[SPARK-24803][SQL] add support for numeric

[wangtao605]

What changes were proposed in this pull request?

numerical is as same with decimal. Spark has already supported decimal，so i think we should add support for numeric to align SQL standards.

(Please fill in changes proposed in this fix)

How was this patch tested?

already exist tests

Please review http://spark.apache.org/contributing.html before opening a pull request.

[AmplabJenkins]

Can one of the admins verify this patch?

[maropu]

@gatorsmile Is it worth adding a numeric type as an alias of decimal? Both types seems to be in the SQL standard and some dbms (e.g., postgresql and sql server) can parse both.
@wangtao605 Can you add tests to check if these types can be parsed correctly?

[gatorsmile]

Although MS SQL Server looks like use them interchangeably, https://docs.microsoft.com/en-us/sql/t-sql/data-types/decimal-and-numeric-transact-sql?view=sql-server-2017, they are not exactly the same in ANSI SQL.

26. NUMERIC species the data type exact numeric, with the decimal precision and scale specified by the and .
27. DECIMAL species the data type exact numeric, with the decimal scale specified by the and the implementation-de ned decimal precision equal to or greater than the value of the specified .

[maropu]

aha, I didn't know that and postgresql also uses them interchangeably, too.

[dmateusp]

Just checked out the PR,

scala> spark.sql("SELECT CAST(1 as NUMERIC)")
res0: org.apache.spark.sql.DataFrame = [CAST(1 AS DECIMAL(10,0)): decimal(10,0)]

scala> spark.sql("SELECT NUMERIC(1)")
org.apache.spark.sql.AnalysisException: Undefined function: 'NUMERIC'. This function is neither a registered temporary function nor a permanent function registered in the database 'default'.; line 1 pos 7

I imagine some tests could be added here:

• sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/parser/DataTypeParserSuite.scala
• sql/core/src/test/resources/sql-tests/inputs/

Do you think it's worth having a separate DataType or just have it as an alias?

[rxin]

Why did you need this change? Given it's very difficult to revert the change (or introduce a proper numeric type if ever needed in the future), I would not merge this pull request unless there are sufficient justification.

[wangtao605]

@dmateusp Actually i think "Numeric" has no essential difference with "Decimal". May be just have it as an alias is better,i will add some tests if you agree.

[wangtao605]

@rxin In order to support sql syntax better and align SQL standards. I think it is worth to add a numeric type as an alias of decimal.

[gatorsmile]

@wangtao605 Do you mind documenting our behavior in our Spark SQL doc?

[wangtao605]

@wangtao605 Do you mind documenting our behavior in our Spark SQL doc?

Yes, it's ok.