Add support for Snowflake identifier function

[yoavcloud]

Expand the support for the IDENTIFIER function in Snowflake, which is used to generate identifiers dynamically. See here: https://docs.snowflake.com/en/sql-reference/identifier-literal

The ObjectNamePart enum was extended with a new part type to support the use of the identifier function as an object name. Some work went into organizing Parser:parse_object_name a bit.

A few examples:

-- identifier for table name
create or replace temp table identifier('test')(c1 int, identifier int); 
-- identifier for column name
select identifier('identifier') from test;
-- identifier for table name
select c1 from identifier('test');
-- identifier for table alias
select identifier('alias1').* from test as alias1;

[iffyio]

wondering can we instead extende the else clause to do something like this?

else {
    let ident = self.parse_identifier()?;
    let part = if self.dialect.is_identifier_generating_function_name(&ident) {
          self.expect_token(&Token::LParen)?;
          let args = ...
          self.expect_token(&Token::RParen)?;
         ObjectNamePartFunction { name:ident, args }
    } else {
          ObjectNamePart::Identifier(ident));
    }
}

thinking that way we skip this extra branch, and the extra error scenario in the function part parsing

[iffyio]

Could we add a test scenario for how this interacts with the double dot feature?

[yoavcloud]

In what way? I believe you cannot chain IDENTIFIER. For example, this is illegal: SELECT * FROM IDENTIFIER('db1')..IDENTIFIER('tbl1').

[iffyio]

Yeah was thinking a test as in your example to demonstrate if/how the parser handles it. If its unsupported then it'll be a negative test

[iffyio]

LGTM! Thanks @yoavcloud!
cc @alamb