Skip to content

Latest commit

 

History

History
74 lines (52 loc) · 3.8 KB

README.md

File metadata and controls

74 lines (52 loc) · 3.8 KB
Raw

NUISANCE python interface

Jupyter Notebook Tutorial

You can follow the notebook tutorial that covers the main features of the python API over the following notebooks:

  1. Opening Event Files
  2. HepMC3 Binding Interoperability
  3. Basic Event Loops
  4. Advanced Event Loops
  5. HistFrames

There are also a few short notebooks for miscellania:

  1. Getting NuHepMC Flux Information
  2. Simple Oscillation Weighting
  3. Altering Logging Levels
  4. Serializing HistFrames to/from YAML

API Walking Tour

EventInput

The pyNUISANCE.EventSource interface wraps the NUISANCE event reading functionality.

We generally don't recommend writing event loops directly in python because python looping is very slow, however, it can be useful for prototyping plots or selections. For a faster loop, see Frame.

A simple, unweighted event loop might look like below:

from pyNUISANCE import EventSource

evs = EventSource("path/to/my/inputfile")

if not evs:
  print("Failed to open input file")
  exit(1)

for ev, _ in evs:
  # Do something with each event, ev, which are HepMC3::GenEvents

pyNUISANCE.EventSource is an interface to nuis::EventSourceFactory, which is able to use a number of plugins to try and read an input file and present it as a HepMC3 event vector adhereing to the NuHepMC specification.

Iterating on an event source returns a tuple of an event and its Central Value (CV) weight (as defined in NuHepMC G.R.7). In the above example, the CV weight is ignored in the loop variables. All correctly normalised cross-section predictions must weight each event's properties by the CV weight. After looping through all of the events, the best-estimate of the flux-averaged total cross section (FATX) and the sum of CV weights can be queries. N.B. the internal accumulation of these properties are reset for each loop for consistency, so they should only be used after looping.

for ev, cvw in evs:
  # Do something with each event, ev, and the associated CV weight, cvw

flux_averaged_total_xs_estimate = evs.fatx()
sum_cvweights = evs.sumw()

You do not need to loop over all events in a file and breaking early will leave the FATX accumulator in a good state, which lets you prototype accurately with a smaller number of events, and scale up to more events without any special handling of the cross-section scaling. You can print out the current FATX estimate and the sumw as you go:

for i, (ev, cvw) in enumerate(evs):
  if i and (((i+1)% 50000) == 0):
    print("On event %s, CV weight = %s, FATX best estimate = %s pb, sum CV weights = %s " \
      % (i, cvw, evs.fatx(), evs.sumw()))

The above might give output like:

On event 49999, CV weight = 1.0, FATX best estimate = 0.884602612 pb, sum CV weights = 50000.0 
On event 99999, CV weight = 1.0, FATX best estimate = 0.877038888 pb, sum CV weights = 100000.0 
On event 149999, CV weight = 1.0, FATX best estimate = 0.875219239 pb, sum CV weights = 150000.0 
On event 199999, CV weight = 1.0, FATX best estimate = 0.875539752 pb, sum CV weights = 200000.0 
On event 249999, CV weight = 1.0, FATX best estimate = 0.875984252 pb, sum CV weights = 250000.0 
On event 299999, CV weight = 1.0, FATX best estimate = 0.876886361 pb, sum CV weights = 300000.0 
On event 349999, CV weight = 1.0, FATX best estimate = 0.876789265 pb, sum CV weights = 350000.0