[SPARK-10186][SQL] support postgre array type in JDBCRDD

[cloud-fan]

Add ARRAY support to PostgresDialect.

Nested ARRAY is not allowed for now because it's hard to get the array dimension info. See http://stackoverflow.com/questions/16619113/how-to-get-array-base-type-in-postgres-via-jdbc

Thanks for the initial work from @mariusvniekerk !

Close #9137

[cloud-fan]

cc @marmbrus

[SparkQA]

Test build #45737 has finished for PR 9662 at commit b7db9ec.

• This patch fails MiMa tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):\n * case class ArrayConversion(elementConversion: JDBCConversion) extends JDBCConversion\n

[mariusvniekerk]

I've added write support in #9137 as well if you want to just use it from there.

[SparkQA]

Test build #45739 has finished for PR 9662 at commit 378c5a9.

• This patch fails MiMa tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):\n * case class ArrayConversion(elementConversion: JDBCConversion) extends JDBCConversion\n

[mariusvniekerk]

Is that the same in all backends that support arrays (Oracle etc)?

[cloud-fan]

you are right, it's not always working

[SparkQA]

Test build #45742 has finished for PR 9662 at commit fbaa543.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):\n * case class ArrayConversion(elementConversion: JDBCConversion) extends JDBCConversion\n * abstract class JdbcDialect extends Serializable\n

[mariusvniekerk]

These test failures don't seem to be related?

[cloud-fan]

retest this please

[SparkQA]

Test build #45794 has finished for PR 9662 at commit fbaa543.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):\n * case class ArrayConversion(elementConversion: JDBCConversion) extends JDBCConversion\n * abstract class JdbcDialect extends Serializable\n

[marmbrus]

This is a public API, if we add fields here we are going to break any existing implementations. Do we have to do this?

[SparkQA]

Test build #46050 has finished for PR 9662 at commit ad52183.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):\n * case class ArrayConversion(elementConversion: JDBCConversion) extends JDBCConversion\n * abstract class JdbcDialect extends Serializable\n

[marmbrus]

Is this just Literal.create(null, a.dataType)?

[cloud-fan]

ah yea, we can simply this.

[marmbrus]

Thanks, I'm going to merge this into master and 1.6.