Skip to content
Snippets Groups Projects

WIP: Slicing and refactoring offline

Closed WIP: Slicing and refactoring offline
37-user-parameters-seem-to-be-transposed into master
Closed Tamas Gal requested to merge 37-user-parameters-seem-to-be-transposed into master
Compare
and
Show latest version
2 files
+ 137
− 683
Compare changes
  • Side-by-side
  • Inline
Files
2
km3io/offline.py
+ 118
− 524
from collections import namedtuple
import uproot
import numpy as np
import awkward1 as ak
import warnings
from .definitions import mc_header
MAIN_TREE_NAME = "E"
# 110 MB based on the size of the largest basket found so far in km3net
BASKET_CACHE_SIZE = 110 * 1024**2
BASKET_CACHE = uproot.cache.ThreadSafeArrayCache(BASKET_CACHE_SIZE)
EXCLUDE_KEYS = set(["AAObject", "t", "fBits", "fUniqueID"])
BranchMapper = namedtuple(
"BranchMapper",
['name', 'key', 'extra', 'exclude', 'update', 'attrparser', 'flat'])
['name', 'key', 'extra', 'exclude', 'update', 'attrparser'])
def _nested_mapper(key):
@@ -19,40 +20,60 @@ def _nested_mapper(key):
return '_'.join(key.split('.')[1:])
EXCLUDE_KEYS = set(["AAObject", "t", "fBits", "fUniqueID"])
EVENTS_MAP = BranchMapper("events", "Evt", {
't_sec': 't.fSec',
't_ns': 't.fNanoSec'
}, [], {
'n_hits': 'hits',
'n_mc_hits': 'mc_hits',
'n_tracks': 'trks',
'n_mc_tracks': 'mc_trks'
}, lambda a: a, True)
EVENTS_MAP = BranchMapper(name="events",
key="Evt",
extra={
't_sec': 't.fSec',
't_ns': 't.fNanoSec'
},
exclude=[],
update={
'n_hits': 'hits',
'n_mc_hits': 'mc_hits',
'n_tracks': 'trks',
'n_mc_tracks': 'mc_trks'
},
attrparser=lambda a: a)
SUBBRANCH_MAPS = [
BranchMapper("tracks", "trks", {}, ['trks.usr_data', 'trks.usr'], {},
_nested_mapper, False),
BranchMapper("mc_tracks", "mc_trks", {},
['mc_trks.usr_data', 'mc_trks.usr'], {}, _nested_mapper,
False),
BranchMapper("hits", "hits", {}, ['hits.usr'], {}, _nested_mapper, False),
BranchMapper("mc_hits", "mc_hits", {},
['mc_hits.usr', 'mc_hits.dom_id', 'mc_hits.channel_id'], {},
_nested_mapper, False),
BranchMapper("events", "Evt", {
't_sec': 't.fSec',
't_ns': 't.fNanoSec'
}, [], {
'n_hits': 'hits',
'n_mc_hits': 'mc_hits',
'n_tracks': 'trks',
'n_mc_tracks': 'mc_trks'
}, lambda a: a, True),
BranchMapper(
name="tracks",
key="trks",
extra={},
exclude=['trks.usr_data', 'trks.usr', 'trks.fUniqueID', 'trks.fBits'],
update={},
attrparser=_nested_mapper),
BranchMapper(name="mc_tracks",
key="mc_trks",
extra={},
exclude=[
'mc_trks.usr_data', 'mc_trks.usr', 'mc_trks.rec_stages',
'mc_trks.fitinf', 'mc_trks.fUniqueID', 'mc_trks.fBits'
],
update={},
attrparser=_nested_mapper),
BranchMapper(name="hits",
key="hits",
extra={},
exclude=[
'hits.usr', 'hits.pmt_id', 'hits.origin', 'hits.a',
'hits.pure_a', 'hits.fUniqueID', 'hits.fBits'
],
update={},
attrparser=_nested_mapper),
BranchMapper(name="mc_hits",
key="mc_hits",
extra={},
exclude=[
'mc_hits.usr', 'mc_hits.dom_id', 'mc_hits.channel_id',
'mc_hits.tdc', 'mc_hits.tot', 'mc_hits.trig',
'mc_hits.fUniqueID', 'mc_hits.fBits'
],
update={},
attrparser=_nested_mapper),
]
class cached_property:
"""A simple cache decorator for properties."""
