Interface ExternalSplit

All Superinterfaces:
NamedWriteable, Writeable
All Known Implementing Classes:
CoalescedSplit, FileSplit

public interface ExternalSplit extends NamedWriteable
A serializable, parallelizable unit of work for an external data source. Each split represents a portion of data (e.g., a file slice, a partition) that can be read independently by a single driver and shipped across nodes.
  • Method Details

    • sourceType

      String sourceType()
    • estimatedSizeInBytes

      default long estimatedSizeInBytes()
    • splitStats

      @Nullable default SplitStats splitStats()
      Returns per-split statistics for use by the optimizer (aggregate pushdown, filter pruning). Returns null if statistics are unavailable for this split.

      Implementations that carry row-group or stripe metadata should override this method. FileSplit returns its compact SplitStats when present. CoalescedSplit returns a MergedSplitStats that lazily aggregates its children's statistics.