Class FloatEquiDepthHistogram

  • All Implemented Interfaces:
    Serializable, Cloneable

    public class FloatEquiDepthHistogramextends PersistentObject
    Read-only equi-depth histogram for selectivity estimation. Assume you have collected statistics over a data set, among them a one-dimensional equi-depth histogram (quantiles). Then an applications or DBMS might want to estimate the selectivity of some range query [from,to], i.e. the percentage of data set elements contained in the query range. This class does not collect equi-depth histograms but only space efficiently stores already produced histograms and provides operations for selectivity estimation. Uses linear interpolation.

    This class stores a list l of float values for which holds:

  • Let v be a list of values (sorted ascending) an equi-depth histogram has been computed over.
  • Let s=l.length.
  • Let p=(0, 1/s-1), 2/s-1,..., s-1/s-1=1.0) be a list of the s percentages.
  • Then for each i=0..s-1: l[i] = e : v.contains(e) && v[0],..., v[p[i]*v.length] <= e .
  • (In particular: l[0]=min(v)=v[0] and l[s-1]=max(v)=v[s-1].)
See Also:
Serialized Form

SCaVis 2.2 © jWork.ORG