Fast and easy storing of histograms (std::map or else?)

ROOT
1

In order to manage the large amount of histograms, I am using a
map:

This allows me to store histograms by name

AND saving them using a loop.

Since I have more than one Monte Carlo type of data I end up creating

which leads to string manipulations (which are not as easy in C++):

for (short int i = 0; i < mcsize; i++) {
		string histname;
			histname = mclabel[i] + "_" + "neutrino_pz";
		hists_[histname.c_str()] = new TH1F(histname.c_str(), "reconstructed p_z(#nu)", 50, -500, 500);
		histname = mclabel[i] + "_" + "neutrino_pz_mc";
		hists_[histname.c_str()] = new TH1F(histname.c_str(), "reconstructed p_z(#nu)", 50, -500, 500);;
		histname = mclabel[i] + "_" + "recoWmass";
		hists_[histname.c_str()] = new TH1F(histname.c_str(), "reconstructed leptonic W mass", 50, 0, 500);
	}
for ( std::map<string, TH1F*>::const_iterator iter = hists_.begin();
	      iter != hists_.end(); ++iter )
		iter->second->Sumw2();
}

Now I would like to store them as:

to store them as

This would be fine, but since I heard that maps are SLOW in C++ and
use a lot of memory,
I wonder if there is a better method to do so.

Is TMap much better?
For this I would only need to use TString instead of std::string, right?

Thank you in advance.

2

Hi,

I’d use the following approach: define two enums:

[code]enum EMCType {
kWhatever,
…
kNumMCTypes
};
enum EHist {
kNeutrinoPz,
…
kNumHists
};
const char* mcnames[] = { “Whatever”, … }
const char* histnames[] = { “neutrino_pz”, … }

vector<vector<TH1*> > hists;
hists.resize(EMCType);
for (int mctype = 0; mctype < kNumMCTypes; ++mctype)
hists[mctype].resize(kNumHists);
for (int mctype = 0; mctype < (int) kNumMCTypes; ++i)
for (int ihist = 0; ihist < kNumHists; ++ihist)
hist[mctype][ihist] = new TH1F(TString(mcnames[mctype]) + histnames[ihist], …);
[/code]

You will get rid of all the string manipulations to access the histograms, and you use vectors instead of maps which makes it much faster.

Cheers, Axel.

3

Hi Alex, thanks for your response.

So your approach means, I can access them via:

Not using strings for index, but still clear names.
And the best of all: since the current implementation is already a
vector, I don’t need to change much to start working with it.

Cheers,
Luke