEs within the same partition key are applied atomically and in isolation.

But Lightweight transactions are read before write operation. Actually at the cost of four round trips.

Example of Lightweight transaction :

#Lightweight transaction Insert
INSERT INTO customer_account (customerID, customer_email) 
VALUES (‘LauraS’, ‘[email protected]’)
IF NOT EXISTS;

#Lightweight transaction Update
UPDATE customer_account
SET    customer_email=’[email protected]’
IF     customerID=’LauraS’; 

Both of the above statement are Lightweight transactions

Source : http://docs.datastax.com/en/cql/3.3/cql/cql_reference/cqlUpdate.html#cqlUpdate__description