Histograms as Objects
Paper: 192
Session: A (talk)
Speaker: Adesanya, Adeyemi, CERN, Geneva
Keywords: analysis, C++, class libraries, data presentation
Histograms as Objects
Y.Adesanya
CERN/CN/ASD
Abstract
In the survey of software needed for computing in the LHC era, we did
not find software appropriate for our needs in the area of histogramming.
We started a project to develop a histogramming package. Requirements
gathered from several users governed the analysis and design of the
package. Among the requirements were support for arbitrary dimension of
histograms and profile histograms, filling of various types of data ( e.g.
not only real numbers), correct handling of statistics, extendable design
and a coherent user interface. Ntuples were explicitly excluded from the
package. Persistent storage of histograms shall be handled via an object
database.
The analysis resulted in a separation of the identification of the bin the
data point belongs to from the actual accumulation of data points in the
histogram. Thus the histogram consists of an object cumulating the data
points according to a partition (binning) object for each coordinate.
Various histograms may share the same partition objects.
The implementation in C++ is in progress. Objects created can be stored
in the Objectivity ODBMS. Modules to display histograms in the framework
of IRIS Explorer are being developed in parallel. A first version has
been given to few users.
We will report on the status of this project.