feat: support "replace using" for insert DML

[nlpersimon]

To fix #7949

[geooo109]

We should also add support for:

REPLACE ON boolean_expression

[geooo109]

@nlpersimon any plans to finish this?

[nlpersimon]

@nlpersimon any plans to finish this?

yes. appreciate your comment!
I will implement the support for "replace on" and address the other comments you left by this week.

[georgesittas]

closing this for now, feel free to reopen when ready.

[nlpersimon]

Hi @georgesittas @geooo109 , I am ready, could you reopen this PR? Or should I open a new PR?

[georgesittas]

@nlpersimon go for it

[nlpersimon]

@geooo109 I implemented the support for "replace on" in 1480c2c . I thought it was a trivial work that I can just implement it as

 elif self._match(TokenType.ON):
     replace_on = self._parse_disjunction()
     if replace_on is None:
         self.raise_error("Expected a condition after REPLACE ON")

but this implementation cannot handle the following query

INSERT INTO TABLE students AS t
  REPLACE ON t.name <=> s.name
  (SELECT * FROM people) AS s

because the parser regards s.name (SELECT * FROM people) as an anonymous function.

The solution I came up with is implementing a method _parse_with_token_barrier to restrict the visibility of the parser so that the parser can parse disjunction of the "replace on" clause without misinterpreting the source subquery as an anonymous function.

[geooo109]

@nlpersimon thanks for trying to solve the ON case. This problem is more complex than it seemed. Let's revert the solution for the ON and support the rest. I will do a follow up in the future to resolve this.

[nlpersimon]

@nlpersimon thanks for trying to solve the ON case. This problem is more complex than it seemed. Let's revert the solution for the ON and support the rest. I will do a follow up in the future to resolve this.

reverted. thank you!