For a detailed introduction to the analysis framework see this page. The framework provides classes and basic tools for general-scope analysis. In order to create an analysis, one has to define some xml configuration files that drive the cut flow. The framework provides a tool called Jigsaw that can load a plugin (see below) which auto-configures and run your analysis. Standalone analyses are still possible for your convenience, but this approach is not suggested at the beginning.

Hands on

A successful test has been performed on Bologna Tier3 nodes.
ssh -X ${USER}@uibo-atlas-01.cr.cnaf.infn.it
bsub -q T3_BO -Is bash

In this case, it is strongly recommended to setup first the GCC environment first:

. /afs/cern.ch/sw/lcg/external/gcc/4.3.2/x86_64-slc5-gcc43-opt/setup.sh
Change x86_64 to your needs (i686, ia32).

I assume a directory tree like this:

We will download the code in $HOME/development. The installation will put the headers, some tools and the libraries in $HOME/local/ subdirectories.

Remember to add $HOME/local to your path. With BASH you can do it this way:

export PATH=$HOME/local/bin:$HOME/local/share:${PATH}

Library Dependencies

The framework needs a number of libraries. Most of them are found inside ATHENA so we strongly suggest to setup the ATHENA environment first. (read the ATLAS Workbook first if you don't know what we're talking about!). For example:
export AtlasSetup=/afs/cern.ch/atlas/software/dist/AtlasSetup
alias asetup='source $AtlasSetup/scripts/asetup.sh'
asetup 17.0.3

Non-ATHENA Installation

Go to $HOME/analysis and download the needed packages IF NEEDED:

First install FastJet ( website )

cd $HOME/src
wget http://www.lpthe.jussieu.fr/~salam/fastjet/repo/fastjet-2.4.4.tar.gz
tar -xzvf fastjet-2.4.4.tar.gz
cd fastjet-2.4.4
./configure --prefix=$HOME/local --enable-allplugins
make install

Now install BAT: Bayesian Analysis Toolkit (website)

cd $HOME/src
wget http://www.mppmu.mpg.de/bat/source/BAT-0.4.3.tar.gz
tar -xzvf BAT-0.4.3.tar.gz
cd BAT-0.4.3
./configure --prefix=$HOME/local
make install


cd $HOME/src
svn co svn+ssh://YourCERNUserID@svn.cern.ch/reps/atlasgrp/Institutes/Goettingen/KLFitter/tags/KLFitter-00-05-00 KLFitter #or newer
cd KLFitter
Interlude: you have to modify the KLFitter library/Makefile and extras/Makefile. Add /include to the definition of BATCFLAGS this way
Bear in mind that this may change according to your environment. You might also want to change BATINSTALLDIR for the ATHENA environment, e.g.:
export BATINSTALLDIR=$ATLAS_EXTERNAL/BAT/0.3.2_root5.28.00a/i686-slc5-gcc43/

Now go on with the compilation:

make extras

Now make a symbolic link to libKLFitter.so:

ln -s $HOME/src/KLFitter/library/libKLFitter.so $HOME/local/lib/.
ln -s $HOME/src/KLFitter/extras/libKLFitterExtras.so $HOME/local/lib/.

If you prefer, you can add this directory to $PATH instead:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/src/KLFitter/library

In any case set:

export KLFITTERROOT=$HOME/src/KLFitter

Installing the Jigsaw Analysis Framework

You may need to adjust Makefile.inc according to your environment (e.g. check some path variabiles such as PREFIX or FASTJET ). Now you can download and compile our framework:
cd $HOME/analysis
svn co svn+ssh://YourCERNUserID@svn.cern.ch/reps/atlasgrp/Institutes/Bologna/AnalysisFramework/trunk Jigsaw
cd Jigsaw

No configure script is provided (yet). You have to export KLFITTERROOT and to set the variable PREFIX according to your environment while issuing make (you can also modify it for good in Makefile.inc). In our case we will keep the default:

export KLFITTERROOT=$(HOME)/src/KLFitter

Now you can compile the framework:

make -j<NumberOfCores, eg 16> PREFIX=$(HOME)/local
export ANALYSISFRAMEWORKROOT=${HOME}/local #or your modified PREFIX

The installation process added a library, a tools and the headers to $HOME/local$ subdirectories. Make sure that they are there! It is important that those paths are present in $LD_LIBRARY_PATH$, otherwise the system will not recognize the analysis framework.


