using namespace RooFit ;

void ex9()
{
  // Make an empty workspace that exports its contents to CINT
  RooWorkspace *w = new RooWorkspace("w",kTRUE) ;  

  // *** Declare model parameters ***
  
  // Generally a variable is declared A[lo,hi], A[initval,lo,hi], or A[constval]
  //
  // Here S is the expected Signal event count
  // Here B is the expected Background event count

  w->factory("S[0,0,100]") ; // allowed range [0,100] initial value 0
  w->factory("B[5]") ;       // constant at value 5

  // Declare the model observable (i.e. the quantity we measure, in this case an event count)
  w->factory("N[0,100]") ;

  // *** Construct an expression for the expected signal yield ***
  

  // Generally a function object is created as FunctionType::FunctionName(arguments)
  //
  // A very useful one is FunctionType 'expr' which allows a TFormula-style C++ expression
  // to be interpreted as its value
  // 
  //   expr::FunctionName('<formula_expression>',inputvariable1,inputvariable2)
  //
  // Construct a function named Nexp that adds parameters S and B
  w->factory("expr::Nexp('S+B',S,B)") ;  
  

  // Finally construct a probability model (a special class of functions that can present probabilities
  // by the virtue that int ProbModel(x) dx == 1 [ unitarity requirement for probability interpretation]
  //
  // Generally, all FunctionTypes that exist map one-to-one to existing RooFit function classes
  // In this case the class RooPoisson is used (you can look at it in $ROOTSYS/roofit/roofit/inc/RooPoisson.h
  // and $ROOTSYS/roofit/roofit/src/RooPoisson.cxx) 
  //
  // This class implements the Poisson distribution and has one parameter (here represented by our function expression Nexp
  // as function of model parameters S(floating) and B(constant)
  w->factory("Poisson::model(N,Nexp)") ;


  // Have a look at all the objects created in the workspace
  w->Print("t") ;

  // *** STEP 1 ***
  return ;

  // Now we have a look at our model 

  // Make a an empty plot frame with N on the X-axis with range [0,100]
  RooPlot* frame = w->var("N")->frame(0,20,20) ;
  
  // Now plot our model on top the plot frame ;
  w->pdf("model")->plotOn(frame,Precision(1e-4)) ;
  

  // ( The Precision(1e-4) improves the adaptive plotting
  //   precision which is needed here because the Poisson
  //   has discontinuities, whereas the plotting code
  //   is generally optimized for continuous functions )

  // This draws the plot frame on the canvas
  frame->Draw() ;

  // *** STEP 2 ***
  //return ;

  // Now we change the value of expected signal yield from 0 to 2
  w->var("S")->setVal(2) ;

  // And we plot again, on the same plot frame
  w->pdf("model")->plotOn(frame,Precision(1e-4),LineColor(kRed)) ;

  // This draws the plot frame on the canvas
  frame->Draw() ;

}