Steps to reproduce:
1. Add a S3 connection
2. Select a file from S3 in dataprep browser and wrangle
3. Create a pipeline from the workspace
4. Once in studio make path, accessID and accessKey as macros
5. Add a TPFSAvro sink and publish the pipeline
5. Post publish run the pipeline with appropriate values for macros.
After Step 5 the user should have seen data copied to a CDAP dataset but it fails with NPE. Have attached the trace for reference.