When using the CouchDB Input step with the Adaptive Execution Layer, the following factor affects performance and results:
- Spark processes null values differently than the Pentaho engine. You will need to adjust your transformation to successfully process null values according to Spark's processing rules.