How to dynamically build a filter expressions with reduce(_ and_) about scanamo HOT 2 CLOSED

@tolomaus sorry about that, the two algebras were designed at different times I think. It should be easy but actually is not, first off the type of Array.reduce forces the result type to be a super-type of the element type, which is why it works for updates. Second, AndCondition takes two type parameters. So

filter(0) and filter(1):
  AndCondition[Contains, Contains]

but

filter(0) and filter(1) and filter(2):
  AndCondition[AndCondition[Contains, Contains], Contains]

so, to write a reduce that works for this, you will need to go beyond simple arrays and functions, something like (not tested):

trait ConditionReduce[Xs <: HList] {
  type Z
  def reduceAnd(xs: Xs): Z
}
object ConditionReduce {
  implicit val hNilReduce: ConditionReduce[HNil] = new ConditionReduce[HNil] {
    type Z = True.type
    def reduceAnd(xs: HNil): Z = True // we don't have `True` but it could be written
  }
  implicit def hConsReduce[X, Xs](implicit X: ConditionExpression[X], Xs: ConditionReduce[Xs]): ConditionReduce[X ::: Xs] =
    new ConditionReduce[X ::: Xs] {
      type Z = AndCondition[X, Xs.Z]
      def reduceAnd(xs: X ::: Xs): Z = xs.head and Xs.reduceAnd(xs.tail)
    }
}

This is ugly as hell but given how all the conditions are defined, it is the only way currently. Ha, you will also need a proof that Z has a ConditionExpression so the above is probably not enough. So as you can see, the library needs a bit of a makeover there 😄

from scanamo.

Thanks for your take on this @regiskuckaertz ! I'm not too familiar with the advanced scala features like cats, shapeless etc so I'm afraid I have reached my limits here ;-)

Ideally scanamo would support all possible dynamodb features (like the reduce), but in reality this will not always be possible. In these cases my current workaround is to manipulate the underlying QueryRequest directly.

from scanamo.