/* 

  Actor's award in Picat.

  From "Figaro Tutorial", page 19f
  """
  There are three classes: actors, movies, and appearances relating actors to movies.
  Whether an actor receives an award for an appearance depends on the fame of the actor 
  and the quality of the movie.
  """

  This is a little messy since we have to memoize the values
  of the random variables for each run. 
  (We cannot use table since that it global.)

  Cf my Gamble model gamble_actors_award.rkt

  This program was created by Hakan Kjellerstrand, hakank@gmail.com
  See also my Picat page: http://www.hakank.org/picat/

*/

import ppl_distributions, ppl_utils.
import util.
% import ordset.

main => go.

/*

  Num accepted samples: 100  Total samples: 18956  (0.005%)

  var : famous actor1
  Probabilities:
  false: 0.8600000000000000
  true: 0.1400000000000000
  mean = [false = 0.86,true = 0.14]

  var : famous actor2
  Probabilities:
  false: 0.8700000000000000
  true: 0.1300000000000000
  mean = [false = 0.87,true = 0.13]

  var : famous actor3
  Probabilities:
  true: 1.0000000000000000
  mean = [true = 1.0]

  var : quality movie 1
  Probabilities:
  medium: 0.5500000000000000
  low: 0.2500000000000000
  high: 0.2000000000000000
  mean = [medium = 0.55,low = 0.25,high = 0.2]

  var : quality movie 2
  Probabilities:
  high: 1.0000000000000000
  mean = [high = 1.0]

  var : award appearance 1
  Probabilities:
  false: 0.9399999999999999
  true: 0.0600000000000000
  mean = [false = 0.94,true = 0.06]

  var : award appearance 2
  Probabilities:
  false: 0.7300000000000000
  true: 0.2700000000000000
  mean = [false = 0.73,true = 0.27]

  var : award appearance 3
  Probabilities:
  true: 0.6700000000000000
  false: 0.3300000000000000
  mean = [true = 0.67,false = 0.33]

  var : prob award appearance 1
  Probabilities:
  0.01: 0.5300000000000000
  0.001: 0.2200000000000000
  0.05: 0.1900000000000000
  0.2: 0.0600000000000000
  mean = 0.02702

  var : prob award appearance 2
  Probabilities:
  0.05: 0.8700000000000000
  0.2: 0.1300000000000000
  mean = 0.0695

  var : prob award appearance 3
  Probabilities:
  0.2: 1.0000000000000000
  mean = 0.2

*/
go ?=>
  reset_store,
  run_model(10_000,$model,[show_probs_trunc,mean,
                           % show_percentiles,show_histogram,
                           % show_hpd_intervals,hpd_intervals=[0.94],
                           min_accepted_samples=100,show_accepted_samples=true
                          ]),
  nl,
  % show_store_lengths,nl,
  % fail,
  nl.
go => true.


famous(Actor) = flip(0.1).

quality(Movie) = categorical([0.3,0.5,0.2],[low,medium,high]).

prob_award([Actor,Movie],Quality,Famous) = Res =>
  QualityMovie= Quality.get(Movie),
  FamousActor = Famous.get(Actor),
  T = [QualityMovie,FamousActor],
  Res = case(T,
             [ [[low,false], 0.001],
                [[low,true], 0.01],
                [[medium,false], 0.01],
                [[medium,true], 0.05],
                [[high,false], 0.05],
                [[high,true], 0.2],
                [true,0.999]
                ]).

award(ActorMovie,ProbAward) = flip(ProbAward.get(ActorMovie)).

model() =>
  Appearance1 = [actor1,movie1],
  Appearance2 = [actor2,movie2],
  Appearance3 = [actor3,movie2],
  Appearances = [Appearance1,Appearance2,Appearance3],

  % Memoizing the values.
  % We have to do that to ensure that all call to - say - famous(actor1)
  % give the same result both in the observe and then in the presentation.
  % Unfortunately this makes it a little more messy than necessarily...
  % 
  Famous    = new_map([A=famous(A) : A in [actor1,actor2,actor3]]),
  Quality   = new_map([M=quality(M) : M in [movie1,movie2]]),
  ProbAward = new_map([A=prob_award(A,Quality,Famous) : A in Appearances]),  
  Award     = new_map([A=award(A,ProbAward) : A in Appearances]),


  observe(Famous.get(actor3) == true),
  observe(Quality.get(movie2) == high),  

  % Ensure that exactly one appearance get an award
  observe(1 == [cond(Award.get(A) == true,1,0) :  A in Appearances].sum),
  if observed_ok then
    add("famous actor1",Famous.get(actor1)),
    add("famous actor2",Famous.get(actor2)),
    add("famous actor3",Famous.get(actor3)),
    add("quality movie 1",Quality.get(movie1)),
    add("quality movie 2",Quality.get(movie2)),
    add("award appearance 1",Award.get(Appearance1)),
    add("award appearance 2",Award.get(Appearance2)),
    add("award appearance 3",Award.get(Appearance3)),   
    add("prob award appearance 1",ProbAward.get(Appearance1)),
    add("prob award appearance 2",ProbAward.get(Appearance2)),
    add("prob award appearance 3",ProbAward.get(Appearance3)),    
  end.