def generate_traces(self, device, instance_id, date):
'''
Returns trace:
series: indexed by time with column name 'time',
series is series of average power value
'''
filename = self.path + device + '/dev_' + instance_id + '_' + date + '.csv'
df = pd.read_csv(filename,sep=';',\
header=None,names=['time','1s_W','8s_W'])
df['time'] = pd.to_datetime(df['time'], format='%d/%m/%Y %H:%M:%S')
df.set_index('time', inplace=True)
try:
series = df['1s_W'].resample(self.sample_rate, how='mean')
series = series.map(decimal.Decimal) / decimal.Decimal(1000.0)
series.name = device
except ValueError:
raise SampleError(self.sample_rate)
series_mult = self.split_on_NANs(series)
return [
app.ApplianceTrace(
single_series, {
'source': self.source,
'device_name': device,
'instance_name': instance_id,
'date': date,
'trace_num': i
}) for i, single_series in enumerate(series_mult)
]
def get_trace_in_time_of_day(device_trace,start_time,end_time):
'''
Given a trace and a start and end datetime.time, it returns a trace
within that time period.
'''
print "WARNING: deprecated, "\
"use trace.get_time_of_day(start_time,end_time) instead"
new_series=device_trace.series.ix[start_time:end_time]
return appliance.ApplianceTrace(new_series,device_trace.metadata)
def aggregate_traces(traces, metadata, how="strict"):
'''
Given a list of temporally aligned traces, aggregate them into a single
signal.
'''
if how == "strict":
if not traces_aligned(traces):
raise appliance.AlignmentError
summed_series = traces[0].series
for trace in traces[1:]:
summed_series += trace.series
return appliance.ApplianceTrace(summed_series, metadata)
else:
return NotImplementedError
def resample_trace(trace,sample_rate):
'''
Takes a trace and resamples it to a given sample rate, defined by the
offset aliases described in panda time series.
http://pandas.pydata.org/pandas-docs/stable/timeseries.html#offset-aliases
'''
print "WARNING: deprecated, "\
"use appliance_trace.resample(sample_rate) instead"
try:
new_series=trace.series.astype(float)
new_series=new_series.resample(sample_rate,how='mean')
new_series=new_series.map(decimal.Decimal)
new_series.name=trace.series.name
return appliance.ApplianceTrace(new_series,trace.metadata)
except ValueError:
raise SampleError(self.sample_rate)
def concatenate_traces(traces, metadata=None, how="strict"):
'''
Given a list of appliance traces, returns a single concatenated
trace. With how="strict" option, must be sampled at the same rate and
consecutive, without overlapping datapoints.
'''
if not metadata:
metadata = traces[0].metadata
if how == "strict":
# require ordered list of consecutive, similarly sampled traces with no
# missing data.
return appliance.ApplianceTrace(pd.concat([t.series for t in traces]),
metadata)
else:
raise NotImplementedError
def split_trace_into_rate(trace,rate):
'''
Given a single trace, a list of traces are returned that are each
from a unique date.
'''
print "WARNING: deprecated, "\
"use appliance trace.split_into_rate(rate) instead"
# set rate to group by
if rate == 'D' or rate == '1D':
groupby_rate = trace.series.index.date
elif rate == 'W' or rate == '1W':
groupby_rate = trace.series.index.week
else:
raise NotImplementedError('Looking for "week" or "day"')
traces=[]
for i, group in enumerate(trace.series.groupby(groupby_rate)):
metadata = dict.copy(trace.metadata)
metadata['trace_num'] = i
traces.append(appliance.ApplianceTrace(group[1],metadata))
return traces