Using Tuples with the Java driver

Cassandra allows to use tuple data types in tables and user-defined types:

CREATE TABLE ks.collect_things (
  pk int,
  ck1 text,
  ck2 text,
  v tuple<int, text, float>,
  PRIMARY KEY (pk, ck1, ck2)
);

Fetching Tuples from Rows results

The DataStax Java driver exposes a special TupleValue class to handle such columns. TupleValue exposes getters allowing to extract from the tuple all the data types supported by Cassandra:

Row row = session.execute("SELECT v FROM ks.collect_things WHERE pk = 1").one();

TupleValue tupleValue = row.getTupleValue("v");

int firstValueInTuple = tupleValue.getInt(0);

String secondValueInTuple = tupleValue.getString(1);

Float thirdValueInTuple = tupleValue.getFloat(2);

Using tuples as statement parameters

A prepared statement may contain a Tuple as a query parameter. In such cases, users will need to create or gather a TupleType first, in order to be able to create a TupleValue to bind:

PreparedStatement ps = session.prepare("INSERT INTO ks.collect_things (pk, ck1, ck2, v) VALUES (:pk, :ck1, :ck2, :v)");

TupleType tupleType = cluster.getMetadata().newTupleType(DataType.cint(), DataType.text(), DataType.cfloat());

BoundStatement bs = ps.bind();
bs.setInt("pk", 1);
bs.setString("ck1", "1");
bs.setString("ck2", "1");
bs.setTupleValue("v", tupleType.newValue(1, "hello", 2.3f));

session.execute(bs);

The method newValue(Object...) follows the same rules as new SimpleStatement(String, Object...); there can be ambiguities due to the fact that the driver will infer the data types from the values given in parameters of the method, whereas the data types required may differ (numeric literals are always interpreted as int).

To avoid such ambiguities, a TupleValue returned by newValue() also exposes specific setters for all the existing Cassandra data types:

TupleType tupleType = cluster.getMetadata().newTupleType(DataType.bigint(), DataType.text(), DataType.cfloat());

TupleValue value = tupleType.newValue().setLong(0, 2).setString(1, "hello").setDouble(2, 2.3f);

More use cases

Users can also define single-usage tuples in SELECT queries with the IN keyword (called a “multi-column IN restriction”), usually for tables with composite clustering keys. In this case, a tuple will be usable the same way it was for prepared statements’ parameters:

TupleType oneTimeUsageTuple = cluster.getMetadata().newTupleType(DataType.text(), DataType.text());

PreparedStatement ps = session.prepare("SELECT * FROM ks.collect_things WHERE pk = 1 and (ck1, ck2) IN (:t)");

BoundStatement bs = ps.bind();
bs.setTupleValue("t", oneTimeUsageTuple.newValue("1", "1"));

session.execute(bs);

More generally, the IN keyword in a SELECT query will be used to define a list of desired values of the filtered clustering keys, those would simply be bound as a list of TupleValue with the Java driver:

TupleType oneTimeUsageTuple = cluster.getMetadata().newTupleType(DataType.text(), DataType.text());

PreparedStatement ps = session.prepare("SELECT * FROM ks.collect_things WHERE pk = 1 AND (ck1, ck2) IN :l");

BoundStatement bs = ps.bind();
bs.setList("l", Arrays.asList(oneTimeUsageTuple.newValue("1", "1"), oneTimeUsageTuple.newValue("1", "2"), oneTimeUsageTuple.newValue("2", "1")));

session.execute(bs);