Package org.apache.drill.exec.vector.accessor.writer.dummy

This package provides a "dummy" set of writers.

See: Description

Class Summary

Class	Description
DummyArrayWriter	Dummy scalar array writer that allows a client to write values into the array, but discards all of them.
DummyArrayWriter.DummyOffsetVectorWriter
DummyDictWriter
DummyScalarWriter	Represents a non-projected column.

Package org.apache.drill.exec.vector.accessor.writer.dummy Description

This package provides a "dummy" set of writers. The dummy writers provide the same API as the "real" writers, but the dummy writers simply discard their data. The dummy writers are used when implementing projection: non-projected columns may still have to be processed (as in a CSV file, say), but their values are not needed. One way to do this is to do an if-statement for each value:

 if (column-a-is-projected) {
   aWriter.setSomething(value);
 }

The dummy writers convert the if-statement into a virtual function call, same as is done to handle the type-specific nature of vectors:

 aWriter.setSomething(value);
 

The theory is that the virtual function dispatch is simpler, and faster, than doing continual if-checks everywhere in the code.

The dummy writers reside in this package so that the various factory methods can automatically build the dummy versions when given a null value vector (which we then interpret to mean that there is no physical backing to the column.)

At present, most methods that return a value simply return zero or null. Experience will show whether it is worthwhile implementing some basics, such as a value type or index. For now, these return null, assuming that the caller won't do anything with the column other than set a value.

Some simpler dummy writers appear as nested classes inside the "real" writers.