Populating TTree with 2 open files fails in PyROOT but works in C++

[jbattat]

I’m reading values from one TTree (in an existing file), computing a new value to write to another TTree (and save in a new file). The Fill() call on the second tree does not work (when done in PyROOT), but works fine in C++.

Here’s the C++ code:

// Read in the ROOT file, compute a quantity and write it to a "Friend" tree                                                                 
TFile* f1 = new TFile("test.root");
TTree* t1 = (TTree*)f1->Get("t1");
t1->Scan();

double n;
double m;
t1->SetBranchAddress("val1", &n);
t1->SetBranchAddress("val2", &m);

TFile* f2 = new TFile("testfriend.root", "RECREATE");
TTree* t2 = new TTree("t1f", "a friend tree");

double w;
t2->Branch("val3", &w, "val3/D");

Int_t nn = t1->GetEntries();
for (int ii=0; ii<nn; ii++) {
  t1->GetEvent(ii);
  w = n+m;
  cout << n << " + " << m << " = " << w	<< endl;
  t2->Fill();
}

t2->Scan();

with output

************************************
*    Row   *      val1 *      val2 *
************************************
*        0 *         2 *         3 *
*        1 *         4 *         6 *
*        2 *         6 *         9 *
************************************
2 + 3 = 5
4 + 6 = 10
6 + 9 = 15
************************
*    Row   *      val3 *
************************
*        0 *         5 *
*        1 *        10 *
*        2 *        15 *
************************

(notice that val3 has the expected contents: 5, 10, 15).

When I do the analagous thing in PyROOT, I get all zeros for the val3 entries in the new TTree:

import ROOT
import numpy as np
f1 = ROOT.TFile('test.root')
t1 = f1.Get('t1')
t1.Scan()

f2 = ROOT.TFile('testfriend.root', 'RECREATE')
t2 = ROOT.TTree('t1f', 'a friend tree')
w = np.zeros(1, dtype='float')
t2.Branch('val3', w, 'val3/D')

for event in t1:
    w = event.val1 + event.val2
    print "%d + %d = %d" % (event.val1, event.val2, w)
    t2.Fill()

t2.Scan()

gives

************************************
*    Row   *      val1 *      val2 *
************************************
*        0 *         2 *         3 *
*        1 *         4 *         6 *
*        2 *         6 *         9 *
************************************
2 + 3 = 5
4 + 6 = 10
6 + 9 = 15
************************
*    Row   *      val3 *
************************
*        0 *         0 *
*        1 *         0 *
*        2 *         0 *
************************

Notice how the val3 entries in the new TTree are all zero.

[wlav]

w[0] = event.val1 + event.val2