TKey, TObject and GetListOfKeys for only newest ttrees

[jwsmith]

Hi there,

I have the requirement that I have root files with a large number of trees that I need to loop over and insert a branch in each (systematics trees).
The problem arises if there are multiples trees of the same name (probably due to some autosave feature outside of my control), i.e.
treeA
treeA:1
…

I don’t want to loop and save both of them, so how do I just get the newest ones?

At the moment I have something like

     // Do a loop over all trees that are not nominal
      TList* list = oldFile->GetListOfKeys() ;
      if (!list) { printf("<E> No keys found in file\n") ; exit(1) ; }
      TIter next(list) ;
      TKey* key ;
      TObject* obj ;

      while ( key = (TKey*)next() ) {

        obj = key->ReadObj() ;
        // skip if not a TTree or sumWeights TTree
        if ( (strcmp(obj->IsA()->GetName(),"TTree")!=0) || (strcmp("sumWeights",obj->GetName()) == 0)) {
          printf("Not running over: %s \n",obj->GetName()); continue;
        }

        newtree = fChain->CloneTree(0);  // some stuff with tree
      }

but this will loop over ALL trees.

Thanks in for the help!
Josh

[Wile_E_Coyote]

Aux;1 aux;2 aux3;

[pcanal]

Try with something like:

   TString oldkeyname;

         while ( (key = (TKey*)nextkey())) {

            // Keep only the highest cycle number for each key for mergeable objects. They are stored
            // in the (hash) list consecutively and in decreasing order of cycles, so we can continue
            // until the name changes. We flag the case here and we act consequently later.
            bool alreadyseen = (oldkeyname == key->GetName());

            if (alreadyseen) continue;
           
           alreadyseen = key->GetName();
          
           ....
      }

Cheers,
Philippe.

[jwsmith]

Thanks for this, Philippe.

However, after testing, it seems alreadyseen will always evaluate to false given tree names I use, i.e.

treeA
treeA:1

[pcanal]

I do not understand. What is the :1 suffix? ‘key->GetName()’ returns the name of the TTree without the cycle and should be self consistent (i.e. alreadyseen is assigned the value of key->GetName() and then it is compared with the same).

Cheers,
Philippe.

[jwsmith]

The trees look like the attached.
Yes, I’m also surprsed this isn’t working.

The loop looks like this. At no point do I ever reach “ERROR”.

 // Do a loop over all trees that are not nominal
      TList* list = oldFile->GetListOfKeys() ;
      if (!list) { printf("<E> No keys found in file\n") ; exit(1) ; }
      TIter next(list) ;
      TKey* key ;
      TObject* obj ;


      std::string oldkeyname;
      while ( (key = (TKey*)next()) ) {

        bool alreadyseen = (oldkeyname == key->GetName());

        std::cout <<alreadyseen <<std::endl;
        if (alreadyseen){
        std::cout << "ERROR"<<std::endl;
        continue;
        }
        alreadyseen = key->GetName();
        TChain *fChain=nullptr;
        TTree *newtree=nullptr;

        obj = key->ReadObj() ;
        if ( (strcmp(obj->IsA()->GetName(),"TTree")!=0) || (strcmp("sumWeights",obj->GetName()) == 0)
          || (strcmp("nominal",obj->GetName()) == 0) || (strcmp("nominal_Loose",obj->GetName()) == 0)) {
          printf("Not running over: %s \n",obj->GetName()); continue;
        }

        fChain = new TChain(obj->GetName());

        fChain->Add((file).c_str());

        newtree = fChain->CloneTree(0);
       if(fChain->GetEntries() == 0){
         std::cout<<"No events, skipping"<<std::endl;
         continue;
       }
        m_add_branches(fChain,newtree,m_neuralNet); // some external function

        newfile->cd();
        newtree->Write();

        delete newtree;
        delete fChain;

      } // end loop over trees

Cheers,
Josh

[pcanal]

There was a fatal bug in my code snippet
Replace

alreadyseen = key->GetName();

with

oldkeyname = key->GetName();

Cheers,
Philippe.

[jwsmith]

Wow. I should have seen that, sorry!
Thanks, it’s working now.
Cheers,
Josh