Switch SingleStore (memsql) JDBC driver from MariaDB to official driver

[jhlodin]

See memsql plugin:

trino/plugin/trino-memsql/pom.xml

Line 72 in 8e9da98

<groupId>org.mariadb.jdbc</groupId>

Currently we package the MariaDB JDBC driver to connect to SingleStore, but there is now an official SingleStore JDBC driver.

@findepi @ebyhr Any insight as to why we use the MariaDB driver or otherwise if this is something worth doing?

[ebyhr]

The reason is there was no such dedicated JDBC driver when we introduced the connector. It worth switching the driver if it's stable and the license is fine.

[leveyja]

TestMemSqlTypeMapping.testUnsignedTypes() is failing with the SingleStore driver. I'll investigate and see if we need a code change to support it (or if the driver needs a fix itself). Thanks for mentioning it @ebyhr ! 👍

2022-02-28T02:54:30.690-0600 SEVERE Error processing Split 20220228_085430_00322_4ysf7.1.0.0-0 io.trino.plugin.jdbc.JdbcSplit@41161598 (start = 1899250.09428, wall = 23 ms, cpu = 0 ms, wait = 0 ms, calls = 1): JDBC_ERROR: byte overflow
io.trino.spi.TrinoException: byte overflow
	at io.trino.plugin.jdbc.JdbcRecordCursor.handleSqlException(JdbcRecordCursor.java:299)

[leveyja]

There is a bug in the field decoding in v1.0.0 and v.1.0.1 of the new SingleStore driver.
0.1.0-preview release of SingleStore doesn't have the issue. I'm investigating how to file a bug or submit a fix to their repo: https://github.com/memsql/S2-JDBC-Connector/releases - but until then this looks like a blocker for switching (and slightly worrying).

[ebyhr]

@leveyja You can handle the byte overflow by fixing MemSqlClient.getUnsignedMapping(). The typeHandle.getJdbcTypeName() contains precisions after the upgrade likes tinyint(3) unsigned

[leveyja]

Thank you! I was in MemSqlClient trying to determine how/why it worked in 0.1.0-preview versus 1.0.0/1.0.1.
That's exactly the issue - name is now "TINTINT(3) UNSIGNED", not "TINYINT UNSIGNED".
Should be straight-forward (ish) to fix 👍