Package mondrian.rolap.cache

Interface SegmentCacheIndex

All Known Implementing Classes:
SegmentCacheIndexImpl
public interface SegmentCacheIndex
Data structure that identifies which segments contain cells.

Not thread-safe.

Author:
Julian Hyde

  • Method Details

    • locate

      List<SegmentHeader> locate(String schemaName, ByteString schemaChecksum, String cubeName, String measureName, String rolapStarFactTableName, BitKey constrainedColsBitKey, Map<String,Comparable> coordinates, List<String> compoundPredicates)
      Identifies the segment headers that contain a given cell.
      Parameters:
      schemaName - Schema name
      schemaChecksum - Schema checksum
      cubeName - Cube name
      measureName - Measure name
      rolapStarFactTableName - Fact table table
      constrainedColsBitKey - Bit key
      coordinates - Coordinates
      compoundPredicates - Compound predicates
      Returns:
      Empty list if not found; never null

    • findRollupCandidates

      List<List<SegmentHeader>> findRollupCandidates(String schemaName, ByteString schemaChecksum, String cubeName, String measureName, String rolapStarFactTableName, BitKey constrainedColsBitKey, Map<String,Comparable> coordinates, List<String> compoundPredicates)
      Returns a list of segments that can be rolled up to satisfy a given cell request.
      Parameters:
      schemaName - Schema name
      schemaChecksum - Schema checksum
      cubeName - Cube name
      measureName - Measure name
      rolapStarFactTableName - Fact table table
      constrainedColsBitKey - Bit key
      coordinates - Coordinates
      compoundPredicates - Compound predicates
      Returns:
      List of candidates; each element is a list of headers that, when combined using union, are sufficient to answer the given cell request

    • intersectRegion

      List<SegmentHeader> intersectRegion(String schemaName, ByteString schemaChecksum, String cubeName, String measureName, String rolapStarFactTableName, SegmentColumn[] region)
      Finds a list of headers that intersect a given region.

      This method is used to find out which headers need to be trimmed or removed during a flush.

      Parameters:
      schemaName - Schema name
      schemaChecksum - Schema checksum
      cubeName - Cube name
      measureName - Measure name
      rolapStarFactTableName - Fact table table
      region - Region
      Returns:
      List of intersecting headers

    • add

      void add(SegmentHeader header, SegmentBuilder.SegmentConverter converter, boolean loading)
      Adds a header to the index.
      Parameters:
      header - Segment header
      loading - Whether segment is pending a load from SQL
      converter - Segment converter

    • update

      void update(SegmentHeader oldHeader, SegmentHeader newHeader)
      Updates a header in the index. This is required when some of the excluded regions have changed.
      Parameters:
      oldHeader - The old header to replace.
      newHeader - The new header to use instead.

    • loadSucceeded

      void loadSucceeded(SegmentHeader header, SegmentBody body)
      Changes the state of a header from loading to loaded.

      The segment must have previously been added by calling add(mondrian.spi.SegmentHeader, mondrian.rolap.agg.SegmentBuilder.SegmentConverter, boolean) with a not-null value of the bodyFuture parameter; neither loadSucceeded nor loadFailed(mondrian.spi.SegmentHeader, java.lang.Throwable) must have been called.

      Informs anyone waiting on the future supplied with add(mondrian.spi.SegmentHeader, mondrian.rolap.agg.SegmentBuilder.SegmentConverter, boolean).

      Parameters:
      header - Segment header
      body - Segment body

    • loadFailed

      void loadFailed(SegmentHeader header, Throwable throwable)
      Notifies the segment index that a segment failed to load, and removes the segment from the index.

      The segment must have previously been added using add(mondrian.spi.SegmentHeader, mondrian.rolap.agg.SegmentBuilder.SegmentConverter, boolean) with a not-null value of the bodyFuture parameter; neither loadSucceeded(mondrian.spi.SegmentHeader, mondrian.spi.SegmentBody) nor loadFailed must have been called.

      Informs anyone waiting on the future supplied with add(mondrian.spi.SegmentHeader, mondrian.rolap.agg.SegmentBuilder.SegmentConverter, boolean).

      Parameters:
      header - Header
      throwable - Error message

    • remove

      void remove(SegmentHeader header)
      Removes a header from the index.
      Parameters:
      header - Segment header

    • printCacheState

      void printCacheState(PrintWriter pw)
      Prints the state of the cache to the given writer.
      Parameters:
      pw - Print writer

    • getFuture

      Future<SegmentBody> getFuture(Execution exec, SegmentHeader header)
      Returns a future slot for a segment body, if a segment is currently loading, otherwise null. This is the method to use to get segments 'hot out of the oven'.

      When this method is invoked, the execution instance of the thread is automatically added to the list of clients for the given segment. The calling code is responsible for calling cancel(Execution) when it is done with the segments, or else this registration will prevent others from canceling the running SQL statements associated to this segment.

      Parameters:
      header - Segment header
      Returns:
      Slot, or null

    • cancel

      void cancel(Execution exec)
      This method must remove all registrations as a client for the given execution. This must terminate all SQL activity for any orphaned segments.
      Parameters:
      exec - The execution to unregister.

    • contains

      boolean contains(SegmentHeader header)
      Tells whether or not a given segment is known to this index.

    • linkSqlStatement

      void linkSqlStatement(SegmentHeader header, Statement stmt)
      Allows to link a Statement to a segment. This allows the index to cleanup when cancel(Execution) is invoked and orphaned segments are left.
      Parameters:
      header - The segment.
      stmt - The SQL statement.

    • getConverter

      SegmentBuilder.SegmentConverter getConverter(String schemaName, ByteString schemaChecksum, String cubeName, String rolapStarFactTableName, String measureName, List<String> compoundPredicates)
      Returns a converter that can convert the given header to internal format.
      Parameters:
      schemaName - Schema name
      schemaChecksum - Schema checksum
      cubeName - Cube name
      rolapStarFactTableName - Fact table
      measureName - Measure name
      compoundPredicates - Compound predicates
      Returns:
      Converter

    • setConverter

      void setConverter(String schemaName, ByteString schemaChecksum, String cubeName, String rolapStarFactTableName, String measureName, List<String> compoundPredicates, SegmentBuilder.SegmentConverter converter)
      Sets a converter that can convert headers in for a given measure to internal format.
      Parameters:
      schemaName - Schema name
      schemaChecksum - Schema checksum
      cubeName - Cube name
      rolapStarFactTableName - Fact table
      measureName - Measure name
      compoundPredicates - Compound predicates
      converter - Converter to store