Impure Infix Clauses are removed via column pruning

[deusaquilus]

Version: (e.g. 3.4.4-SNAPSHOT)
Module: (e.g. quill-sql)
Database: (e.g. ALL)

The current behavior of column pruning removes a column from all queries and subqueries where it is not used. For example:

case class Person(id:Int, name:String, other:Int)
run { query[Person].map(p => (p.id, p.name, p.other)).map(t => (t._1, t._2)) }
// SELECT p.id, p.name FROM person p

Note that this behavior happens even with nested queries:

run { query[Person].map(p => (p.id, p.name, p.other)).nested.map(t => (t._1, t._2)) }
// SELECT p.id, p.name FROM (SELECT x.id, x.name FROM person x) AS p

The same behavior occurs with infixes:

run { query[Person].map(p => (p.id, p.name, infix"foobar${p.other}")).map(t => (t._1, t._2)) }
// SELECT p.id, p.name FROM person p

While this kind of behavior should be fine for pure infixes, infixes such as rank and distinct on totally break by this behavior:

run { query[Person].map(p => (infix"DISTINCT ON (${p.other})".as[Int], p.name, p.id)).map(t => (t._2, t._3)) }
// SELECT p.name, p.id FROM person p

In general, it seems that for impure-infixes, the assumption that the infix clause can be pruned with the containing column is incorrect. The SQL normalization needs to be changed accordingly.

@getquill/maintainers

[deusaquilus]

Fixed via #1597