Package org.apache.druid.query.groupby.epinephelinae.column

Interface DimensionIdCodec<DimensionType>

Type Parameters:
DimensionType - Type of the dimension holder
All Known Implementing Classes:
DictionaryBuildingGroupByColumnSelectorStrategy.UniValueDimensionIdCodec
public interface DimensionIdCodec<DimensionType>
Dimension to integer id encoder - decoder i.e. it is an interface for converters of dimension to dictionary id and back. It only handles single value dimensions. Handle multi-value dimensions (i.e. strings) using the KeyMappingMultiValueGroupByColumnSelectorStrategy.

Encoding
The caller is expected to handle non-null values. Null values must be filtered by the caller, and assigned GroupByColumnSelectorStrategy.GROUP_BY_MISSING_VALUE 1. DimensionIdCodec extracts the dimension from the selector 2. The value gets encoded into a dictionaryId, using lookupId(DimensionType) 3. The callers can use this integer dictionaryID to represent the grouping key

Decoding
Converts back the dictionaryId to the dimension value. The implementations are not expected to handle GroupByColumnSelectorStrategy.GROUP_BY_MISSING_VALUE. The callers should handle those values appropriately ontheir own, and filter those out before trying to convert the dictionary id back to value.

  • Method Details

    • lookupId

      MemoryFootprint<Integer> lookupId(DimensionType dimension)
      Returns:
      DictionaryId of the object at the given index and the memory increase associated with it

    • idToKey

      DimensionType idToKey(int id)

    • canCompareIds

      boolean canCompareIds()
      Returns if the object comparison can be optimised directly by comparing the dictionaryIds, instead of decoding the objects and comparing those. Therefore, it returns true iff the "dict" function defined by dict(id) = value is monotonically increasing. Ids backed by dictionaries built on the fly can never be compared, therefore those should always return false.

    • reset

      void reset()