Page 1 of 1

Transformer erasing predecessor sort information

Posted: Sat Sep 27, 2014 9:25 am
by zulfi123786
Hi,

I am reading a sorted sequential file and have a couple of transformers which are using keychange detection functions. I have added a dummy sort over sort keys to mark that the data is sorted but it looks like the transfomer is erasing this information on the output link.

This makes sense as a column can be modified inside transformer, is there any way to avoid inserting dummy sort stages following every transformer ?

Thanks

Posted: Tue Sep 30, 2014 2:37 am
by priyadarshikunal
It should not get lost unless you have some operation on keys, are you re-partitioning etc in transformer, is the partitioning set to propagate etc. Check the score to see if its partitioning or doing something unnoticed from job design.

Posted: Tue Sep 30, 2014 6:37 am
by zulfi123786
Nothing done on the keys, Dont think the score would report information of sort order preservance across operators.

Posted: Wed Oct 01, 2014 4:11 am
by priyadarshikunal
score will report re-partition if happening before any transformer as you say. As it will partition data after reading from seq file anyways. So you have to have one sort operator just to mark that data is already sorted.