| 1 |
bbetchar |
1.1 |
SimpleTree Documentation
|
| 2 |
|
|
-------------------------
|
| 3 |
|
|
|
| 4 |
|
|
* Justification and Philosophy
|
| 5 |
|
|
|
| 6 |
|
|
Before SimpleTree, there are two formats that CMSSW can produce which
|
| 7 |
|
|
one might use for analysis: the EDM tree and the Histogram. Neither
|
| 8 |
|
|
of these formats is appropriate for the kind of interactive
|
| 9 |
|
|
exploration of a dataset I find most useful in preliminary analysis.
|
| 10 |
|
|
|
| 11 |
|
|
Histograms are static displays of information, good for communicating
|
| 12 |
|
|
information or answering single well posed questions. However,
|
| 13 |
|
|
interactive exploration via histograms produced directly from CMSSW
|
| 14 |
|
|
requires rerunning CMSSW for each new question and subsequent cut, or
|
| 15 |
|
|
alternatively, an immense number of histograms and anticipation of the
|
| 16 |
|
|
correct questions and cuts.
|
| 17 |
|
|
|
| 18 |
|
|
The EDM tree is browseable as a rootuple, and so can be explored
|
| 19 |
|
|
dynamically. However, interacting dynamically with an EDM tree is
|
| 20 |
|
|
impractical, due to the prohibitively large file size, the long/deep
|
| 21 |
|
|
variable names, and the prohibitively slow processing of
|
| 22 |
|
|
PoolOutputModule. The EDM tree is an excellent format for consistent,
|
| 23 |
|
|
traceable, reproducible, massive parallel processing in CMSSW, but the
|
| 24 |
|
|
requirements of these features make interactive browsing impractical.
|
| 25 |
|
|
|
| 26 |
|
|
SimpleTree is meant to be a minimal, flexible format which allows
|
| 27 |
|
|
dynamic exploration. It is not meant for the production of private
|
| 28 |
|
|
"standard" rootuples but rather as an "n-dimensional" Histogram. It
|
| 29 |
|
|
is meant to reduce the frequency of running CMSSW and the grid. It is
|
| 30 |
|
|
NOT meant to facilitate abandonment of CMSSW. SimpleTrees are meant
|
| 31 |
|
|
to produce plots with simple cuts. If you find yourself constructing
|
| 32 |
|
|
complicated variables from the leaves of a SimpleTree, or writing a
|
| 33 |
|
|
ROOT macro to loop over the events in a SimpleTree, you are clearly
|
| 34 |
|
|
doing something wrong. Those activities are best handled by the
|
| 35 |
|
|
massive parallelism of CMSSW on the grid.
|
| 36 |
|
|
|
| 37 |
|
|
|
| 38 |
|
|
|
| 39 |
|
|
|
| 40 |
|
|
|
| 41 |
|
|
* Advantages
|
| 42 |
|
|
|
| 43 |
|
|
SimpleTree offers dynamic exploration via browsing and cuts.
|
| 44 |
|
|
|
| 45 |
|
|
SimpleTree is a shallow: no need to hunt for the information.
|
| 46 |
|
|
|
| 47 |
|
|
SimpleTree is much smaller than a comparable EDM tree.
|
| 48 |
|
|
High statistics should fit on your local computer.
|
| 49 |
|
|
|
| 50 |
|
|
SimpleTree separates the construction of variables from their output format.
|
| 51 |
|
|
|
| 52 |
|
|
SimpleTree is easy to use.
|
| 53 |
|
|
|
| 54 |
|
|
|
| 55 |
|
|
|
| 56 |
|
|
|
| 57 |
|
|
|
| 58 |
|
|
* Disadvantages
|
| 59 |
|
|
|
| 60 |
|
|
SimpleTree files are much larger than Histogram files.
|
| 61 |
|
|
|
| 62 |
|
|
SimpleTrees store information inefficiently due to lack of hierarchy.
|
| 63 |
|
|
|
| 64 |
|
|
SimpleTree files retain no provenance information (by design).
|
| 65 |
|
|
|
| 66 |
|
|
|
| 67 |
|
|
|
| 68 |
|
|
|
| 69 |
|
|
|
| 70 |
|
|
* outputCommands
|
| 71 |
|
|
|
| 72 |
|
|
SimpleTree takes one configuration option, the cms.vstring
|
| 73 |
|
|
outputCommands. This is a series of "keep" and "drop" statements
|
| 74 |
|
|
which uses the same software fragment as the PoolOutputModules. You
|
| 75 |
|
|
can find documentation here:
|
| 76 |
|
|
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideSelectingBranchesForOutput
|
| 77 |
|
|
|
| 78 |
|
|
|
| 79 |
|
|
|
| 80 |
|
|
|
| 81 |
|
|
|
| 82 |
|
|
* Chaining the CRAB results
|
| 83 |
|
|
|
| 84 |
|
|
You can easily combine the results of multiple jobs by using a TChain.
|
| 85 |
|
|
It is especially easy to use the wildcard (*) notation, as follows:
|
| 86 |
|
|
|
| 87 |
|
|
$ root -l
|
| 88 |
|
|
root [0] TChain ch("chain_name");
|
| 89 |
|
|
root [1] ch.Add("file_name_*.root/subdir/tree");
|
| 90 |
|
|
root [2] ch.Draw("someVar", "someOtherVar>cut");
|
| 91 |
|
|
|
| 92 |
|
|
|
| 93 |
|
|
|
| 94 |
|
|
|
| 95 |
|
|
|
| 96 |
|
|
* Typical Workflow
|
| 97 |
|
|
|
| 98 |
|
|
1. Write an EDProducer which puts the variables you want into the EDM
|
| 99 |
|
|
tree as C++ standard types or std::vectors of C++ standard types.
|
| 100 |
|
|
2. Write a config file with a Source, a TFileService, your EDProducer, and a SimpleTree.
|
| 101 |
|
|
3. Run CMSSW, locally or on the grid, using the config file of step (2).
|
| 102 |
|
|
4. Interactively browse the resulting rootuple, making cuts and comparisons.
|
| 103 |
|
|
5. Write a ROOT macro to format and output any histograms you want to present.
|
| 104 |
|
|
|
| 105 |
|
|
|
