How to save TGraphErrors in a tree?

Vladislav · October 26, 2015, 10:34am

Hello!

I want to save TGraphErrors in the tree, but my method don’t work.

First of all, I explain you what I want.
I want to make fit of TGraphErrors object and save TGraphErrors (object) and fit parameters (double) in the tree.

I know how to save TGraphErrors objects only:

TMultiGraph *mg = new TMultiGraph();
TGraphErrors *gr = new TGraphErrors(xv.size(), &xv[0], &yv[0], &xverr[0], &yverr[0]);
mg->Add(gr);
...
TObjArray Hlist(0);
Hlist.Add(mg);

I know how to save double values only:

TNtuple ntuple("ntuple", "fit results", "x0:y0:chi2");
...
ntuple.Fill(x0, y0, chi2);
TFile f_ntuple(string_ntuple.c_str(), "RECREATE");
ntuple.Write();
f_ntuple.Close();

I have tried the next code in order to save both: graph and doubles:

void CreateTree()
{
	TFile f("D:\\Data_work\\tree.root","recreate");
	TTree t1("t1","a simple Tree with simple variables");
	
	Double_t x0, y0, amp;
	
	//create vectors
	vector<double> xv;
	vector<double> yv;
	vector<double> xverr;
	vector<double> yverr;
	
	//loop to create several graphs
	for(int j = 0; j < 10; j++)
	{	
		//fill vectors		
		for (Int_t i= -10; i<= 10; i++) 
		{
			xv.push_back(i);
			yv.push_back(0.2 * i*i + gRandom->Gaus(0, 1));
			xverr.push_back(0);
			yverr.push_back(1);
		}
		
		TGraphErrors *gr = new TGraphErrors(xv.size(), &xv[0], &yv[0], &xverr[0], &yverr[0]);
		gr->SetMarkerColor(4);
		gr->SetMarkerStyle(kFullCircle);
		
		TF1 *fitFcn = new TF1("fitFcn", fitFunction, -10, 10, 3);
		gr->Fit("fitFcn", "R");
		
		x0 = fitFcn->GetParameter(2);
		y0 = fitFcn->GetParameter(1);
		amp = fitFcn->GetParameter(0);
		
		//cout << "x0 = " << x0 << " y0 = " << y0 << " amp = " << amp << endl;
		//gr->Draw("AP");
		
		t1.Branch("x0",&x0,"x0/D");
		t1.Branch("y0",&y0,"y0/D");
		t1.Branch("amp",&amp,"amp/D");
		t1.Branch("gr","TGraphErrors",&gr, 128000, 0);
		
		t1.Fill();
		
		xv.clear();
		yv.clear();
		xverr.clear();
		yverr.clear();
	}
	
	
	t1.Write();
}

But the result is differs from my expectation: I can’t open graph object like 1) method.

prntscr.com/8vgg9f (via TObjArray); picture just for example
prntscr.com/8vggw1 (via TNtuple )
my attempt (via TTree) prntscr.com/8vghs2

I expected to find in the tree the array of these graphs prntscr.com/8vgisx

Thanks for advance.

dpiparo · October 26, 2015, 8:48pm

Hi,

to write a tree containing per entry a TGraphErrors and the values of the parameters of a linear function, a starting point which you could adapt to your case is:

/// Just to make all graphs different from each other
void fillRandomAndFit(TGraphErrors& g){
   for (int i = 0;i<10;++i){ 
      g.SetPoint(i, i + gRandom->Uniform(-.5,.5), gRandom->Uniform(-.5,.5));
   }
   g.Fit("fitFunc", "N");
}

void treeOfTGraphs() {

	TGraphErrors g(10);
	TF1 fitFunc("fitFunc","[0]+[1]*x",0,10);
        double m,q;
      
	TFile ofile("myFile.root","RECREATE");
	TTree tree("treeOfGraphs","TreeOfGraphs");
	tree.Branch("graph",&g);
	tree.Branch("q",&q);
	tree.Branch("m",&m);


    for (int i=0;i<100;++i){
       fillRandomAndFit(g);
       q = fitFunc.GetParameter(0);
       m = fitFunc.GetParameter(1);
       tree.Fill();
    }

    tree.Write();
    ofile.Close();
}

Cheers,
Danilo

Vladislav · October 27, 2015, 6:06am

Hello!