def __init__(self, function):
@@ -67,7 +88,7 @@ class cached_property:
class OfflineReader:
"""reader for offline ROOT files"""
def __init__(self, file_path=None, fobj=None, data=None, index=None):
def __init__(self, file_path=None):
""" OfflineReader class is an offline ROOT file wrapper
Parameters
@@ -77,52 +98,19 @@ class OfflineReader:
path-like object that points to the file.
"""
self._index = index
if file_path is not None:
self._fobj = uproot.open(file_path)
self._tree = self._fobj[MAIN_TREE_NAME]
self._data = self._tree.lazyarrays(
basketcache=uproot.cache.ThreadSafeArrayCache(
BASKET_CACHE_SIZE))
else:
self._fobj = fobj
self._tree = self._fobj[MAIN_TREE_NAME]
self._data = data
self._fobj = uproot.open(file_path)
self._tree = self._fobj[MAIN_TREE_NAME]
@cached_property
def events(self):
return Branch(self._tree, mapper=EVENTS_MAP, index=self._index, subbranchmaps=SUBBRANCH_MAPS)
@classmethod
def from_index(cls, source, index):
"""Create an instance with a subtree of a given index
Parameters
----------
source: ROOTDirectory
The source file object.
index: index or slice
The index or slice to create the subtree.
"""
instance = cls(fobj=source._fobj,
data=source._data[index],
index=index)
return instance
def __getitem__(self, index):
return OfflineReader.from_index(source=self, index=index)
def __len__(self):
tree = self._fobj[MAIN_TREE_NAME]
if self._index is None:
return len(tree)
else:
return len(
tree.lazyarrays(basketcache=uproot.cache.ThreadSafeArrayCache(
BASKET_CACHE_SIZE))[self.index])
"""The `E` branch, containing all offline events."""
return Branch(self._tree,
mapper=EVENTS_MAP,
subbranchmaps=SUBBRANCH_MAPS)
@cached_property
def header(self):
"""The file header"""
if 'Head' in self._fobj:
header = {}
for n, x in self._fobj['Head']._map_3c_string_2c_string_3e_.items(
@@ -132,378 +120,36 @@ class OfflineReader:
else:
warnings.warn("Your file header has an unsupported format")
def get_best_reco(self):
"""returns the best reconstructed track fit data. The best fit is defined
as the track fit with the maximum reconstruction stages. When "nan" is
returned, it means that the reconstruction parameter of interest is not
found. for example, in the case of muon simulations: if [1, 2] are the
reconstruction stages, then only the fit parameters corresponding to the
stages [1, 2] are found in the Offline files, the remaining fit parameters
corresponding to the stages 3, 4, 5 are all filled with nan.
Returns
-------
numpy recarray
a recarray of the best track fit data (reconstruction data).
"""
keys = ", ".join(self.keys.fit_keys[:-1])
empty_fit_info = np.array(
[match for match in self._find_empty(self.tracks.fitinf)])
fit_info = [
i for i, j in zip(self.tracks.fitinf, empty_fit_info[:, 1])
if j is not None
]
stages = self._get_max_reco_stages(self.tracks.rec_stages)
fit_data = np.array([i[j] for i, j in zip(fit_info, stages[:, 2])])
rows_size = len(max(fit_data, key=len))
equal_size_data = np.vstack([
np.hstack([i, np.zeros(rows_size - len(i)) + np.nan])
for i in fit_data
])
return np.core.records.fromarrays(equal_size_data.transpose(),
names=keys)
def _get_max_reco_stages(self, reco_stages):
"""find the longest reconstructed track based on the maximum size of
reconstructed stages.