We provide a skeleton for your analyses.

cd $HOME/analysis
svn co svn+ssh://YourCERNUserID@svn.cern.ch/reps/atlasgrp/Institutes/Bologna/TutorialAnalysis/trunk TutorialAnalysis
cd TutorialAnalysis

You can modify makeLinks2Datasets.sh or just create a directory containing symbolic links to your datasets. In our example try to run:

Jigsaw -a Tutorial -n 1000 -d /gpfs_data/storm/atlas/localgroupdisk/user/rsoualah/mc10_7TeV/user.rsoualah.mc10_7TeV.105860.TTbar_PowHeg_Jimmy.merge.NTUP_TOP.e600_s933_s946_r2302_r2300_p572.28July_v5_0.110801015827/ -w Rachik
The -d option specifies the directory containing the input ROOT files, while the -f option specifies a text file containing a list of data files to be processed (useful when running in parallel on a cluster). The -p option specifies the configuration file. The default one is share/Analysis.xml. Some other examples are present for your convenience.

If everything goes smoothly, you should see something like this:

File contains 9992 events
Initialize tree for file no. 0
100 / 1000(10%)
200 / 1000(20%)
300 / 1000(30%)
400 / 1000(40%)
500 / 1000(50%)
600 / 1000(60%)
700 / 1000(70%)
800 / 1000(80%)
900 / 1000(90%)
Finalizing analysis...
Final statistics for selection ELE:
                    All events ( 0)       1000 (      1000)
                     Dummy Cut ( 1)       1000 (prev: 1.00) (eff: 1.00) (unw:       1000) 
                EF_e15_medium  ( 2)        329 (prev: 0.33) (eff: 0.33) (unw:        329) 
            # pvx (4 trk) >= 1 ( 3)        329 (prev: 1.00) (eff: 0.33) (unw:        329) 
                         SL_el ( 4)        186 (prev: 0.57) (eff: 0.19) (unw:        186) 
                  MET > 35 GeV ( 5)        135 (prev: 0.73) (eff: 0.14) (unw:        135) 
                 W_mT > 25 GeV ( 6)        119 (prev: 0.88) (eff: 0.12) (unw:        119) 
                          4j25 ( 7)         71 (prev: 0.60) (eff: 0.07) (unw:         71) 
                 1 bTag > 5.85 ( 8)         50 (prev: 0.70) (eff: 0.05) (unw:         50) 
                  Jet cleaning ( 9)         50 (prev: 1.00) (eff: 0.05) (unw:         50) 

Cut flow ELE dumped to histogram ELE_cutflow

Final statistics for selection MUON:
                    All events ( 0)       1000 (      1000)
                     Dummy Cut ( 1)       1000 (prev: 1.00) (eff: 1.00) (unw:       1000) 
EF_mu10_MSonly EF_mu13 EF_mu13_tight  ( 2)        350 (prev: 0.35) (eff: 0.35) (unw:        350) 
            # pvx (4 trk) >= 1 ( 3)        350 (prev: 1.00) (eff: 0.35) (unw:        350) 
                         SL_mu ( 4)        208 (prev: 0.59) (eff: 0.21) (unw:        208) 
                  MET > 20 GeV ( 5)        197 (prev: 0.95) (eff: 0.20) (unw:        197) 
             MET+W_mT > 60 GeV ( 6)        184 (prev: 0.93) (eff: 0.18) (unw:        184) 
                          4j25 ( 7)        114 (prev: 0.62) (eff: 0.11) (unw:        114) 
                 1 bTag > 5.85 ( 8)         86 (prev: 0.75) (eff: 0.09) (unw:         86) 
                  Jet cleaning ( 9)         86 (prev: 1.00) (eff: 0.09) (unw:         86) 

Cut flow MUON dumped to histogram MUON_cutflow

Histograms have been saved in a file called by default BaselineAnalysis.root. You can change its name using the -o <OutputFileName.root> option.

And Now?

Now that we make sure that Jigsaw runs on your system, you are free to create your own analysis. We set up three ways to do this.

Modify TutorialAnalysis

You can tweak TutorialAnalysis according to your needs, adding new event manipulators and histogram fillers. This analysis should work correctly just out-of-the-box. The makefiles provided by us should be smart enough to recognize and compile them if their names contain *EventManipulator* and *HistogramFiller*. Two not-too-basic examples are provided.

cd $HOME/analysis
svn co svn+ssh://YourCERNUserID@svn.cern.ch/reps/atlasgrp/Institutes/Bologna/TutorialAnalysis/trunk MyAnalysis
cd MyAnalysis

Create a Stub Analysis

Just issue Jigsaw's analysis-creation script (in this example it is called TTbar):
cd $HOME/analysis
jigsaw-createCutBasedAnalysis.py TTbar
cd TTbarAnalysis
The script creates the following files in TTbarAnalysis:

As usual, this stub analysis should run out of the box. We will add some scripts for the creation of event manipulators and histogram fillers asap. Stay tuned!

