CSV文件复制到保留了文件名的HDFS

Question

Most probably it's a dummy question but I can't find an appropriate example.

Please correct my way of thinking if I'm lost. To proceed with that I created a job and stream to trigger it:

job create hdfsjob 
    --definition "filepollhdfs --directory=/tmp/xd --names=*" --deploy
stream create csvStream 
    --definition "file --ref=true --dir=/data/in --pattern=*.txt 
    > queue:job:hdfsjob" --deploy

I can manage the file name and extension through fileName and fileExtension in the xd source filepollhdfs but how can I preserve the original name of it?

To prevent rollover is it a good idea to set the rollover parameter to a very high value, or is it something that can ruin memory consumption for example?

Ideal would be to travel through the source directory structure and recreate the same structure in hdfs.