RankChecker Class Reference

Inheritance diagram for RankChecker:

Public Member Functions
	initialize (self)
	event (self)

Public Attributes
	plist = Belle2.PyStoreObj("e-")
	particle list object

Detailed Description

Check if the ranks are actually what we want

Definition at line 44 of file best_candidate_selection.py.

Member Function Documentation

event()

event

(

self

)

And check all the ranks

Definition at line 52 of file best_candidate_selection.py.

    def event(self):
        """And check all the ranks"""
        # make a list of all the values and a dict of all the extra infos
        px = []
        py = []
        einfo = defaultdict(list)
        for particle in self.plist:
            px.append(particle.getPx())
            py.append(particle.getPy())
            # get all names of existing extra infos but convert to a native list of python strings to avoid
            # possible crashes if the std::vector returned by the particle goes out of scope
            names = [str(n) for n in particle.getExtraInfoNames()]
            for n in names:
                einfo[n].append(particle.getExtraInfo(n))
 
        # check the default name is set correctly if we don't specify an output variable
        print(list(einfo.keys()))
        assert 'M_rank' in einfo.keys(), "Default name is not as expected"
 
        # Now determine the correct ranks if multiple values are allowed:
        # create a dictionary which will be value -> rank for all unique values
        # in theory we just need loop over the sorted(set(values)) but we have
        # special treatment for nans which should go always to the end of the
        # list so sort with a special key that replaces nan by inf or -inf
        # depending on sort order
        px_value_ranks = {v: i for i, v in enumerate(sorted(set(px), reverse=True,
                                                            key=lambda v: -math.inf if math.isnan(v) else v), 1)}
        py_value_ranks = {v: i for i, v in enumerate(sorted(set(py),
                                                            key=lambda v: math.inf if math.isnan(v) else v), 1)}
 
        # Ok, test if the rank from extra info actually corresponds to what we
        # want
        for v, r in zip(px, einfo["px_high_multi"]):
            print(f"Value: {v}, rank: {r}, should be: {px_value_ranks[v]}")
            assert r == px_value_ranks[v], "Rank is not correct"
 
        for v, r in zip(py, einfo["py_low_multi"]):
            print(f"Value: {v}, rank: {r}, should be: {py_value_ranks[v]}")
            assert r == py_value_ranks[v], "Rank is not correct"
 
        # so we checked multiRank=True. But for multiRank=False this is more
        # complicated because ranking a second time will destroy the order
        # of the previous sorts. But we can at least check if all the ranks
        # form a range from 1..n if we sort them
        simple_range = list(range(len(px)))
        px_single_ranks = list(sorted(int(r) - 1 for r in einfo["px_high_single"]))
        assert simple_range == px_single_ranks, "sorted ranks don't form a range from 1..n"
        # but the second two rankings are on the same variable in the same
        # order so they need to keep the order stable. so for py_low_single the
        # ranks need to be the range without sorting
        py_single_ranks = list(int(r) - 1 for r in einfo["py_low_single"])
        assert simple_range == py_single_ranks, "ranks don't form a range from 1..n"
 
 

initialize()

initialize

(

self

)

Create particle list object

Definition at line 47 of file best_candidate_selection.py.

    def initialize(self):
        """Create particle list object"""
        
        self.plist = Belle2.PyStoreObj("e-")
 

Member Data Documentation

plist

plist = Belle2.PyStoreObj("e-")

particle list object

Definition at line 50 of file best_candidate_selection.py.

---

The documentation for this class was generated from the following file:

• analysis/tests/best_candidate_selection.py