Parameters
----------
reco_stages : chunked array
chunked array of all the reconstruction stages of all tracks.
In km3io, it is accessed with
km3io.OfflineReader(my_file).tracks.rec_stages .
Returns
-------
numpy array
array with 3 columns: *list of the maximum reco_stages
*lentgh of the maximum reco_stages
*position of the maximum reco_stages
"""
empty_reco_stages = np.array(
[match for match in self._find_empty(reco_stages)])
max_reco_stages = np.array(
[[max(i, key=len),
len(max(i, key=len)),
i.index(max(i, key=len))]
for i, j in zip(reco_stages, empty_reco_stages[:, 1])
if j is not None])
return max_reco_stages
def get_reco_fit(self, stages, mc=False):
"""construct a numpy recarray of the fit information (reconstruction
data) of the tracks reconstructed following the reconstruction stages
of interest.
Parameters
----------
stages : list
list of reconstruction stages of interest. for example
[1, 2, 3, 4, 5].
mc : bool, optional
default is False to look for fit data in the tracks tree in offline files
(not the mc tracks tree). mc=True to look for fit data from the mc tracks
tree in offline files.
Returns
-------
numpy recarray
a recarray of the fit information (reconstruction data) of
the tracks of interest.
Raises
------
ValueError
ValueError raised when the reconstruction stages of interest
are not found in the file.
"""
keys = ", ".join(self.keys.fit_keys[:-1])
if mc is False:
rec_stages = np.array(
[match for match in self._find_rec_stages(stages, mc=False)])
fitinf = self.tracks.fitinf
if mc is True:
rec_stages = np.array(
[match for match in self._find_rec_stages(stages, mc=True)])
fitinf = self.mc_tracks.fitinf
mask = rec_stages[:, 1] != None
if np.all(rec_stages[:, 1] == None):
raise ValueError(
"The stages {} are not found in your file.".format(
str(stages)))
else:
fit_data = np.array(
[i[k] for i, k in zip(fitinf[mask], rec_stages[:, 1][mask])])
rec_array = np.core.records.fromarrays(fit_data.transpose(),
names=keys)
return rec_array
def get_reco_hits(self, stages, keys, mc=False):
"""construct a dictionary of hits class data based on the reconstruction
stages of interest. For example, if the reconstruction stages of interest
are [1, 2, 3, 4, 5], then get_reco_hits method will select the hits data
from the events that were reconstructed following these stages (i.e
[1, 2, 3, 4, 5]).
Parameters
----------
stages : list
list of reconstruction stages of interest. for example
[1, 2, 3, 4, 5].
keys : list of str
list of the hits class attributes.
mc : bool, optional
default is False to look for hits data in the hits tree in offline files
(not the mc_hits tree). mc=True to look for mc hits data in the mc hits
tree in offline files.
Returns
-------
dict
dictionary of lazyarrays containing data for each hits attribute requested.
Raises
------
ValueError
ValueError raised when the reconstruction stages of interest
are not found in the file.
"""
lazy_d = {}
if mc is False:
rec_stages = np.array(
[match for match in self._find_rec_stages(stages, mc=False)])
hits_data = self.hits
if mc is True:
rec_stages = np.array(
[match for match in self._find_rec_stages(stages, mc=True)])
hits_data = self.mc_hits
mask = rec_stages[:, 1] != None
if np.all(rec_stages[:, 1] == None):
raise ValueError(
"The stages {} are not found in your file.".format(
str(stages)))
else:
for key in keys:
lazy_d[key] = getattr(hits_data, key)[mask]
return lazy_d
def get_reco_events(self, stages, keys, mc=False):
"""construct a dictionary of events class data based on the reconstruction
stages of interest. For example, if the reconstruction stages of interest
are [1, 2, 3, 4, 5], then get_reco_events method will select the events data
that were reconstructed following these stages (i.e [1, 2, 3, 4, 5]).
Parameters
----------
stages : list
list of reconstruction stages of interest. for example
[1, 2, 3, 4, 5].
keys : list of str
list of the events class attributes.
mc : bool, optional
default is False to look for the reconstruction stages in the tracks tree
in offline files (not the mc tracks tree). mc=True to look for the reconstruction
data in the mc tracks tree in offline files.
Returns
-------
dict
dictionary of lazyarrays containing data for each events attribute requested.
Raises
------
ValueError
ValueError raised when the reconstruction stages of interest
are not found in the file.
"""
lazy_d = {}
if mc is False:
rec_stages = np.array(
[match for match in self._find_rec_stages(stages, mc=False)])
if mc is True:
rec_stages = np.array(
[match for match in self._find_rec_stages(stages, mc=True)])
mask = rec_stages[:, 1] != None
if np.all(rec_stages[:, 1] == None):
raise ValueError(
"The stages {} are not found in your file.".format(
str(stages)))
else:
for key in keys:
lazy_d[key] = getattr(self.events, key)[mask]
return lazy_d
def get_reco_tracks(self, stages, keys, mc=False):
"""construct a dictionary of tracks class data based on the reconstruction
stages of interest. For example, if the reconstruction stages of interest
are [1, 2, 3, 4, 5], then get_reco_tracks method will select tracks data
from the events that were reconstructed following these stages (i.e
[1, 2, 3, 4, 5]).
Parameters
----------
stages : list
list of reconstruction stages of interest. for example
[1, 2, 3, 4, 5].
keys : list of str
list of the tracks class attributes.
mc : bool, optional
default is False to look for tracks data in the tracks tree in offline files
(not the mc tracks tree). mc=True to look for tracks data in the mc tracks
tree in offline files.
Returns
-------
dict
dictionary of lazyarrays containing data for each tracks attribute requested.
Raises
------
ValueError
ValueError raised when the reconstruction stages of interest
are not found in the file.
"""
lazy_d = {}
if mc is False:
rec_stages = np.array(
[match for match in self._find_rec_stages(stages, mc=False)])
tracks_data = self.tracks
if mc is True:
rec_stages = np.array(
[match for match in self._find_rec_stages(stages, mc=True)])
tracks_data = self.mc_tracks
mask = rec_stages[:, 1] != None
if np.all(rec_stages[:, 1] == None):
raise ValueError(
"The stages {} are not found in your file.".format(
str(stages)))
else:
for key in keys:
lazy_d[key] = np.array([
i[k] for i, k in zip(
getattr(tracks_data, key)[mask], rec_stages[:,
1][mask])
])
return lazy_d
def _find_rec_stages(self, stages, mc=False):
"""find the index of reconstruction stages of interest in a
list of multiple reconstruction stages.
Parameters
----------
stages : list
list of reconstruction stages of interest. for example
[1, 2, 3, 4, 5].
mc : bool, optional
default is False to look for reconstruction stages in the tracks tree in
offline files (not the mc tracks tree). mc=True to look for reconstruction
stages in the mc tracks tree in offline files.
Yields
------
generator
the track id and the index of the reconstruction stages of
interest if found. If the reconstruction stages of interest
are not found, None is returned as the stages index.
"""
if mc is False:
stages_data = self.events.tracks.rec_stages
if mc is True:
stages_data = self.events.mc_tracks.rec_stages
for trk_index, rec_stages in enumerate(stages_data):
try:
stages_index = rec_stages.index(stages)
except ValueError:
stages_index = None
yield trk_index, stages_index
continue
yield trk_index, stages_index
def _find_empty(self, array):
"""finds empty lists/arrays in an awkward array
Parameters
----------
array : awkward array
Awkward array of data of interest. For example:
km3io.OfflineReader(my_file).tracks.fitinf .
