Tools for fitting femtoscopic correlation functions

The art and science of femtoscopy relies on the accurate fitting of correlation functions to data gathered by various collider experiments. This package hopes to provide both a standard and easy-to-use base on which fits may be performed.

There are two components of this library:

libFemtoFitter.so

A C library which interfaces with ROOT for reading and writing .root files and performing fits with TMinuit objects, the standard for fitting femtoscopic analysis data.

femtofitter

A python package which wraps libFemtoFitter with an easy to use interface

This library can be built from source or installed via the python package manager pip which bundles both the libFemtoFitter.so library and femtofitter python package.

This library does not have the goal of backwards compatibility with older systems, as fitting is often done on users' computers where installation of up-to-date packages is expected.

Building requires ROOT6 with the following features:

• python
• cxx11 or cxx14
• threads
• minuit2

You can check your root installation by running the following command; if it returns 'yes' for each parameter it is correctly configured (or switch cpp11)

$ root-config --has-python --has-threads --has-minuit2 --has-cxx14

Other Requirements:

cmake

Like ROOT itself, femtofitter is configured and built with the program cmake_. Minimum version is 3.6

python

This should be installed by your operating system; python3.7 is recommended

pip

Probably already installed, but use sudo apt install python3-pip (or equivalent for your package manager) if python3 -m pip returns a "module not found" error

misc

There are python <-> ROOT support libraries that will be installed automatically.

# pull from repository
$ git clone https://gitlab.cern.ch/akubera/FemtoFitter
$ cd FemtoFitter

# standard build - installs to $HOME/.local
$ make

# explicit in-source cmake build
$ cmake -DCMAKE_INSTALL_PREFIX=/custom/install/path . && make

Rather than installing from source, pre-built binaries are provided using the python-package-index (https://pypi.python.org/pypi/femtofitter) and easily downloadable using the pip installer (installed via your operating system's installer such as sudo apt install python3-pip

Note: If you are using alternative python installers, such as conda, please use the standard instructions rather than installing potentially conflicting files with pip.

# install from stored binaries (may not work on incompatible machines)
$ pip install --user femtofitter

# automatically build from source
$ pip install --user --no-binary=all femtofitter

While libFemtoFitter.so can be used directly by users' macros and other libraries/executables, the expected usage is via the femtofitter executable provided with this package. This program is designed to simply be handed the names of input files, the classes which shall be used to fit, and output destinations and it performs all indicated operations.

While flexibility is built into the design, there are limits to the structure of datafiles; this will be elaborated on more in the documentation.

The standard usage is to create a configuration file which has three responsibilities:

• lists the input files and how to group histogram objects in 'analysis-units'
• indicates which fitters to be used and set all options for this fitter
• specify the output root-file and how to organize 'output-units'

---
- input:
    file: data.root
    matching:
      num: foo/bar
  output:
    file: fit-results-%(timestamp).root

  fits:
  # Standard gaussian-fit, no coulomb
  - name: Gauss3D_NoCoulomb
    class: Gauss3D
    coulomb: False

  # Standard gaussian-fit including CoulombFactor provided by
  # the CoulombInterp3D class created with file CoulombInterpData.root
  - name: Gauss3D
    class: Gauss3D
    coulomb:
      class: CoulombInterp3D
      file: CoulombInterpData.root

Issues, suggestions, and patches are welcome. Please make sure that submitted code follows the style guidelines enforced by clang-format and editorconfig files in this directory, and that all unittests pass when merging into the master branch.

This code is released under the conditions of the LGPL-2.1 free-software license, the contents of which are provided in the LICENSE file of this repository.

Copyright is held by Andrew Kubera (mailto:[email protected]).