Problem with TF1 pointer

ROOT
1

what i intended to do is to have one master fit function (in this case the fit_func_x). For every new idx at jdx=0, i call the master fit function to do the fit, then clone the master function to proceed with the rest of jdx. Therefore, we should see at every jdx=0 the pointer of the master function is being called, and another pointer for jdx = 2~4 (different for all idx, 2nd output in my posting).

void debug(){
    vector <int> Filters = {6,4}; //{1,2,3,4,6};
    int Ni= Filters.size();

    vector <double> EndPoints = {40,150,350,550,650};
    int Nj = EndPoints.size();

    // TF1* fit_func_x = new TF1(Form(func_name_x, muonID, 0, 0), func_2freq_constA_constW, ta, tb, init_vals_x.size());
    TF1* fit_func_x = new TF1("func_00", "[0] + [1]*cos([2]*x + [3]) + [4]*cos([5]*x + [6])",30,40);
    fit_func_x->SetNpx(500000);
    fit_func_x->SetLineColor(2);


    // GRid scan !
    cout << "    scanning x graphs ... " << endl; //debug

    for(int idx=0; idx<Ni; idx++){  // loop through fittiing of a graph at particular pts/Tc. clone the func at the beginning (8pts)
        int pts = int(48/Filters[idx]);

        TF1* fit_func_x_working=nullptr;
        cout << "    ptr of master func (every muonID, we use a new func): "<< fit_func_x  << endl; //debug

        for(int jdx=0; jdx<Nj; jdx++){ // for the same graph of pts, we study the change in fitting range
            double endpt = EndPoints[jdx];

            // cout << Form("        %d_%.0f",pts,endpt)<< endl;

            if(jdx == 0){ 
                fit_func_x_working = fit_func_x;
            };

            // I skip fitting codes here ...

cout << Form("        %d_%.0f. name : %s ; ptr is: (1 != 2=3=4=...)",pts,endpt, fit_func_x_working->GetName()) << fit_func_x_working  << endl; //debug
            cout <<  "       "<< fit_func_x_working->GetName() << endl; // debug

            if(jdx == 0){ // clone the state after the first fit. we want to maintain the state of the master func to the next pts (idx)
                fit_func_x_working = (TF1*) fit_func_x->Clone(); //Form("%d_%d",idx,jdx)
                // cout << fit_func_x_working << endl;
            };
            
        };

        delete fit_func_x_working;         
}
}

with the last line delete fit_func_x_working, i got the following which is not expected:

ptr of master func (every muonID, we use a new func): 0x55a09cb5f330
        8_40. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09cb5f330
        8_150. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09c5f9eb0
        8_350. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09c5f9eb0
        8_550. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09c5f9eb0
        8_650. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09c5f9eb0
    ptr of master func (every muonID, we use a new func): 0x55a09cb5f330
        12_40. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09cb5f330
        12_150. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09c5f9eb0
        12_350. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09c5f9eb0
        12_550. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09c5f9eb0
        12_650. name : func_00 ; ptr is: (1 != 2=3=4=...)0x55a09c5f9eb0

when i commented that line, i got expected result instead,

ptr of master func (every muonID, we use a new func): 0x561560a74e00
        8_40. name : func_00 ; ptr is: (1 != 2=3=4=...)0x561560a74e00
        8_150. name : func_00 ; ptr is: (1 != 2=3=4=...)0x56156055eba0
        8_350. name : func_00 ; ptr is: (1 != 2=3=4=...)0x56156055eba0
        8_550. name : func_00 ; ptr is: (1 != 2=3=4=...)0x56156055eba0
        8_650. name : func_00 ; ptr is: (1 != 2=3=4=...)0x56156055eba0
    ptr of master func (every muonID, we use a new func): 0x561560a74e00
        12_40. name : func_00 ; ptr is: (1 != 2=3=4=...)0x561560a74e00
        12_150. name : func_00 ; ptr is: (1 != 2=3=4=...)0x561560ab4dd0
        12_350. name : func_00 ; ptr is: (1 != 2=3=4=...)0x561560ab4dd0
        12_550. name : func_00 ; ptr is: (1 != 2=3=4=...)0x561560ab4dd0
        12_650. name : func_00 ; ptr is: (1 != 2=3=4=...)0x561560ab4dd0

please help. omit fit quality for now .

ROOT Version: 6.26.00 (conda)
Platform: centos7
Compiler: gcc9

2

Hi @NgJunKai,

You’re deleting the pointer fit_func_x_working while you’re still in the larger loop: for(int idx=0; idx<Ni; idx++), so that’s why you get the same memory address twice, i.e. 0x55a09c5f9eb0 in your exact example above.

Since you’re not initialising your fit_func_x_working pointer with new you don’t need to delete it. You should delete the fit_func_x pointer before closing the final curly bracket (your example is missing this one final bracket as well).

Cheers,
Marta

3

sorry for incomplete description and typo in the code (already updated).
using the same pointer for jdx=2,3,4 is what i am trying to do. but what concerned is this pointer of (jdx=2,3,4) should not be the same across idx ,which is something similar to my 2nd output.

4

so doing this without the delete will not cause memory leakage ?
ps: i am actually using fit_func_x_working as a temporary “container” to switch between the master function and its clone. i am not sure about the validity of doing so…

5

Hi,

yes, I understood that.

The pattern you want is in a way:

A
A
B
B
B
B
A
A
C
C
C
C

The issue appears when you have both

TF1* fit_func_x_working=nullptr;
and
delete fit_func_x_working;

if you don’t initialize to nullptr but do TF1* fit_func_x_working instead and then delete the pointer it will also work as expected.

Cheers,
Marta

6

tried the following combination. seems like the only thing that matters is the delete line …

A = TF1* fit_func_x_working=nullptr; ?
B = delete fit_func_x_working; ?

A | B | result
0 0 expected
0 1 not
1 0 expected
1 1 not

7

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.