About Column Projection in PXF

A newer version of this documentation is available. Click here to view the most up-to-date release of the Greenplum 5.x documentation.

PXF supports column projection, and it is always enabled. With column projection, only the columns required by a SELECT query on an external table are returned from the external data source. This process can improve query performance, and can also reduce the amount of data that is transferred to Greenplum Database.

Note: Some external data sources do not support column projection. If a query accesses a data source that does not support column projection, the query is instead executed without it, and the data is filtered after it is transferred to Greenplum Database.

Column projection is automatically enabled for the pxf external table protocol. PXF accesses external data sources using different connectors, and column projection support is also determined by the specific connector implementation. The following PXF connector and profile combinations support column projection on read operations:

  • PXF Hive Connector, HiveORC profile
  • PXF JDBC Connector, Jdbc profile
  • PXF Hadoop and Object Store Connectors, hdfs:parquet, adl:parquet, gs:parquet,s3:parquet, and wasbs:parquet profiles

Note: PXF may disable column projection in cases where it cannot successfully serialize a query filter; for example, when the WHERE clause resolves to a boolean type.

To summarize, all of the following criteria must be met for column projection to occur:

  • The external data source that you are accessing must support column projection. For example, Hive supports column projection for ORC-format data, and certain SQL databases support column projection.
  • The underlying PXF connector and profile implementation must also support column projection. For example, the PXF Hive and JDBC connector profiles identified above support column projection, as do the PXF connectors that support reading Parquet data.
  • PXF must be able to serialize the query filter.