Re: Event/TTree reading

Dmitry,

If your events are so large that one single event does not fit in memory, you should make
one tree per event and call TTree::Fill for each Track. That's what we do in the case of Alice where a simulated event can be Gigabytes.

Rene Brun

Dmitry Naumov wrote:
> Hi Rene,
>
> let me be more precise. My Event has a structure like this:
> class Event {
> <>
> private:
> Header *fHeader;
> Hits *fHits;
> <>
> }
>
>
> in the simulation chain (VMC ) we do something like this:
>
> // generate events: propagate tracks. fEvent is Event* object
> if ( fEvent->GetTracksNumber() > kMaxTracksNumber) {
> // reached limit: save the event, clear it, save to the TTree and
> continue simulation saving the information into the new entry of TTree
> tree->Fill();
> fEvent->Clear();
> }
>
> The output Tree is saved with Event as branch. As the result we have
> something like this in the output TTree:
>
> entry 0 Event 0
> entry 1 Event 1
> entry 2 Event 2 (event is splitted and continued in the next entries)
> entry 3 Event 2 (continued from entry 2)
> entry 4 Event 2 (continued from entry 2)
> entry 5 Event 3
> entry 6 Event 4
>
> and so on.
>
>
> Now if I want to read these events in the analysis code I have to
> handle somehow that entries 0,1,5,6 return me fully contained
> Event(s), while entries 2,3,4 belong to the same Event 2
> Therefore if I want say to fill two histos (one with energy of initial
> particle, another with energy released in the detector) I encounter a
> problem:
> for (int i(0); i<tree->GetEntries(); i++) {
> tree->GetEntry(i);
> h1->Fill(ev->GetHeader()->Energy());// this will fill the
> histogram three times with the same Event 2 (entries 2,3,4)
> h2->Fill(ev->GetHits()->GetTotalEnergyReleased()); // this will
> fill the histogram with smaller energy released because Event 2 is
> stored as three entries 2,3,4
>
> }
>
> Therefore I have to take care somehow about a correct handling of
> these events when I read them.
>
>
> Thank you for your considerations, Dmitry
>
>
> Rene Brun wrote:
>> Hi Dmitry,
>>
>> I am afraid that you will have to provide more information if you
>> want to get a meaningful answer.
>> Did you look at the friend trees. See TTree::AddFriend.
>>
>> Rene Brun
>>
>> Dmitry Naumov wrote:
>>> Dear Rooters,
>>>
>>> may I ask your advice on the following problem. We save simulated
>>> events into a TTree object. However sometimes the events are too
>>> large and we split them into pieces and save each piece as an entry
>>> of the TTree. This works OK. What would be a clever way to read
>>> these events from the TTree?
>>> A simple code which we had before (before splitting) is not useful
>>> anymore:
>>>
>>> // open the tree
>>> Event *ev=0;
>>> tree->SetBranchAddress("fEvent",&ev);
>>> do (int i(0); i<tree->GetEntries()l i++) {
>>> tree->GetEntry(i);
>>> // do something with ev: fill its global information (energy,
>>> position, etc), fill its hits
>>> }
>>>
>>>
>>> Following this way the concept of entry as an event is no more
>>> working, and we have to organize something like "SubEvent" and do
>>> cycle over ev->GetSubEvents() entries or so. This is the way I have
>>> in mind. However before trying to realize it it would be great if
>>> ROOT/C++ gurus could give an advice on that. [Building the whole
>>> event from different entries of TTree is not OK because the event
>>> might not fit into the memory].
>>>
>>> Thank you in advance, Dmitry
Received on Fri Dec 15 2006 - 15:45:03 MET