scala - Spark union weird behavior -

i' m experiencing strange behavior union method of rdd class. , can't understand why happens.

i have class, in there data , filtereddata vars. first 1 initialized textfile , other map , filter, second instead sc.emptyrdd[point]. have method this:

do{   /**   * here val lastfiltered computed   */    logger.debug("filtered "+lastfiltered.count()+" points")    filtereddata=filtereddata.union(lastfiltered)    logger.debug("filtered far "+filtereddata.count()+" points")    data = data.subtract(lastfiltered)    /** here data repartitioned **/ }while(/** here there condition equivalent lastfiltered.count() == 0 **/) logger.info("preprocessing has filtered "+filtereddata.count()+" points") 

what i'm getting logger very strange me:

 filtered 13 points  filtered far 13 points  filtered 4 points  filtered far 4834 points  filtered 0 points  filtered far 0 points  preprocessing has filtered 0 points 

of course first 2 lines expected...but later looks strange me. moreover, subtract method on data rdd seems work fine (the counts expected).

can me in understanding happening?

thank you! marco