Thanks for the answer!
But I can’t understand the principial differense between our programms.

Ok, I try to explain in more details my problem.

For example, I have x vs y dependency.
I create TGraphErrors object and fit if.
I will have the vector of parameters pv, chi2 value and TGraphErrors object.
I save all these in the tree.
Then I read the tree and cut a subset:

	TChain *chain = new TChain("ntuple");
	chain->Add("D:\\Data_work\\myFile.root");
	TCut cut = "p1 > 0.2 && chi2 < 1";

I can Draw the correlation between parameters i and j:

chain->Draw("p1:p2", cut);

I want to Draw several TGraphErrors from this subset.
How to do this?

Thank you in advance.

dpiparo · October 27, 2015, 6:34am

There is no difference: mine is a skeleton to show the way of achieving what you ask in a general and minimal way.

This is is too much to ask to TTree::Draw
In this case, you can loop on the entries yourself and draw the graphs for those that pass your selection.

Cheers,
Danilo

Vladislav · October 27, 2015, 7:13am

Danilo, I am very grateful to you for your help.

I don’t have enough experiense with trees.
Could you give me the simple example how to get graph from the tree and organize the loop?

Thank you in advance.

dpiparo · October 27, 2015, 8:23am

Hi Vladislav,

no problem.
So, the documentation can be found here: root.cern.ch/root/htmldoc/guide … ding-trees
For your particular case, since you are not using ROOT6 (it did not make it yet on Windows), you can ignore TTreeReader: this class does not exist in ROOT5.
To be more concrete, to read your tree and apply a cut, you need to do:
in python

import ROOT
f = ROOT.TFile("myFile.root")
for e in f.treeOfGraphs:
   if e.q > 0.1: continue
   # Here the work 
raw_input("Press a key to continue.")

in C++

void readTree(){
   TFile f("myFile.root");
   TTree* t;
   f.GetObject("treeOfGraphs",t);
   TGraphErrors* graph = new TGraphErrors();
   double q;
   t->SetBranchAddress("graph",&graph);
   t->SetBranchAddress("q",&q);
   for (int i=0;i<t->GetEntries();++i){
      t->GetEntry(i);
      if (q>.1) continue;
      cout << "q is " << q << endl;
      // do work
      graph->Dump();
   }

}

Cheers,
Danilo

Vladislav · October 27, 2015, 1:26pm

Danilo, your links are very useful.

But I still have two questions. Maybe you can help me.

I did not undestand what this code do:

Does it just print in concole all properties of the object?

I have understood how to realize my desire: I shoud save my graphs into the new .root file.

void readTree()
{
	TObjArray Hlist(0);

	TFile f("D:\\Data_work\\myFile.root");
	TTree* t;
	f.GetObject("treeOfGraphs",t);

	TGraphErrors* graph = new TGraphErrors();
	double q;
	t->SetBranchAddress("graph",&graph);
	t->SetBranchAddress("q",&q);
	  
	for(int i = 1; i < t->GetEntries(); ++i)
	{
		t->GetEntry(i);
		Hlist.Add(graph); 
	}   

	TFile ofile_Hlist("D:\\Data_work\\graphs.root", "RECREATE");
	Hlist.Write();
	ofile_Hlist.Close();   
}

But all graphs in the graphs.root are equal, so there is an error.
Is it right code in the loop?

Hlist.Add(graph);

Thank you in advance.

dpiparo · October 27, 2015, 2:17pm

This is a place holder to show you that the graphs are properly deserialised
The code was saving N time the last graph read. The solution is

void readTree2()
{
   TObjArray Hlist(0);
   Hlist.SetOwner(kTRUE);

   TFile f("myFile.root");
   TTree* t;
   f.GetObject("treeOfGraphs",t);

   TGraphErrors* graph = new TGraphErrors();
   double q;
   t->SetBranchAddress("graph",&graph);
   t->SetBranchAddress("q",&q);
    
   for(int i = 1; i < t->GetEntries(); ++i)
   {
      t->GetEntry(i);
      Hlist.Add(new TGraphErrors(*graph));
   }   

   TFile ofile_Hlist("graphs.root", "RECREATE");
   Hlist.Write();
   ofile_Hlist.Close();   
}

Vladislav · October 28, 2015, 5:24am

Danilo, thank you very much!