Double reads in BAMInputFormat

[serash]

I used the BAMInputFormat to read a bam file in Spark:

val tmprdd = sc.newAPIHadoopFile(input, classOf[BAMInputFormat], classOf[LongWritable], classOf[SAMRecordWritable], conf_).map(_._2.get())

And the number of entries isn't correct. In this RDD I had 222 more reads than in the original BAM file. It seems that these reads are all doubles, exactly the same read appears twice in the RDD.