feat: support multiple value for pivot

[chenkovsky]

https://spark.apache.org/docs/latest/sql-ref-syntax-qry-select-pivot.html

currently this sql is not supported.

SELECT * FROM person
    PIVOT (
        SUM(age) AS a, AVG(class) AS c
        FOR (name, age) IN (('John', 30) AS c1, ('Mike', 40) AS c2)
    );

[iffyio]

Suggested change

	value_column: Vec<Expr>, // Expr is a identifier or a compound identifier
	value_column: Vec<Expr>,

[iffyio]

hmm this assertion is probably not needed, I think verified_stmt already does the same assertion

[iffyio]

similar to unpivot can we call parse_expr directly here?

[chenkovsky]

no, we cannot do it here. (name, age) IN (('John', 30) AS c1, ('Mike', 40) AS c2) will be parsed as an expr.

[iffyio]

Ah right that makes sense, I think we can do something this instead to reuse existing capabilities of the parser?

let value_column = if self.peek_token_ref().token == Token::LParen {
    self.parse_parenthesized_column_list_inner(Mandatory, false, |p| {
         p.parse_subexpr(self.dialect.prec_value(Precedence::Between))
     })?
} else {
    vec![self.parse_subexpr(self.dialect.prec_value(Precedence::Between))?]
};

[iffyio]

Ah right that makes sense, I think we can do something this instead to reuse existing capabilities of the parser?

let value_column = if self.peek_token_ref().token == Token::LParen {
    self.parse_parenthesized_column_list_inner(Mandatory, false, |p| {
         p.parse_subexpr(self.dialect.prec_value(Precedence::Between))
     })?
} else {
    vec![self.parse_subexpr(self.dialect.prec_value(Precedence::Between))?]
};

[iffyio]

hmm heads up that this representation wouldn't be ideal - a compound identifier would be expected to have more than one ident in it

[iffyio]

can we add tests to demonstrate the new behavior? it seems like those are currently lacking in the PR?

[iffyio]

LGTM! Thanks @chenkovsky!
FYI @chenkovsky please re-request review explicitly when ready after addressing comments to avoid the PR getting missed in the queue