Running DaVinci

From DaVinci/v62r0 the DaVinci configuration has been modernized and revisited in order to improve the user accessibility and hide all technicalities the user doesn’t need to deal with. The two major changes with respect to the old configuration are:

The configuration of your job is now declared in two files:

  • A Python function that takes an options returns the PyConf configuration

  • A YAML file which declares data specific configuration and which is used to populate the options object

DaVinci can then be run using:

lb-run DaVinci/vXrY lbexec my_module:my_function options.yaml

Replacing lb-run DaVinci/vXrY with a specific version, or in the case of development builds replacing it with with the ./run script. Alternatively, it can be run after loading the DaVinci environment in a bash shell by using:

lb-run DaVinci/vXrY bash
lb-exec my_module:my_function options.yaml

Minimal example

Make a file named my_module.py that contains a function that takes an options argument and returns the result of DaVinci.make_config`:

from DaVinci import Options, make_config
from DaVinci.algorithms import create_lines_filter
from PyConf.Algorithms import PrintDecayTree
from PyConf.reading import get_particles


def print_decay_tree(options: Options):
    turbo_line = "Hlt2BsToJpsiPhi_JPsi2MuMu_PhiToKK_Line"
    input_data = get_particles(f"/Event/HLT2/{turbo_line}/Particles")

    user_algorithms = [
        create_lines_filter("HDRFilter_SeeNoEvil", lines =[ f"{turbo_line}"]),
        PrintDecayTree(name="PrintBsToJpsiPhi", Input=input_data)
    ]

    return make_config(options, user_algorithms)

Also make a file named options.yaml containing:

input_files:
- root://eoslhcb.cern.ch//eos/lhcb/wg/dpa/wp3/tests/hlt2_passthrough_thor_lines.dst
input_manifest_file: root://eoslhcb.cern.ch//eos/lhcb/wg/dpa/wp3/tests/hlt2_passthrough_thor_lines.tck.json
input_type: ROOT
evt_max: 100
ntuple_file: davinci_ntuple.root
input_process: TurboPass
print_freq: 1
simulation: true
conddb_tag: sim-20180530-vc-md100
dddb_tag: dddb-20180815

This example can then be run using:

lb-run DaVinci/vXrY lbexec my_module:print_decay_tree options.yaml

For a more detailed explanation of this job, as well as many more examples, see the DaVinci tutorials repository.

Options YAML

The full schema with which the options.yaml file is parsed can be found in Options.

Running in Ganga

Ganga does not support lbexec, yet! A workaround has been prepared, as follows:

  1. Add this to the end of your DaVinci options file e.g. my_module.py

import sys
if sys.argv[0].endswith("gaudirun.py"):
    import yaml
    from DaVinci import Options
    with open("options.yaml", "r") as f:
        options = Options(**yaml.safe_load(f))
    with options.apply_binds():  # required as gaudirun.py does not apply binds!
        print_decay_tree(options)  # make sure to rename print_decay_tree to whatever your function is called

The function apply_binds(), a member function of the LbExec Options class [https://gitlab.cern.ch/lhcb/LHCb/-/blob/master/GaudiConf/python/GaudiConf/LbExec/options.py](LbExec/options.py), is automatically called when running DaVinci through lbexec. When running through gaudirun.py the binds must be applied within the DaVinci options, otherwise DaVinci will encounter reconstruction issues.

  1. Add my_module.py to your GaudiExec application options list, and options.yaml to your job.inputfiles (input sandbox.)

j = Job()
j.backend = Dirac()  # as required
j.application = GaudiExec(
    dir="DaVinci_v62r0",
    platform="x86_64_v3-centos7-gcc11-opt+g"
)
j.application.options = ["my_module.py"]
j.inputdata = ...  # as required
j.inputfiles = ["hlt2.tck.json", "options.yaml"]  # change as required
j.outputfiles = [DiracFile("*.root"), LocalFile("stdout")]  # change as required

# ... configure as required ...

j.submit()

  1. Submit your jobs as usual. Note that in options.yaml, input_files should be set to a dummy value, as it is overwritten later.