feat: path expressions inside filter of scoped queries

[patricebender]

Support path expressions in infix filter conditions along a FROM clause.

• Filter conditions at from.ref.at(-1) (only for to-one assocs):

  • SELECT: Render an INNER JOIN for each association traversal and attach it to the main query's FROM clause.
  • UPDATE/DELETE: Render an INNER JOIN for each association traversal and attach it to the subquery's FROM clause. Transform UPDATE/DELETE statements with path expressions into UPDATE … WHERE <Primary Key(s)> IN (<subquery>).

• Filter conditions along from.ref (excluding the leaf ref.where):

  • SELECT/UPDATE/DELETE: Render an INNER JOIN for each association traversal and attach it to the corresponding EXISTS <subquery>.

[stewsk]

What about books that don't have a genre? Should they also appear in the result?

[patricebender]

hmm now that I think about it… it should have the same semantics as if you would write the expression into the queries .where clause. For this case, also books without a genre will be part of the result set due to the left join. What do you think?

[stewsk]

There should be a test where there is a to-many association in the filter of the last path step.
Simply translating that to a JOIN would not be right, in my opinion, as this would potentially increase the size of the result set rather than reducing it (this is a mistake that we have already made when dealing with to-many assocs in the WHERE clause

[patricebender]

There should be a test where there is a to-many association in the filter of the last path step. Simply translating that to a JOIN would not be right, in my opinion, as this would potentially increase the size of the result set rather than reducing it (this is a mistake that we have already made when dealing with to-many assocs in the WHERE clause

👍 having an inner join and not an left outer join for all filter path expressions should take care of that, shouldnt it?

[stewsk]

having an inner join and not an left outer join for all filter path expressions should take care of that, shouldnt it?

no, concerning "duplication" there is no difference between INNER and LEFT OUTER