If I read from dataflow source, say cosmosDB, and branch the source data into two streams of dataset based on filter conditions, does dataflow reads the source twice to apply two filter conditions ? As dataflow uses spark, I believe unless you save the source data in spark cluster, the read has to happen twice. Does ADF dataflow saves intermediate transformations as well ?
If you Add the same source twice using two source transformation , then only it will read the source data twice like below:
If you use conditional split, or filter transformation, and split the data, it wont read it twice. It will consider it as single source