H2root and variable length array

ROOT
1

Dear rooters,
I have the following problem: I have a PAW ntuple which contains among the other variables these two:

ntrack integer

al(5,ntrack) float

I want to translate this ntuple into a rootple. h2root creates a rootple which contains

Ntrack integer

Al[Ntrack][5] float

For many reasons (after long discussions with other collaborators) we would prefer to have in the rootple something like

ntrack integer

al[5][ntrack] float

that is the same as above but with slightly different names AND with the correct order of columns and rows in the matrix. So I have written some code which open the ntuple read the content and try to write in the “correct” way a rootple. To fill the rootple I am using a structure like:

struct trklev {
Int_t ntrk;
Float_t al[5][50]; // ntrk always less than 50
}

and I book the rootple this way

struct trklev trklev2;
TFile *hfile=0;
TTree *tree=0;
hfile = new TFile(filename,“RECREATE”,“my data”);
tree = new TTree(“TrkLev2”,“my data”);
tree->Branch(“ntrk”,&trklev2.ntrk,“ntrk/I”);
tree->Branch(“al”,trklev2.al,“al[5][ntrk]/F”);

the program works (doesn’t crash) and I am sure I have in trklev2.al all the correct values (just to try I have fixed them) but when I look at the rootple content (with ntrk=1 for the moment) I find that only al[0][0] is correct while al[1][0]…al[4][0] contain random (?) numbers (like 3.503e-44, 8.579e-39, 7.553e+31,…).
I have tried to modify my program inverting rows with columns in the matrix and that worked perfectly, but it is not what I need.
My question is: is it possible to save a matrix which has variable length columns and not rows? why do h2root switch rows and columns? is it because fortran and C use the inverted convention for rows and columns or because it is not possible to do otherwise?

Many thanks,
Emiliano Mocchiutti (for the PAMELA collaboration)

2

Emiliano,

h2root converts your table from the Fortran convention to the C/C++
storage convention. We have no plans to change this.

Rene

3

Dear Rene,
I understand you will not change h2root, perhaps I did not explain well my problem.
A part h2root, is there any intrinsic reason for which I cannot create a branch in a tree which contains a matrix which has a variable length number columns instead of rows?
I mean, if I do this way

struct trklev trklev2;
TFile *hfile=0;
TTree *tree=0;
hfile = new TFile(filename,“RECREATE”,“my data”);
tree = new TTree(“TrkLev2”,“my data”);
tree->Branch(“ntrk”,&trklev2.ntrk,“ntrk/I”);
tree->Branch(“al”,trklev2.al,“al[ntrk][5]/F”);

it works, variable “al” contains the correct values.
If I do instead

struct trklev trklev2;
TFile *hfile=0;
TTree *tree=0;
hfile = new TFile(filename,“RECREATE”,“my data”);
tree = new TTree(“TrkLev2”,“my data”);
tree->Branch(“ntrk”,&trklev2.ntrk,“ntrk/I”);
tree->Branch(“al”,trklev2.al,“al[5][ntrk]/F”); <— different here!

it doesn’t work, variable “al” contains random numbers.
If you tell me it is the way ROOT fills matrixes and cannot be change I am happy with the answer, if you know it is possible to do it my question is how can I do it, since if I do this way it doesn’t work! I am using ROOT Version 4.03/02 9 February 2005.
Thanks anyway for your answer!
Emiliano