def aggregate(dataset, timeline, aggregator=None, **aggregator_param):
# nothing to aggregate if dataset and/or timeline are/is empty
if dataset.isEmpty() or timeline.isEmpty():
return PYODataset.Empty()
# params will be passed to the aggregator function
# it is meant to contain the user-provided parameters
# and additional parameters (such as the current timerange)
params = {}
for p, param in enumerate(aggregator_param):
params[param] = aggregator_param[param]
nTimeranges = timeline.getNumberOfTimeranges()
datas = [None] * nTimeranges
number = np.zeros((nTimeranges, ), dtype=np.int32)
for t, timerange in enumerate(timeline):
subdataset = sub(dataset, timerange)
# try and aggregate subdataset:
# some aggregator may fail in case of empty dataset
# some other may return something
try:
if aggregator == None:
datas[t] = np.copy(subdataset.data)
else:
params['timerange'] = timerange
datas[t] = aggregator(subdataset, **params)
number[t] = datas[t].shape[0]
except Exception, e:
raise e
datas[t] = None
number[t] = 0
def sub(dataset, extent):
timeline = pinocchIO.utils.timeline.sub(dataset.getTimeline(), extent)
if timeline.isEmpty():
return PYODataset.Empty()
trIDs = dataset.getTimeline().indexOfTimerangesInPeriod(extent, strict=False)
# note: trIDs cannot be empty -- otherwise timeline would have been empty
# and we would have already returned an empty dataset
number = dataset.getNumber()[trIDs]
data = np.concatenate([dataset[i] for i in trIDs])
return PYODataset.PYODataset(data, number, timeline)
