Package org.apache.drill.exec.vector.accessor.reader

Interface NullStateReader

All Known Implementing Classes:
NullStateReaders.AlwaysNullStateReader, NullStateReaders.ComplexMemberStateReader, NullStateReaders.ListIsSetVectorStateReader, NullStateReaders.MemberNullStateReader, NullStateReaders.NullableIsSetVectorStateReader, NullStateReaders.RequiredStateReader, NullStateReaders.TypeVectorStateReader
public interface NullStateReader
Internal mechanism to detect if a value is null. Handles the multiple ways that Drill represents nulls:
  • Required and repeated modes: value is never null.
  • Optional mode: null state is carried by an associated "bits" vector.
  • Union: null state is carried by both the bits state of the union itself, and the null state of the associated nullable vector. (The union states if the column value itself is null; the vector state is if that value is null, which will occur if either the column is null, or if the type of the column is something other than the type in question.)
  • List, with a single data vector: null state is carried by the list vector and the associated nullable data vector. Presumably the list vector state takes precedence.
  • List, with a union data vector (AKA variant array or union array): the null state is carried by all three of a) the list vector, b) the union vector, and c) the type vectors. Presumably, the list vector state has precedence.

The interface here allows each reader to delegate the null logic to a separate component, keeping the data access portion itself simple.

As with all readers, this reader must handle both the single-batch and the hyper-batch cases.

  • Method Details

    • bindIndex

      void bindIndex(ColumnReaderIndex rowIndex)

    • isNull

      boolean isNull()

    • bindBuffer

      void bindBuffer()