Yields
------
generator
the empty list id and the index of the empty list. When
data structure (list) is simply empty, None is written in the
corresponding index. However, when data structure (list) is not
empty and does not contain an empty list, then False is written in the
corresponding index.
"""
for i, rs in enumerate(array):
try:
if len(rs) == 0:
j = None
if len(rs) != 0:
j = rs.index([])
except ValueError:
j = False # rs not empty but [] not found
yield i, j
continue
yield i, j
class Usr:
"""Helper class to access AAObject usr stuff"""
"""Helper class to access AAObject `usr`` stuff"""
def __init__(self, name, tree, index=None):
# Here, we assume that every event has the same names in the same order
# to massively increase the performance. This needs triple check if it's
# always the case; the usr-format is simply a very bad design.
# print("initialising usr for {}".format(name))
# print("Setting up usr")
# to massively increase the performance. This needs triple check if
# it's always the case; the usr-format is simply a very bad design.
self._name = name
try:
tree['usr'] # This will raise a KeyError in old aanet files
# which has a different strucuter and key (usr_data)
# We do not support those...
self._usr_names = [
n.decode("utf-8") for n in tree['usr_names'].lazyarray()[0]
n.decode("utf-8") for n in tree['usr_names'].lazyarray(
basketcache=BASKET_CACHE)[0]
]
except (KeyError, IndexError): # e.g. old aanet files
print("The `usr` fields could not be parsed for the '{}' branch."
.format(name))
self._usr_names = []
else:
# print(" checking usr data")
self._usr_idx_lookup = {
name: index
for index, name in enumerate(self._usr_names)
}
data = tree['usr'].lazyarray(
basketcache=uproot.cache.ThreadSafeArrayCache(
BASKET_CACHE_SIZE))
data = tree['usr'].lazyarray(basketcache=BASKET_CACHE)
if index is not None:
data = data[index]
self._usr_data = data
# print(" adding attributes")
for name in self._usr_names:
# print(" setting {}".format(name))
setattr(self, name, self[name])
def __getitem__(self, item):
@@ -538,7 +184,7 @@ def _to_num(value):
class Header:
"""The online header"""
"""The header"""
def __init__(self, header):
self._data = {}
for attribute, fields in mc_header.items():
@@ -564,7 +210,12 @@ class Header:
class Branch:
"""Branch accessor class"""
def __init__(self, tree, mapper, index=None, subbranches=None, subbranchmaps=None):
def __init__(self,
tree,
mapper,
index=None,
subbranchmaps=None,
keymap=None):
self._tree = tree
self._mapper = mapper
self._index = index
@@ -572,19 +223,23 @@ class Branch:
self._branch = tree[mapper.key]
self._subbranches = []
self._initialise_keys() #
if keymap is None:
self._initialise_keys() #
else:
self._keymap = keymap
if subbranches is not None:
self._subbranches = subbranches
if subbranchmaps is not None:
for mapper in subbranchmaps:
subbranch = Branch(self._tree, mapper=mapper, index=self._index)
subbranch = Branch(self._tree,
mapper=mapper,
index=self._index)
self._subbranches.append(subbranch)
for subbranch in self._subbranches:
setattr(self, subbranch._mapper.name, subbranch)
def _initialise_keys(self):
"""Create the keymap and instance attributes for branch keys"""
# TODO: this could be a cached property
keys = set(k.decode('utf-8') for k in self._branch.keys()) - set(
self._mapper.exclude) - EXCLUDE_KEYS
self._keymap = {
@@ -597,8 +252,7 @@ class Branch:
del self._keymap[k]
for key in self._keymap.keys():
# print("setting", self._mapper.name, key)
setattr(self, key, self[key])
setattr(self, key, None)
def keys(self):
return self._keymap.keys()
@@ -607,111 +261,51 @@ class Branch:
def usr(self):