Run Default CutBasedAnalysis

If you don't need to perform special operations such as event-per-event dump, there is no need at all to code an analysis library. This part is still in development and not all the features are enabled yet. Anyway, if your analysis is based on default event manipulators and histogram fillers, you just neeed to provide the xml configuration files (usually cardfiles/Analysis.xml and share/histograms.xml ). Just run:
#> Jigsaw --params-xml cardfiles/Analysis.xml  -nevtmax 100

[14:48:56] INFO: Config: Running on 100 events
[14:48:56] INFO: Jigsaw: Found WrapperNtuplePlugin: libSTDHEPWrapperNtuplePlugin.so / STDHEPWrapperNtuplePlugin
[14:48:56] INFO: Found custom ntuple wrapper STDHEPWrapperNtuplePlugin
[14:48:56] INFO: Jigsaw: Found Analyzer: libBlackHoleRemnantsAnalyzer.so / BlackHoleRemnantsAnalyzer
[14:48:56] INFO: Plugin loaded...
Initializing BlackHoleRemnantsAnalyzer

How to Deal with Runtime Options

The most important configuration file is the one defining the runtime behaviour of your analysis. The default name is share/Analysis.xml (take a look). The basic syntax is:

<analysis name="MyAnalysis" >

    <schedule hfiller="FillEventWeight" />
    <schedule hfiller="FillFinalStateObjects" />

<cutflow key="ELE">
  <cut type="trigger" L1="" L2="" EF="EF_e20_medium" />
  <cut type="leptoncut" ptmin=25. etamax=2.5 n=1 flavor="ele" > 
  <cut type="jetcut" n=2 ptmin=25 />


In many cases, you will run on ROOT ntuples. To speed up the execution time, you can specify in a separate cardfile (e.g. share/MyWrapper.xml) the branches you want to activate for a given wrapper. You could also assign more than one configuration to the same wrapper: the same wrapper could be useful for real data, data/MC comparison or a dedicated truth Monte Carlo study. In this case, the config file looks like this:

<ntuple wrapper="MyWrapper" >
     <activate branch="el_pt" />
     <activate branch="el_eta" />
     <activate branch="el_phi" />
     <activate branch="el_e" />


Jigsaw has too many options!

Jigsaw --help
Or, if you sourced jigsaw-setenv.sh or jigsaw-autocompletion.sh, just press TAB while issuing the command.

I can't remember all those keys for default cuts, event manipulators, histogram fillers and ntuple wrappers. How do I find out?

jigsaw-config [--installdir] [--cuts|-c] [--event-manipulators|-em] [--ntuple-wrappers|-nw] [--histogram-fillers|-hf]

Registered cuts: 25
* btag
* dummy
* emuoverlap
* etmiss
* hfoverlap
* jetcleaning
* jetcut
* larError
* leptoncut
* leptonveto
* mTW
* noncollisionbkg
* optjetcut
* pvx
* recoparticlecut
* recoquantitycut
* sl
* svx
* topchess
* toptriangular
* toptrigger
* toptruth
* trigger
* triggermatch
* zmass

What if I want to run Jigsaw on the grid with Panda/prun?

Try to download our submission package AnalysisOnGrid. Customize user.config according to your local setup. This version is highly developmental and we can give only a very limited support. You asked for that!
svn co svn+ssh://YourCERNUserID@svn.cern.ch/reps/atlasgrp/Institutes/Bologna/AnalysisOnGrid/trunk AnalysisOnGrid

The configuration file looks like this:

# define configurable options

framework     = Jigsaw
grid_user     = RiccardoDiSipio
sw_dir        = /home/disipio/development
framework_src = /home/disipio/development/Jigsaw
install_dir   = /home/disipio/local
grl           = share/Top_GRL_K.xml
params        = cardfiles/MCChallenge_2011_EPS.xml
wrapper       = Rachik
klfitter      = /home/disipio/KLFitter
compile       = local  # remote | local | afs

You are very likely to modfiy options grid_user, sw_dir, framework_src, install_dir and klfitter (perhaps wrapper, too). We suggest to leave the compile option untouched if you are running on a Linux machine.

To run:

./submit_analysis -a AnalysisName -r RunFile.list 

During the submission, a file called logs/submission_YYYYMMDDHHMM.log is created. It can be used for bookkeeping.

-- RiccardoDiSipio - 22-Sep-2011

  • output_historgams.png:
Edit | Attach | Watch | Print version | History: r14 < r13 < r12 < r11 < r10 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r14 - 2011-12-09 - RiccardoDiSipio
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback