import numpy as np
from numpy.typing import ArrayLike
import uproot as ur
class Rootable:
"""
this class uses uproot to load pxd data from root files and converts them into
native python data structures.
it can load the cluster information, uses the digits to generate the adc matrices,
coordinates, layer and ladders and finally also monte carlo data.
"""
def __init__(self) -> None:
# these are the sensor IDs of the pxd modules/panels from the root file, they are
# use to identify on which panels a cluster event happened
self.panelIDs = np.array([ 8480, 8512, 8736, 8768, 8992, 9024, 9248, 9280,
9504, 9536, 9760, 9792, 10016, 10048, 10272, 10304,
16672, 16704, 16928, 16960, 17184, 17216, 17440, 17472,
17696, 17728, 17952, 17984, 18208, 18240, 18464, 18496,
18720, 18752, 18976, 19008, 19232, 19264, 19488, 19520])
# every line in this corresponds to one entry in the array above, this is used
# to put the projected uv plane in the right position
self.panelShifts = np.array([[1.3985 , 0.2652658 , 3.68255],
[ 1.3985 , 0.23238491, -0.88255],
[ 0.80146531, 1.17631236, 3.68255],
[ 0.82407264, 1.15370502, -0.88255],
[-0.2582769 , 1.3985 , 3.68255],
[-0.2322286 , 1.3985 , -0.88255],
[-1.17531186, 0.80246583, 3.68255 ],
[-1.15510614, 0.82267151, -0.88255],
[-1.3985 , -0.2645974 , 3.68255],
[-1.3985 , -0.23012119, -0.88255],
[-0.80591227, -1.17186534, 3.68255],
[-0.82344228, -1.15433536, -0.88255],
[ 0.26975836, -1.3985 , 3.68255],
[ 0.23326624, -1.3985 , -0.88255],
[ 1.1746111 , -0.80316652, 3.68255],
[ 1.15205703, -0.82572062, -0.88255],
[ 2.2015 , 0.26959865, 5.01305],
[ 2.2015 , 0.2524582 , -1.21305],
[ 1.77559093, 1.32758398, 5.01305],
[ 1.78212569, 1.31626522, -1.21305],
[ 0.87798948, 2.03516717, 5.01305],
[ 0.88478563, 2.03124357, -1.21305],
[-0.26129975, 2.2015 , 5.01305],
[-0.25184137, 2.2015 , -1.21305],
[-1.32416655, 1.77756402, 5.01305],
[-1.31417539, 1.78333226, -1.21305],
[-2.03421133, 0.87964512, 5.01305],
[-2.02960691, 0.88762038, -1.21305],
[-2.2015 , -0.25954151, 5.01305],
[-2.2015 , -0.24969109, -1.21305],
[-1.77636043, -1.32625112, 5.01305],
[-1.78138268, -1.31755219, -1.21305],
[-0.87493138, -2.03693277, 5.01305 ],
[-0.8912978 , -2.02748378, -1.21305],
[ 0.26489725, -2.2015 , 5.01305],
[ 0.25364439, -2.2015 , -1.21305],
[ 1.3269198 , -1.7759744 , 5.01305],
[ 1.32258793, -1.77847528, -1.21305],
[ 2.03616649, -0.87625871, 5.01305],
[ 2.02936825, -0.8880338 , -1.21305]])
# every entry here corresponds to the entries in the array above, these are
# used for rotating the projected uv plane
self.panelRotations = np.array([ 90, 90, 135, 135, 180, 180, 225, 225, 270, 270, 315, 315, 360,
360, 405, 405, 90, 90, 120, 120, 150, 150, 180, 180, 210, 210,
240, 240, 270, 270, 300, 300, 330, 330, 360, 360, 390, 390, 420,
420])
# the layer and ladder arrays, for finding them from sensor id
self.panelLayer = np.array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2])
self.panelLadder = np.array([1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 12, 12, 13, 13, 14, 14, 15, 15, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21])
# all transpormaations are stored in a dict, with the sensor id as a keyword
self.transformation = {}
self.layersLadders = {}
for i in range(len(self.panelIDs)):
self.transformation[str(self.panelIDs[i])] = [self.panelShifts[i], self.panelRotations[i]]
self.layersLadders[str(self.panelIDs[i])] = [self.panelLayer[i], self.panelLadder[i]]
# these are the branch names for cluster info in the root file
self.gotClusters = False
self.clusters = ['PXDClusters/PXDClusters.m_clsCharge',
'PXDClusters/PXDClusters.m_seedCharge',
'PXDClusters/PXDClusters.m_clsSize',
'PXDClusters/PXDClusters.m_uSize',
'PXDClusters/PXDClusters.m_vSize',
'PXDClusters/PXDClusters.m_uPosition',
'PXDClusters/PXDClusters.m_vPosition',
'PXDClusters/PXDClusters.m_sensorID']
# these are the branch names for cluster digits in the root file
self.digits = ['PXDDigits/PXDDigits.m_uCellID',
'PXDDigits/PXDDigits.m_vCellID',
'PXDDigits/PXDDigits.m_charge']
# this establishes the relationship between clusters and digits
# because for some reaseon the branch for digits has a different
# size than the cluster branch
self.clusterToDigis = 'PXDClustersToPXDDigits/m_elements/m_elements.m_to'
# these are the branch names for monte carlo data in the root file
self.mcData = ['MCParticles/MCParticles.m_pdg',
'MCParticles/MCParticles.m_momentum_x',
'MCParticles/MCParticles.m_momentum_y',
'MCParticles/MCParticles.m_momentum_z']
# these two establish the relation ship to an from clusters and monte carlo
# there more entries than in the cluster data, but there still mc data missing
# for some cluster files
self.clusterToMC = 'PXDClustersToMCParticles/m_elements/m_elements.m_to'
self.mcToCluster = 'PXDClustersToMCParticles/m_elements/m_elements.m_from'
# this dict stores the data
self.data = {}
def __getitem__(self, index: str | int | ArrayLike) -> np.ndarray | dict:
"""
this makes the class subscriptable, one can retrieve one coloumn by using
strings as keywords, or get a row by using integer indices or arrays
"""
if isinstance(index, str):
return self.data[index]
return {key: value[index] for key, value in self.data.items()}
def loadData(self, file: str) -> None:
"""
reads the file off of the harddrive, it automatically creates event numbers
file: str = it's the whole file path + .root ending
"""
self.eventTree = ur.open(f'{file}:tree')
self._getEventNumbers()
def _getEventNumbers(self) -> None:
"""
a private method that gets called on file import
it generates the event numbers from the jagged arrays
coming from the branches
"""
eventNumbers = []
clusters = self.eventTree.arrays('PXDClusters/PXDClusters.m_clsCharge', library='np')['PXDClusters/PXDClusters.m_clsCharge']
for i in range(len(clusters)):
eventNumbers.append(np.array([i]*len(clusters[i])))
self.data['eventNumbers'] = self._flatten(eventNumbers)
def _getData(self, keyword: str, library: str = 'np') -> np.ndarray:
"""
a private method for converting branches into something useful, namely
into numpy arrays, if the keyward library is set to np.
keyword: str = the full branch name
library: str = can be 'np' (numpy), 'pd' (pandas) or 'ak' (akward)
see uproot documentation for more info
"""
try:
data = self.eventTree.arrays(keyword, library=library)[keyword]
return self._flatten(data)
except:
return KeyError
def _flatten(self, structure: ArrayLike, maxDepth: int = None, currentDepth: int = 0) -> np.ndarray:
"""
this is a private function, that gets called during loading branches
it flattens ragged array, one can set the depths to which one wants to flatten
structure: the list/array to flatten
maxDepth: int = the amount of flattening
currentDepth: int = don't touch this, it's used for recursively calling
"""
flat_list = []
for element in structure:
if isinstance(element, (list, np.ndarray)) and (maxDepth is None or currentDepth < maxDepth):
flat_list.extend(self._flatten(element, maxDepth, currentDepth + 1))
else:
flat_list.append(element)
return np.array(flat_list, dtype=object)
def getClusters(self) -> None:
"""
this uses the array from __init__ to load different branches into the data dict
"""
self.gotClusters = True
for branch in self.clusters:
data = self._getData(branch)
keyword = branch.split('_')[-1]
self.data[keyword] = data
def getMatrices(self) -> None:
"""
loads the digit branches into arrays and converts them into adc matrices
"""
uCellIDs = self.eventTree.arrays(self.digits[0], library='np')[self.digits[0]]
vCellIDs = self.eventTree.arrays(self.digits[1], library='np')[self.digits[1]]
cellCharges = self.eventTree.arrays(self.digits[2], library='np')[self.digits[2]]
# this establishes the relation between digits and clusters, it's still
# shocking to me, that this is necessary, why aren't digits stored in the
# same way as clusters, than one wouldn't need to jump through hoops just
# to have the data in a usable und sensible manner
# root is such a retarded file format
clusterDigits = self.eventTree.arrays(self.clusterToDigis, library='np')[self.clusterToDigis]
self.data['cluster'] = self._getMatrices(uCellIDs, vCellIDs, cellCharges, clusterDigits).astype('int')
def _getMatrices(self, uCellIDs: ArrayLike, vCellIDs: ArrayLike, cellCharges: ArrayLike, clusterDigits: ArrayLike, matrixSize: tuple = (9, 9)) -> np.ndarray:
"""
this takes the ragged/jagged digit arrays and converts them into 9x9 matrices
it's a rather slow process because of all the looping
"""
plotRange = np.array(matrixSize) // 2
events = []
for event in range(len(cellCharges)):
adcValues = []
digitsU = np.array(uCellIDs[event])
digitsV = np.array(vCellIDs[event])
digitsCharge = np.array(cellCharges[event])
digitIndices = clusterDigits[event]
for indices in digitIndices:
cacheImg = np.zeros(matrixSize)
maxChargeIndex = digitsCharge[indices].argmax()
uMax, vMax = digitsU[indices[maxChargeIndex]], digitsV[indices[maxChargeIndex]]
uPos, vPos = digitsU[indices] - uMax + plotRange[0], digitsV[indices] - vMax + plotRange[1]
valid_indices = (uPos >= 0) & (uPos < matrixSize[0]) & (vPos >= 0) & (vPos < matrixSize[1])
cacheImg[uPos[valid_indices].astype(int), vPos[valid_indices].astype(int)] = digitsCharge[indices][valid_indices]
adcValues.append(cacheImg)
events.extend(adcValues)
return np.array(events, dtype=object)
def getCoordisnate(self) -> None:
"""
converting the uv coordinates, together with sensor ids, into xyz coordinates
"""
if self.gotClusters is False:
self.getClusters()
xcoords, ycoords, zcoords = self._getCartesian(self.data['uPosition'], self.data['vPosition'], self.data['sensorID'])
self.data['xPosition'] = xcoords
self.data['yPosition'] = ycoords
self.data['zPosition'] = zcoords
def _getCartesian(self, uPositions: ArrayLike, vPositions: ArrayLike, sensorIDs: ArrayLike) -> tuple[np.ndarray]:
"""
a private method for transposing/converting 2d uv coords into 3d xyz coordinates
"""
length = len(sensorIDs)
xArr, yArr, zArr = np.zeros(length), np.zeros(length), np.zeros(length)
# iterting over the cluster arrays
for index, (u, v, sensor_id) in enumerate(zip(uPositions, vPositions, sensorIDs)):
# grabbing the shift vector and rotation angle
shift, angle = self.transformation[str(sensor_id)]
# setting up rotation matrix
theta = np.deg2rad(angle)
rotMatrix = np.array([[np.cos(theta), -np.sin(theta), 0], [np.sin(theta), np.cos(theta), 0], [0, 0, 1]])
# projecting uv coordinates into 3d space
point = np.array([u, 0, v])
# shifting and rotating the projected vector
shifted = rotMatrix.dot(point) + shift
xArr[index], yArr[index], zArr[index] = shifted
return xArr, yArr, zArr
def getLayers(self) -> None:
"""
looks up the corresponding layers and ladders for every cluster
"""
if self.gotClusters is False:
self.getClusters()
layers, ladders = [], []
for id in self.data['sensorID']:
layer, ladder = self.layersLadders[str(id)]
layers.append(layer)
ladders.append(ladder)
self.data['layers'] = np.array(layers)
self.data['ladder'] = np.array(ladders)
def getMCData(self) -> None:
"""
this loads the monte carlo from the root file
"""
# the monte carlo data, they are longer than the cluster data
pdg = self.eventTree.arrays(self.mcData[0], library='np')[self.mcData[0]]
momentumX = self.eventTree.arrays(self.mcData[1], library='np')[self.mcData[1]]
momentumY = self.eventTree.arrays(self.mcData[2], library='np')[self.mcData[2]]
momentumZ = self.eventTree.arrays(self.mcData[3], library='np')[self.mcData[3]]
# this loads the relation ships to and from clusters and mc data
# this is the same level of retardedness as with the cluster digits
clusterToMC = self.eventTree.arrays(self.clusterToMC, library='np')[self.clusterToMC]
mcToCluster = self.eventTree.arrays(self.mcToCluster, library='np')[self.mcToCluster]
# it need the cluster charge as a jagged/ragged array, maybe I could simply
# use the event numbers, but I am too tired to fix this shitty file format
clsCharge = self.eventTree.arrays('PXDClusters/PXDClusters.m_clsCharge', library='np')['PXDClusters/PXDClusters.m_clsCharge']
# reorganizing MC data
momentumXList = []
momentumYList = []
momentumZList = []
pdgList = []
clusterNumbersList = []
for i in range(len(clusterToMC)):
# _fillMCList fills in the missing spots, because there are not mc data for
# every cluster, even though there are more entries in this branch than
# in the cluster branch... as I said, the root format is retarded
fullClusterReferences = self._fillMCList(mcToCluster[i], clusterToMC[i], len(clsCharge[i]))
clusterNumbersList.append(fullClusterReferences)
pdgs, xmom, ymom, zmom = self._getMCData(fullClusterReferences, pdg[i], momentumX[i], momentumY[i], momentumZ[i])
momentumXList.append(xmom)
momentumYList.append(ymom)
momentumZList.append(zmom)
pdgList.append(pdgs)
self.data['momentumX'] = self._flatten(momentumXList)
self.data['momentumY'] = self._flatten(momentumYList)
self.data['momentumZ'] = self._flatten(momentumZList)
self.data['pdg'] = self._flatten(pdgList)
self.data['clsNumber'] = self._flatten(clusterNumbersList)
def _findMissing(self, lst: list, length: int) -> list:
"""
a private method for finding missing elements in mc data arrays
"""
return sorted(set(range(0, length)) - set(lst))
def _fillMCList(self, fromClusters: ArrayLike, toClusters: ArrayLike, length: ArrayLike) -> list:
"""
a private method for filling MC data arrays where clusters don't have
any information
"""
missingIndex = self._findMissing(fromClusters, length)
testList = [-1] * length
fillIndex = 0
for i in range(len(testList)):
if i in missingIndex:
testList[i] = -1
else:
try:
testList[i] = int(toClusters[fillIndex])
except TypeError:
testList[i] = int(toClusters[fillIndex][0])
fillIndex += 1
return testList
def _getMCData(self, toClusters: ArrayLike, pdgs: ArrayLike, xMom: ArrayLike, yMom: ArrayLike, zMom: ArrayLike) -> tuple[np.ndarray]:
"""
after filling and reorganizing MC data arrays one can finally collect the
actual MC data, where there's data missing I will with zeros
"""
pxList, pyList, pzList = [], [], []
pdgList = []
for references in toClusters:
if references == -1:
pxList.append(0)
pyList.append(0)
pzList.append(0)
pdgList.append(0)
else:
pxList.append(xMom[references])
pyList.append(yMom[references])
pzList.append(zMom[references])
pdgList.append(pdgs[references])
return np.array(pdgList,dtype=list), np.array(pxList,dtype=list), np.array(pyList,dtype=list), np.array(pzList,dtype=list)
def getStructuredArray(self) -> np.ndarray:
"""
this converts the data dict of this class into a structured numpy array
"""
# Create a list to hold the dtype specifications
dtype = []
# Iterate through the dictionary keys and values
for key, value in self.data.items():
# Determine the data type of the first value in the list
sampleValue = value[0]
if isinstance(sampleValue, np.ndarray):
# If the value is an array, use its shape and dtype
fieldDtype = (sampleValue.dtype, sampleValue.shape)
else:
# Otherwise, use the type of the value itself
fieldDtype = type(sampleValue)
# Append the key and data type to the dtype list
dtype.append((key, fieldDtype))
# Convert the dictionary to a list of tuples
keys = list(self.data.keys())
dataList = [tuple(self.data[key][i] for key in keys) for i in range(len(self.data[keys[0]]))]
# Create the structured array
structuredArray = np.array(dataList, dtype=dtype)
return structuredArray