return Usr(self._mapper.name, self._branch, index=self._index)
def __getitem__(self, item):
"""Slicing magic a la numpy"""
if isinstance(item, slice):
return self.__class__(self._tree, self._mapper, index=item, subbranches=self._subbranches)
if isinstance(item, int):
# A bit ugly, but whatever works
if self._mapper.flat:
if self._index is None:
dct = {
key: self._branch[self._keymap[key]].array()
for key in self.keys()
}
else:
dct = {
key:
self._branch[self._keymap[key]].array()[self._index]
for key in self.keys()
}
for subbranch in self._subbranches:
dct[subbranch._mapper.name] = subbranch
return BranchElement(self._mapper.name, dct)[item]
else:
if self._index is None:
dct = {
key: self._branch[self._keymap[key]].array()[item]
for key in self.keys()
}
else:
dct = {
key:
self._branch[self._keymap[key]].array()[self._index,
item]
for key in self.keys()
}
for subbranch in self._subbranches:
dct[subbranch._mapper.name] = subbranch
return BranchElement(self._mapper.name, dct)
if isinstance(item, tuple):
return self[item[0]][item[1]]
if isinstance(item, str):
item = self._keymap[item]
def __getattribute__(self, attr):
if attr.startswith("_"): # let all private and magic methods pass
return object.__getattribute__(self, attr)
out = self._branch[item].lazyarray(
basketcache=uproot.cache.ThreadSafeArrayCache(
BASKET_CACHE_SIZE))
if attr in self._keymap.keys(): # intercept branch key lookups
out = self._branch[self._keymap[attr]].lazyarray(
basketcache=BASKET_CACHE)
if self._index is not None:
out = out[self._index]
return out
return self.__class__(self._tree, self._mapper, index=np.array(item), subbranches=self._subbranches)
return object.__getattribute__(self, attr)
def __getitem__(self, item):
"""Slicing magic"""
if isinstance(item, (int, slice)):
return self.__class__(self._tree,
self._mapper,
index=item,
keymap=self._keymap,
subbranchmaps=SUBBRANCH_MAPS)
if isinstance(item, tuple):
return self[item[0]][item[1]]
return self.__class__(self._tree,
self._mapper,
index=np.array(item),
keymap=self._keymap,
subbranchmaps=SUBBRANCH_MAPS)
def __len__(self):
if self._index is None:
return len(self._branch)
elif isinstance(self._index, int):
return 1
else:
return len(
self._branch[self._keymap['id']].lazyarray()[self._index])
return len(self._branch[self._keymap['id']].lazyarray(
basketcache=BASKET_CACHE)[self._index])
def __str__(self):
return "Number of elements: {}".format(len(self._branch))
def __repr__(self):
return "<{}[{}]: {} parsed elements>".format(self.__class__.__name__,
self._mapper.name,
len(self))
class BranchElement:
"""Represents a single branch element
Parameters
----------
name: str
The name of the branch
dct: dict (keys=attributes, values=arrays of values)
The data
index: slice
The slice mask to be applied to the sub-arrays
"""
def __init__(self, name, dct, index=None, subbranches=[]):
self._dct = dct
self._name = name
self._index = index
self.ItemConstructor = namedtuple(self._name[:-1], dct.keys())
if index is None:
for key, values in dct.items():
setattr(self, key, values)
else:
for key, values in dct.items():
setattr(self, key, values[index])
def __getitem__(self, item):
if isinstance(item, slice):
return self.__class__(self._name, self._dct, index=item)
if isinstance(item, int):
if self._index is None:
return self.ItemConstructor(
**{k: v[item]
for k, v in self._dct.items()})
else:
return self.ItemConstructor(
**{k: v[self._index][item]
for k, v in self._dct.items()})
def __repr__(self):
return "<{}[{}]>".format(self.__class__.__name__, self._name)
length = len(self)
return "<{}[{}]: {} element{}>".format(self.__class__.__name__,
self._mapper.name, length,
's' if length > 1 else '')