python.histsampling Namespace Reference

Classes
class	TH1
class	TH2

Functions
	load_hist (*args)
	get_sampling_vars (h)
	get_random_bin (globalbins, cheights)
	get_random_x (h, globalbins, cheights, globalbin_to_axisbin)
	get_random_xy (h2, globalbins, cheights, globalbin_to_axisbin)

Variables
str	__author__ = "Andy Buckley <andy.buckley@cern.ch>"

Detailed Description

Tools for histogram sampling, in particular inverse transform sampling which is
missing from ROOT's TH2 classes.

Function Documentation

get_random_bin()

get_random_bin	(		globalbins,
			cheights )

Choose a random bin from the cumulative distribution list of nbins+1 entries.

TODO: Search more efficiently (lin and log guesses, then lin search or
binary split depending on vector size).

Definition at line 59 of file histsampling.py.

def get_random_bin(globalbins, cheights):
    """
    Choose a random bin from the cumulative distribution list of nbins+1 entries.
 
    TODO: Search more efficiently (lin and log guesses, then lin search or
    binary split depending on vector size).
    """
    assert len(cheights) == len(globalbins)+1
    randomheight = random.uniform(0, cheights[-1])
    for i, iglobal in enumerate(globalbins):
        if randomheight >= cheights[i] and randomheight < cheights[i+1]:
            return iglobal
    raise Exception("Sample fell outside range of cumulative distribution?!?!")
 
 

get_random_x()

get_random_x	(		h,
			globalbins,
			cheights,
			globalbin_to_axisbin )

Choose a random bin via get_random_bin, then pick a uniform random x
point in that bin (without any attempt at estimating the in-bin distribution).

Definition at line 74 of file histsampling.py.

def get_random_x(h, globalbins, cheights, globalbin_to_axisbin):
    """
    Choose a random bin via get_random_bin, then pick a uniform random x
    point in that bin (without any attempt at estimating the in-bin distribution).
    """
    irand = get_random_bin(globalbins, cheights)
    axisids = globalbin_to_axisbin.get(irand)
    assert axisids is not None
    xrand = random.uniform(h.GetXaxis().GetBinLowEdge(axisids[0]), h.GetXaxis().GetBinUpEdge(axisids[0]))
    return xrand
 
 

get_random_xy()

get_random_xy	(		h2,
			globalbins,
			cheights,
			globalbin_to_axisbin )

Choose a random bin via get_random_bin, then pick a uniform random x,y
point in that bin (without any attempt at estimating the in-bin distribution).

Definition at line 86 of file histsampling.py.

def get_random_xy(h2, globalbins, cheights, globalbin_to_axisbin):
    """
    Choose a random bin via get_random_bin, then pick a uniform random x,y
    point in that bin (without any attempt at estimating the in-bin distribution).
    """
    irand = get_random_bin(globalbins, cheights)
    axisids = globalbin_to_axisbin.get(irand)
    assert axisids is not None
    xrand = random.uniform(h2.GetXaxis().GetBinLowEdge(axisids[0]), h2.GetXaxis().GetBinUpEdge(axisids[0]))
    yrand = random.uniform(h2.GetYaxis().GetBinLowEdge(axisids[1]), h2.GetYaxis().GetBinUpEdge(axisids[1]))
    return xrand, yrand
 
 

get_sampling_vars()

get_sampling_vars

(

h

)

Get the following from a histogram h, since the ROOT API sucks:
* list of global bin IDs (not even contiguous for 2D, gee thanks ROOT)
* dict mapping global bin IDs to a tuple of axis bin IDs
* list of nbins+1 cumulative bin values, in the same order as globalbins

Definition at line 33 of file histsampling.py.

def get_sampling_vars(h):
    """
    Get the following from a histogram h, since the ROOT API sucks:
    * list of global bin IDs (not even contiguous for 2D, gee thanks ROOT)
    * dict mapping global bin IDs to a tuple of axis bin IDs
    * list of nbins+1 cumulative bin values, in the same order as globalbins
    """
    globalbin_to_axisbin = {} # for reverse axis bin lookup to get edges
    globalbins = [] # because they aren't easily predicted, nor contiguous
    cheights = [0] # cumulative "histogram" from which to uniformly sample
    if issubclass(type(h), ROOT.TH1):
        for ix in range(1, h.GetNbinsX()+1):
            iglobal = h.GetBin(ix)
            globalbins.append(iglobal)
            globalbin_to_axisbin[iglobal] = (ix,)
            cheights.append(cheights[-1] + h.GetBinContent(iglobal))
    elif issubclass(type(h), ROOT.TH2):
        for ix in range(1, h.GetNbinsX()+1):
            for iy in range(1, h.GetNbinsY()+1):
                iglobal = h.GetBin(ix, iy)
                globalbins.append(iglobal)
                globalbin_to_axisbin[iglobal] = (ix, iy)
                cheights.append(cheights[-1] + h.GetBinContent(iglobal))
    return globalbins, globalbin_to_axisbin, cheights
 
 

load_hist()

load_hist

(

*

args

)

Load a histogram from a filename/TFile and histo name. If a single arg is
provided, it has to be a histo object and will be cloned before return.

Definition at line 13 of file histsampling.py.

def load_hist(*args):
    """
    Load a histogram from a filename/TFile and histo name. If a single arg is
    provided, it has to be a histo object and will be cloned before return.
    """
    h = None
    if len(args) == 1 and issubclass(type(args[0]), ROOT.TH1):
        h = args[0].Clone()
    elif len(args) == 2:
        if isinstance(args[0], str) and isinstance(args[1], str) :
            f = ROOT.TFile.Open(args[0])
            h = copy.deepcopy(f.Get(args[1]).Clone())
            #f.Close()
        elif type(args[0]) is ROOT.TFile and type(args[1]) is str:
            h = args[0].Get(args[1]).Clone()
    if h is None:
        raise Exception("Error in histogram loading from " + args)
    return h
 
 

Variable Documentation

__author__

str python.histsampling.__author__ = "Andy Buckley <andy.buckley@cern.ch>"

private

Definition at line 8 of file histsampling.py.