I'm using TTreeFormula to get around the problem of parsing a set of cuts sent to my TSelector via the outside world. See, I don't want to hard code in my cuts, which can be complicated expressions, in my TSelector. I want a generic TSelector which can be changed from my main program. Chris *** Yesterday, you (Rene Brun) wrote *** >Hi Chris, > >I do not think that it is a good idea to use TTreeFormula within a TSelector. >If you already have a TSelector derived class, you should implement >the function MySelector::ProcessCut. In this function, you should call >the TBranch::GetEntry (not TTree::GetEntry) in the order where you eliminate >as many candidates as possible and minimizing the I/O. >You can see an example in $ROOTSYS/tutorials/h1analysis.C > >Use TBranch::GetEntry as shown in this example instead of the mechanism >via TTree::SetBranchStatus. The pointers to the branches are already declared >and precomputed for you in mySelector.h > >Rene Brun > > >Chris Roat wrote: >> >> Hi Philipe, >> >> Thanks so much for your solution. It worked great and I'm now back up and >> running. I have another question on a related note. >> >> I need to be sure that my TSelector reads in the branches necessary (and >> *only* those branches, since I have 100s) for the TTreeFormulas I am >> using. But I don't see the need to add/change SetBranchStatus lines in my >> TSelector::Init method everytime I add/change my TTreeFormulas. >> >> Is there a way to do this more easily, like by looping over the leaves a >> TTreeFormula needs? >> >> Cheers, >> Chris >> >> *** Today, you (Philippe Canal) wrote *** >> >> >Hi Chris, >> > >> >The problem was that the TTreeFormula were not informed properly that >> >the chain loaded a need TTree. To fix that you need to add at the end of >> >generic::Notify the following: >> > >> > // fChain->LoadTree(fChain->GetReadEntry()); >> > // Also update the formulas: >> > for (Int_t cut=0; cut<_cs->nCuts(); cut++) { >> > ttform[cut]->UpdateFormulaLeaves(); >> > } >> > >> >Now in your sample case (only one file), it should not have been loading >> >any new file in the first place. The reason why it was, is that you have >> >in all your selector's Notify method (like generic::Notify) calls like: >> > >> > fChain->SetBranchStatus("*",kFALSE); >> > >> >This forces a reload of the current file. The SetBranchStatus actually >> >belongs in the generic::Init method because they only need to be once >> >by chain. The chain (at least in newer ROOT) keeps track of the >> >info and pass it to each TTree. If for some reasons (very old ROOT >> >version) this does not work, then uncomment the first like in my >> >first code fragment. >> > >> >Cheers, >> >Philipe. >> > >> > >> >-----Original Message----- >> >From: owner-roottalk@pcroot.cern.ch >> >[mailto:owner-roottalk@pcroot.cern.ch]On Behalf Of Chris Roat >> >Sent: Wednesday, December 19, 2001 1:03 PM >> >To: roottalk@pcroot.cern.ch >> >Subject: [ROOT] TTreeFormula within TSelector::ProcessCut >> > >> > >> >Hi ROOT, >> > >> >I'm having problems finding out if a particular entry in a tree passes a >> >cut. I'm using a TSelector which takes recieves a list of strings (which >> >are cuts, of the form "p>3" or "myBool" from the outside world). During >> >ProcessCut/ProcessFill, I would like to see if the entry I'm on passes any >> >of the strings. >> > >> >I am trying to use TTreeFormula to tackle this, but without prolonged >> >luck. It works fine on the first entry - EvalInstance(0) returns the >> >correct answer, so I'm on the right track. However, EvalInstance(0) seg >> >faults on subsequent entries. >> > >> >Could there be a problem with ProcessCut() and EvalInstance() both getting >> >info from the tree? >> > >> >Thanks for any insight, >> >Chris >> > >> > >> > >
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:51:13 MET