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
+ 19
− 485
Compare changes
  • Side-by-side
  • Inline
Files
2
km3io/offline.py
+ 6
− 340
@@ -103,12 +103,14 @@ class OfflineReader:
@cached_property
def events(self):
"""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(
@@ -118,351 +120,13 @@ 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"""
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.
# 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
@@ -473,6 +137,8 @@ class Usr:
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:
self._usr_idx_lookup = {