Speaker
Dr
J. LIST
(University of Wuppertal)
Description
Analyses in high-energy physics often involve the filling of large
amounts of histograms from n-tuple like data structures, e.g. RooT
trees. Even when using an object-oriented framework like RooT, a the
user code often follows a functional programming approach, where
booking, application of cuts, calculation of weights and
histogrammed quantities and finally the filling of the histogram is
performed separately in different places of the program.
We will present a set of RooT based histogram classes that allow to
define the histogrammed quantity, its weight and the cuts to be
applied at the time of booking.
We use lightweight function object classes to define plotted
quantities and cut conditions; the "self-filling" histograms hold
references to these objects, and evaluate them in a fill method that
thus needs no parameters. The use of function objects rather than
strings to define plotted quantities and cuts permits error
detection at compile rather than run time, and allows the
implementation of caching mechanisms if costly computations are to
be performed. Arithmetic and logical expressions are implemented by
operator overloading. Histograms can be grouped in collections. We
apply the visitor pattern to perform operations like filling,
writing, fitting or attribute setting on such a group, without
having to extend the collection class each time a new functionality
is needed.
Although developed within the object oriented analysis framework of
the H1 experiment, this toolkit can be used on any RooT tree.
