Dmitry Konstantinov created CASSANDRA-19814:
-----------------------------------------------
Summary: UDT codec decode is too restrictive in decoding of
unknown fields
Key: CASSANDRA-19814
URL: https://issues.apache.org/jira/browse/CASSANDRA-19814
Project: Cassandra
Issue Type: Bug
Components: Client/java-driver
Reporter: Dmitry Konstantinov
Assignee: Dmitry Konstantinov
The current UDT code logic in driver 4.x expects the number of UDT fields
exactly equal to the number of fields in the schema version used to create the
codec.
As a result it makes applications less friendly to a live upgrade logic when we
change DB schema and do a rolling restart of applications to switch from an old
app version to a new one. In this scenario we can add a new field to UDT
(backward compatible change) and a new application version can write this UDT
with a new field filled but the old application version will fail to decode
such UDT with the extra unknown field.
Driver 3.x UDT codec has a different behaviour - the UDT codec logic reads
known UDT fields received from network and ignores the unknown tail, which
makes in tolerant to such schema changes:
https://github.com/apache/cassandra-java-driver/blob/3.x/driver-core/src/main/java/com/datastax/driver/core/TypeCodec.java#L2202
A code example to illustrate the issue:
Schema
{code:java}
CREATE TYPE test_keyspace.basicInfoInner (
weight text,
height text
);
CREATE TYPE test_keyspace.basicInfo (
birthday timestamp,
nationality text,
inner frozen<basicInfoInner>
);
CREATE TABLE test_keyspace.tableWithUdt
(
id varchar,
value frozen<list<basicInfo>>,
PRIMARY KEY (id)
); {code}
Code:
{code:java}
CqlSession session = getSession();
UdtCodec udtCodec = new UdtCodec(userDefinedType);
TypeCodec<List<UdtValue>> listOfUdtCodec = TypeCodecs.listOf(udtCodec);
{
ResultSet resultSet = session.execute("select id, value from tableWithUdt");
for (Row row : resultSet) {
String id = row.getString(0);
List<UdtValue> udtValueList = row.get(1, listOfUdtCodec);
}
}
session.execute("ALTER TYPE basicInfo ADD details text");
session.execute("INSERT INTO tableWithUdt (id, value) VALUES ('4', [\n" +
" { birthday : '1993-06-18', " +
" nationality : 'New Zealand', " +
" inner: { weight : '70', height : '70' } " +
" },\n" +
" { birthday : '1993-06-19', " +
" nationality : null, " +
" inner: { weight : '70', height : null }, " +
" details: 'added details'" + // <=== insert a value for the new
added UDT field
" }\n" +
" ])");
{
ResultSet resultSet = session.execute("select id, value from tableWithUdt");
for (Row row : resultSet) {
String id = row.getString(0);
// ==== UdtCodec throws IllegalArgumentException: Too many fields in
encoded UDT value, expected 3 ====
List<UdtValue> udtValueList = row.get(1, listOfUdtCodec);
}
}
private static UserDefinedType getUserDefinedType(CqlSession session, String
udtName) {
return session.getMetadata()
.getKeyspace(session.getKeyspace().orElseThrow(() -> new
IllegalArgumentException("Udt name " + udtName + " can't be loaded due to null
keyspace")))
.orElseThrow(() -> new IllegalArgumentException("Metadata is not
found"))
.getUserDefinedType(udtName)
.orElseThrow(() -> new IllegalArgumentException("Udt name " +
udtName + " isn't found"));
}
{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
