What is this tool?

Let's all be frank, root sucks and the root file format is horrible. It's among humanities worst pieces of software. With this small tool I hope to fix the damage that was done, at least a little, by converting root files into native Python formats.

It's using Numpy and a library called Uproot to read and process these damn root files. So far it is specialist for one task and I will have to work on it to make it actually viable for more use cases. That task is to extract PXD data from Belle 2 data files.

This tool is still in early development, which means that the source code is horrible and that not all features work properly or aren't even fully implemented. Right now only PXD is supported by this tool. In the future I plan to include more detectors.

How to use this?

This is a single class, that needs to be instantiated, it doesn't take any arguments. Just import it like this:

from rootable import Rootable

Then you can create an instance:

loadFromRoot = Rootable()

and load the root file and all the data:

loadFromRoot.loadData('/root-files/slow_pions_2.root')
loadFromRoot.getClusters()
loadFromRoot.getCoordinates()
loadFromRoot.getLayers()
loadFromRoot.getMatrices()
loadFromRoot.getMCData()

One can now specify that ROI unselected digits should be read and to reconstruct the cluster data from them. this is still iffy, after including ROI unselected clusters, one cannot load monte carlo information and the u/v mapping is still very wonky.

loadFromRoot.loadData('/root-files/slow_pions_2.root', includeUnSelected=True)

The 'get' commands don't have any return value, but instead work in-place. Then all data is stored inside the object as dict:

loadFromRoot.data

Here follows a list of keywords contained in the dict:

• cluster data:
  • 'eventNumber': int
  • 'clsCharge': int
  • 'seedCharge': int
  • 'clsSize': int
  • 'uSize': int
  • 'vSize': int
  • 'uPosition': float
  • 'vPosition': float
  • 'sensorID': int
  • 'detector': str
  • 'roiSelected': bool
• coordinates:
  • 'xPosition': float
  • 'yPosition': float
  • 'zPosition': float
• layers:
  • 'layer': int
  • 'ladder': int
• matrices:
  • 'cluster': array
• Monte Carlo data:
  • 'momentumX': float
  • 'momentumY': float
  • 'momentumZ': float
  • 'pdg': int
  • 'clsNumber': int

Since the class is subscriptable one can access every element directly using the keywords like this:

loadFromRoot['eventNumber']

or

loadFromRoot[0]

will return either the array containing the event numbers of the first entry of every array contained in the classes dict.

It is possible to filter through the data:

loadFromRoot.where('clsSize == 1')
loadFromRoot.where('clsSize > 1')
loadFromRoot.where('clsSize > 1', 'layer == 1')

or even:

loadFromRoot.where('eventNumbers in [0,1,2]')

And finally you can convert the dict into a structured Numpy array by simply writing:

loadFromRoot.getStructuredArray()

This last command returns a Numpy array. From there the user can save it using Numpys build-in functions, convert it to Pandas or use it in any way that is compatible with Numpy.

The class itself is iterable, it's a bit different from typical python dicts, I iterate over rows and return it as a dict, not sure if that's actually useful.

In certain instances it can be very usefull to stack certain columns together, for example when one wants to calculate the distance from the origin. Then one can stack the positions:

loadFromRoot.stack('xPosition', 'yPosition', 'zPosition', toKey: 'position')

Installation

You will need to the wheel and setuptools packages of python in order to install Download the repo, navigate in the terminal to the folder and run the following script:

python3 setup.py sdist

and then:

pip3 install .