Re: [rules-users] [planner] Drools planner & Scala => Solution.getScore():Score<?>
Hi Adam,
Michael Neale (the drools guvnor guy) is also using Scala and Drools
Planner together. His code is at github.
He doesn't use such a hack (maybe he's still at drools 5.0.1 where it
was a double instead of Score). Why doesn't he?
If the issue still makes sense, feel free to open an issue for the
org.drools.planner.core.solution.Solution#getScore():Score<?>
discussion (and copy this part of the discussion).
Why should it be Score<?> and not
2) Score<S extends Score>
3) or Score<<S extends Score<S extends Score> >> ?
4) Or maybe it should be, just like EnumSet?
class Solution<S extends Score> {
public S getScore()
}
I don't like to fiddle trial&error with generics just "so it gives less
compiler issues". Only when the other alternatives are proven worse
(especially by referring to JDK class examples/conventions) and if it
makes sense, I 'll be happy to apply a fix and adjust the examples and
UpgradeRecipe. Better yet: a patch to do that is welcome too (if the
discussion proves it to be the best option).
With kind regards,
Geoffrey De Smet
Adam Warski schreef:
Hello,
I'm using Planner in a Scala program and it almost works without problems; the only
one is in implementing the Solution interface, as it uses raw (erased) version of Score,
which is normally generic (Score<S extends Score>). Scala generally doesn't like
raw types and in most cases it's impossible to extend/implement interfaces which use
them. A work-around is to introduce an "adapter" in Java, which you can later
extend in Scala:
public abstract class ScalaSolution implements Solution {
public Score getScore() { return getScoreScala(); }
public abstract Score<?> getScoreScala();
public void setScore(Score score) { setScoreScala(score); }
public abstract void setScoreScala(Score<?> score);
}
and then in Scala:
class Schedule(...) extends ScalaSolution {
var score: Score[_] = _
def setScoreScala(score: Score[_]) = { this.score = score }
def getScoreScala = score
...
}
So the fix is to replace the raw Score type with Score<?>. Maybe it would be
possible to apply that in the code? I tried it locally and it compiles without problems.
And a side question: do you have any plans on implementing multithreading in planner? So
that several moves could be explored concurrently? I saw there's an open Jira issue
for some time, but it didn't have any timeframe information.