predictionio優化

predictionio 的validScores方法 有點問題，因為model對象中的一個屬性類型為Map<String, Item> items，在數據量很大的時候，肯定是不行，所以優化一下，采用join+filter的模式，代碼如下

return all.mapToPair(new PairFunction<ItemScore, String, ItemScore>() {
            @Override
            public Tuple2<String, ItemScore> call(ItemScore idUser) throws Exception {
                return new Tuple2<>(idUser.getItemEntityId(), idUser);
            }
        }).join(model.getItems()).filter(new Function<Tuple2<String,Tuple2<ItemScore,Item>>, Boolean>() {
            @Override
            public Boolean call(Tuple2<String,Tuple2<ItemScore,Item>> itemScore) throws Exception {
                /*Item item = items.get(itemScore.getItemEntityId());*/

                Item item =itemScore._2._2;
                //logger.info("join end");

                return (item!=null && /*item != null 
                        && */passWhitelistCriteria(whitelist, itemScore._2._1.getItemEntityId())
                        && passBlacklistCriteria(blacklist, itemScore._2._1.getItemEntityId())
                       /* && passCategoryCriteria(categories, item)*/
                       /* && passUnseenCriteria(seenItemEntityIds, itemScore.getItemEntityId())*///看過或買過的不要
                       /* && passAvailabilityCriteria(unavailableItemEntityIds, itemScore.getItemEntityId())*/);
            }
        }).map(new Function<Tuple2<String, Tuple2<ItemScore,Item>>, ItemScore>() {
             @Override
             public ItemScore call(Tuple2<String, Tuple2<ItemScore,Item>> userItemCount) throws Exception {
                 return userItemCount._2._1;
             }
         });