Sets partition paths that flow partitions data by. Can be empty. Partition path can be a simple string column (e.g. "color") or a path pointing nested string field (e.g. "user.address.postcode"). Partition path is used to extract data from the entity and to create a tree of subdirectories for partitioned files. Using aforementioned partitions effects in creation of (example) following tree:

../color=blue
      /user.address.postcode=XY1234/
      /user.address.postcode=AB4321/
  /color=green
      /user.address.postcode=XY1234/
      /user.address.postcode=CV3344/
      /user.address.postcode=GH6732/

Take note:

• PartitionBy must point a string field.
• Partitioning removes partition fields from the schema. Data is stored in name of subdirectory instead of Parquet file.
• Partitioning cannot end in having empty schema. If you remove all fields of the message you will get an error.
• Partitioned directories can be filtered effectively during reading.

Adds a handler after record writes, exposing some of the internal state of the flow. Intended for lower level monitoring and control.

Please note that the handler is invoked after each input element is processed and not after each write. It is so because postWriteHandler may produce multiple records for a single input element.