boostvector.py

 
 
'''
Validation of the Boost Vector calibration
'''
 
 
from prompt import ValidationSettings
import ROOT
import sys
import json
 
import numpy as np
import scipy.stats
import matplotlib.pyplot as plt
 
import os
from glob import glob
from math import sqrt, frexp, hypot, atan2
 
from datetime import datetime, timedelta
from ROOT.Belle2 import Unit
 
 
settings = ValidationSettings(name='BoostVector Calibrations',
                              description=__doc__,
                              download_files=['stdout'],
                              expert_config={})
 
 
# Get the eigen parameters of the matrix
def getBVvalsErrs(bv, bve):
 
    bv = np.array(bv)
    bve = np.array(bve)
 
    mag = np.linalg.norm(bv)
    xzAngle = atan2(bv[0], bv[2])
    yzAngle = atan2(bv[1], bv[2])
 
    magUnc = np.linalg.norm(bv * bve) / mag
    xzAngleUnc = 1. / (1 + (bv[0] / bv[2])**2) * hypot(bve[0] / bv[2], -bve[1] * bv[0] / bv[2]**2)
    yzAngleUnc = 1. / (1 + (bv[1] / bv[2])**2) * hypot(bve[1] / bv[2], -bve[1] * bv[1] / bv[2]**2)
 
    tomrad = Unit.rad / Unit.mrad
    return [1e3 * mag, tomrad * xzAngle, tomrad * yzAngle], [1e3 * magUnc, tomrad * xzAngleUnc, tomrad * yzAngleUnc]
 
 
# Load the values of the Boost Vector parameters into the list
def getBVvalues(path):
    runDict = {}
    with open(path + '/database.txt') as fDB:
        for ll in fDB:
            ll = ll.strip()
            ll = ll.split(' ')
 
            # deriving file name
            fN = ll[0].replace('/', '_') + '_rev_' + ll[1] + '.root'
 
            r = ll[2].split(',')
            runDict[fN] = (int(r[0]), int(r[1]))
 
    arr = []
 
    fList = glob(path + '/*.root')
    for fName in fList:
        bName = os.path.basename(fName)
        runExp = runDict[bName]
 
        f = ROOT.TFile.Open(fName)
        bvAll = f.Get("CollisionBoostVector")
        assert(bvAll.ClassName() == "Belle2::EventDependency")
 
        evNums = bvAll.getEventNumbers()
 
        for i in range(len(evNums) + 1):
            bvObj = bvAll.getObjectByIndex(i)
 
            bvV = bvObj.getBoost()
            bv = [bvV(i) for i in range(3)]
            bveV = bvObj.getBoostCovariance()
            bve = [sqrt(bveV(i, i)) for i in range(3)]
 
            tStart = int(frexp(bveV(0, 1))[0] * 2**53) % 2**32 / 3600.
            tEnd = int(frexp(bveV(0, 2))[0] * 2**53) % 2**32 / 3600.
 
            bvV, bvE = getBVvalsErrs(bv, bve)
 
            arr.append((runExp, tStart, tEnd, bvV, bvE))
        f.Close()
 
    arr = sorted(arr, key=lambda x: x[1])
 
    return arr
 
# Print the Boost Vector parameters to the text file
 
 
def printToFile(arr):
 
    with open('bvData.csv', 'w') as outFile:
        outFile.write('exp  run  tStart  tEnd  mag  xzAngle  yzAngle    magUnc  xzAngleUnc  yzAngleUnc \n')
 
        for e in arr:
            outFile.write(
                # exp  run
                str(e[0][0]) + ' ' +
                str(e[0][1]) + ' ' +
                # tStart  tEnd
                str(e[1]) + ' ' +
                str(e[2]) + ' ' +
                # mag  xzAngle  yzAngle
                str(e[3][0]) + ' ' +
                str(e[3][1]) + ' ' +
                str(e[3][2]) + ' ' +
                # magUnc  xzAngleUnc  yzAngleUnc
                str(e[4][0]) + ' ' +
                str(e[4][1]) + ' ' +
                str(e[4][2]) + '\n')
        return arr
 
 
# Create a plot with the specified variable
def plotVar(arr, limits, vName, getterV, getterE=None):
 
    tVals = []
    Vals = []
    Errs = []
 
    tGapVals = []
    GapVals = []
 
    for i, el in enumerate(arr):
        s, e = el[1], el[2]
        s = datetime.utcfromtimestamp((s + 9) * 3600)  # Convert to the JST (+9 hours)
        e = datetime.utcfromtimestamp((e + 9) * 3600)
        tVals.append(s)
        tVals.append(e)
        Vals.append(getterV(el))
        Vals.append(getterV(el))
 
        if getterE is not None:
            Errs.append(getterE(el))
            Errs.append(getterE(el))
 
