SAF files Treatment

A small library to process SAF files output by MadAnalysis 5. It provides a representation of histograms that can then be manipulated in order to produce customized plots. It relies on the Python library xml.etree.

Install

The package can be install through pip command:

git clone https://github.com/marcorossi5/saft
cd saft
pip install -e .

It requires numpy python package.

Features

File importation

The main class is the SAF_File class, which represents the file as a whole.

Basic usage: mySAF = SAF_File(<filename>) to import a SAF file.

Note: The file is modified to have a <SAF>...</SAF> as a root if needed.

Histograms

SAF_File objects have an attribute called histos which is a list of SAF_Histo objects representing the histograms defined in the saft module:

from saft.saft import SAF_File
saf_file = SAF_File("MadAnalysis5job.saf")

The available histograms can be inspected with the get_title_list and get_from_title methods:

available = saf_file.get_title_list() # get available titles
hist = saf_file.get_from_title(available[0])

hist is a SAF_Histo instance, whose most important attributes are:

• data is a numpy array that contains the entries of the histogram
• bins is a list that contains the histogram bin edges
• bin_size is the size of the binsize, nbins is the number of bins
• xmin and xmax are the bounds of the histogram's x-axis
• overflow and underflow are the sums of weights of events right and left of the histogram bounds.
• title holds the name of the histogram.

The following methods are implemented

---

Important note

Forked from here and turned into a python package.

At the moment SAFT only works out of the box with SAF files for MA5 analyses that do not contain cuts due to a parsing issue with "<" symbols inside tag texts. One needs to find them and remove them by hand in the cut descriptions in order to be able to handle cuts.

---

Todo list

• Add a Cut class
• Handle errors with "<" symbols in cut descriptions which cause parsing failures
• Better handling of the data (weights, normalization, cross sections etc.)
• Add error handling
• Better importation procedures: find a way to get the information on the content of the histograms