A mapping is run in high precision mode in a Hadoop environment for Hive 0.11 and above.
In a Hadoop environment, sources that have data errors in a column result in a null value for the column. In the native environment, the Data Integration Service does not process the rows that have data errors in a column.
When you cancel a mapping that reads from a flat file source, the file copy process that copies flat file data to HDFS may continue to run. The Data Integration Service logs the command to kill this process in the Hive session log, and cleans up any data copied to HDFS. Optionally, you can run the command to kill the file copy process.
When you set a limit on the number of rows read from the source for a Blaze mapping, the Data Integration Service runs the mapping with the Hive engine instead of the Blaze engine.