        # Add breaks for longer gap if not the last interval
        if i >= len(arr) - 1:
            continue
 
        dt = (arr[i + 1][1] - arr[i][2]) * 3600
 
        # only consider gaps longer than 10 mins
        if dt < 10 * 60:
            continue
 
        # start-time of gap, end-time of gap
        gS = datetime.utcfromtimestamp((arr[i][2] + 9) * 3600)  # Convert to the JST (+9 hours)
        gE = datetime.utcfromtimestamp((arr[i + 1][1] + 9) * 3600)
 
        tVals.append(gS + (gE - gS) / 2)
        Vals.append(np.nan)
        if getterE is not None:
            Errs.append(np.nan)
 
        # store curve connecting gaps
        tGapVals.append(gS - timedelta(seconds=1))
        tGapVals.append(gS)
        tGapVals.append(gE)
        tGapVals.append(gE + timedelta(seconds=1))
 
        GapVals.append(np.nan)
        GapVals.append(getterV(arr[i]))
        GapVals.append(getterV(arr[i + 1]))
        GapVals.append(np.nan)
 
    plt.plot(tVals, Vals, linewidth=2, color='C0')
    plt.plot(tGapVals, GapVals, linewidth=2, color='C0', alpha=0.35)
 
    Vals = np.array(Vals)
    Errs = np.array(Errs)
 
    if getterE is not None:
        plt.fill_between(tVals, Vals - Errs, Vals + Errs, alpha=0.2)
 
    plt.xlabel('time')
    if 'Angle' in vName:
        plt.ylabel(vName + ' [mrad]')
    else:
        plt.ylabel(vName + r' $[10^{-3}]$')
 
    loc = 'plots/allData'
    if limits is not None:
        plt.xlim(datetime.strptime(limits[0], '%Y-%m-%d'), datetime.strptime(limits[1], '%Y-%m-%d'))
        loc = 'plots/' + limits[0] + 'to' + limits[1]
 
    # if not os.path.isdir(loc):
    os.makedirs(loc, exist_ok=True)
 
    plt.savefig(loc + '/' + vName + '.png')
    plt.clf()
 
 
def plotPullSpectrum(arr, vName, getterV, getterE):
 
    from itertools import groupby
    vals = np.array([k for k, g in groupby([getterV(v) for v in arr])])
    errs = np.array([k for k, g in groupby([getterE(v) for v in arr])])
    assert(len(vals) == len(errs))
 
    diffs = (vals[1:] - vals[:-1]) / np.hypot(errs[1:], errs[:-1])
 
    # Get 1sigma quantiles of the normal distribution
    n1, n2 = scipy.stats.norm.cdf([-1, 1])
 
    q1 = np.quantile(diffs, n1)
    q2 = np.quantile(diffs, n2)
 
    plt.hist(diffs, bins=np.linspace(-10, 10, 20), label=f'q1={round(q1,1)}, q2={round(q2,1)}')
    plt.xlabel(vName)
    plt.legend()
 
    loc = 'plots/allData'
    plt.savefig(loc + '/' + vName + '.png')
    plt.clf()
 
 
def run_validation(job_path, input_data_path, requested_iov, expert_config):
    '''
    Run the validation.
    '''
 
    # Expert config can contain the time ranges of the plots
    if expert_config != '':
        expert_config = json.loads(expert_config)
 
    allLimits = [None]
    if expert_config is not None and 'plotsRanges' in expert_config:
        allLimits += expert_config['plotsRanges']
 
    # Path to the database.txt file and to the payloads.
    dbFile = glob(f'{job_path}/**/database.txt', recursive=True)
    dbFile = [db for db in dbFile if 'algorithm_output' not in db]
    assert(len(dbFile) == 1)
    dbFile = dbFile[0]
    inputDir = dbFile[:dbFile.rfind('/')]
 
    arr = getBVvalues(inputDir)
 
    # print the results to the CSV file
    printToFile(arr)
 
    plt.figure(figsize=(18, 9))
 
    # plot the results
    for limits in allLimits:
        plotVar(arr, limits, 'boostMag', lambda e: e[3][0], lambda e: e[4][0])
        plotVar(arr, limits, 'xzBoostAngle', lambda e: e[3][1], lambda e: e[4][1])
        plotVar(arr, limits, 'yzBoostAngle', lambda e: e[3][2], lambda e: e[4][2])
 
    plotPullSpectrum(arr, 'boostMagPulls', lambda e: e[3][0], lambda e: e[4][0])
    plotPullSpectrum(arr, 'xzBoostAnglePulls', lambda e: e[3][1], lambda e: e[4][1])
    plotPullSpectrum(arr, 'yzBoostAnglePulls', lambda e: e[3][2], lambda e: e[4][2])
 
 
if __name__ == "__main__":
    run_validation(*sys.argv[1:])