Package mondrian.rolap.agg

Class Aggregation

java.lang.Object
mondrian.rolap.agg.Aggregation
public class Aggregation extends Object
A Aggregation is a pre-computed aggregation over a set of columns.

Rollup operations:

  • drop an unrestricted column (e.g. state=*)
  • tighten any restriction (e.g. year={1997,1998} becomes year={1997})
  • restrict an unrestricted column (e.g. year=* becomes year={1997})

Representation of aggregations. Sparse and dense representations are necessary for different data sets. Should adapt automatically. Use an interface to hold the data set, so the segment doesn't care.

Suppose we have a segment {year=1997, quarter={1,2,3}, state={CA,WA}}. We want to roll up to a segment for {year=1997, state={CA,WA}}. We need to know that we have all quarters. We don't. Because year and quarter are independent, we know that we have all of the ...

Suppose we have a segment specified by {region=West, state=*, year=*}, which materializes to ({West}, {CA,WA,OR}, {1997,1998}). Because state=*, we can rollup to {region=West, year=*} or {region=West, year=1997}.

The space required for a segment depends upon the dimensionality (d), cell count (c) and the value count (v). We don't count the space required for the actual values, which is the same in any scheme.

Since:
28 August, 2001
Author:
jhyde

  • Constructor Details

    • Aggregation

      public Aggregation(AggregationKey aggregationKey)
      Creates an Aggregation.
      Parameters:
      aggregationKey - the key specifying the axes, the context and the RolapStar for this Aggregation

  • Method Details

    • getCreationTimestamp

      public Date getCreationTimestamp()
      Returns:
      Returns the timestamp when the aggregation was created

    • load

      public void load(SegmentCacheManager cacheMgr, int cellRequestCount, RolapStar.Column[] columns, List<RolapStar.Measure> measures, StarColumnPredicate[] predicates, GroupingSetsCollector groupingSetsCollector, List<Future<Map<Segment,SegmentWithData>>> segmentFutures)
      Loads a set of segments into this aggregation, one per measure, each constrained by the same set of column values, and each pinned once.

      A Column and its constraints are accessed at the same level in their respective arrays.

      For example, 

       measures = {unit_sales, store_sales},
 state = {CA, OR},
 gender = unconstrained
      Parameters:
      segmentFutures - List of futures wherein each statement will place a list of the segments it has loaded, when it completes

    • optimizePredicates

      public StarColumnPredicate[] optimizePredicates(RolapStar.Column[] columns, StarColumnPredicate[] predicates)
      Drops predicates, where the list of values is close to the values which would be returned anyway.

    • getStar

      public RolapStar getStar()
      This is called during SQL generation.

    • getConstrainedColumnsBitKey

      public BitKey getConstrainedColumnsBitKey()
      Returns the BitKey for ALL columns (Measures and Levels) involved in the query.