def serie_operatorset(manager, period, perfdatas, timewindow, usenan=True):
"""
Generate set of operators.
:param manager: Serie manager
:type manager: canopsis.serie.manager.Serie
:param period: Period used for timeserie
:type period: canopsis.timeserie.timewindow.Period
:param perfdatas: Perfdata classified by metric id
:type perfdatas: dict
:param timewindow: Time window used for consolidation
:type timewindow: canopsis.timeserie.timewindow.TimeWindow
:param usenan: include nan point values
:type usenan: bool
:returns: operators classified by name as dict
"""
operators = {
key: new_operator(key, manager, period, perfdatas, timewindow, usenan)
for key in get_aggregations()
}
return operators
def _aggregation_value(self, values_to_aggregate, func=None):
"""Get the aggregated value related to input values_to_aggregate, a
specific function and a timestamp.
"""
if func is None:
func = get_aggregations()[self.aggregation]
if len(values_to_aggregate) > 0:
result = round(func(values_to_aggregate), 2)
else:
result = 0 if self.fill else None
return result
def calculate(self, points, timewindow, meta=None, usenan=True):
"""Do an operation on all points with input timewindow.
:param bool usenan: if False (True by default) remove nan point values.
:return: points such as follow:
Let func self aggregation function and
input points of the form: [(T0, V0), ..., (Tn, Vn)]
then the result is [(T0, func(V0, V1)), (T2, func(V2, V3), ...].
"""
result = []
nan = float('nan')
# start to exclude points not in timewindow
# in taking care about round time
if self.round_time:
period = self._get_period(timewindow)
round_starttimestamp = period.round_timestamp(
timestamp=timewindow.start()
)
timewindow = timewindow.reduce(
start=round_starttimestamp,
stop=timewindow.stop()
)
# start to exclude points which are not in timewindow
points = [
point for point in points
if point[0] in timewindow and (usenan or not isnan(point[1]))
]
if not meta:
meta = {}
transform_method = meta.get('value', {}).get('type', None)
points = apply_transform(points, method=transform_method)
points_len = len(points)
func = None
# if no period and max_points > len(points)
if (
(not points) or self.period is None
and self.max_points > points_len
):
result = points # result is points
else: # else calculate points with the right aggregation function
func = get_aggregations()[self.aggregation]
# get timesteps
timesteps = self.timesteps(timewindow)[:-1]
# initialize variables for loop
i = 0
values_to_aggregate = []
last_point = None
len_timesteps = len(timesteps)
# iterate on timesteps to get points in [prev_ts, timestamp[
for index, timestamp in enumerate(timesteps):
# initialize values_to_aggregate
values_to_aggregate = []
# set timestamp and previous_timestamp
if index < (len_timesteps - 1): # calculate the upper bound
next_timestamp = timesteps[index + 1]
else:
next_timestamp = None
# fill the values_to_aggregate array
for i in range(i, points_len): # while points to process
pt_ts, pt_val = points[i]
# leave the loop if _timestamp is for a future aggregation
if next_timestamp is not None and pt_ts >= next_timestamp:
break
else:
# add value to list of values to aggregate
values_to_aggregate.append(pt_val)
else: # leave the loop whatever timesteps
i += 1
# TODO: understand what it means :D
if self.aggregation == "DELTA" and last_point:
values_to_aggregate.insert(0, last_point)
if values_to_aggregate:
# get the aggregated value related to values_to_aggregate
aggregation_value = self._aggregation_value(
values_to_aggregate, func
)
# new point to add to result
if usenan or not isnan(aggregation_value):
aggregation_point = timestamp, aggregation_value
result.append(aggregation_point)
# save last_point for future DELTA checking
last_point = aggregation_point[-1]
elif usenan:
result.append((timestamp, nan))
return result
def calculate(self, points, timewindow, meta=None):
"""
Do an operation on all points with input timewindow.
Return points su as follow:
Let fn self aggregation function and
input points of the form: [(T0, V0), ..., (Tn, Vn)]
then the result is [(T0, fn(V0, V1)), (T2, fn(V2, V3), ...].
"""
result = []
# start to exclude points not in timewindow
# in taking care about round time
if self.round_time:
period = self._get_period(timewindow)
round_starttimestamp = period.round_timestamp(
timestamp=timewindow.start(),
normalize=True
)
timewindow = timewindow.reduce(
start=round_starttimestamp,
stop=timewindow.stop()
)
# start to exclude points which are not in timewindow
points = [point for point in points if point[0] in timewindow]
if not meta:
meta = {}
transform_method = meta.get('value', {}).get('type', None)
points = self.apply_transform(points, method=transform_method)
points_len = len(points)
fn = None
# if no period and max_points > len(points)
if (not points) or self.period is None\
and self.max_points > points_len:
result = points # result is points
else: # else get the right aggregation function
fn = get_aggregations()[self.aggregation]
# if an aggregation is required
if fn is not None:
# get timesteps
timesteps = self.timesteps(timewindow)
# initialize variables for loop
i = 0
values_to_aggregate = []
last_point = None
# iterate on all timesteps in order to get points in [prev_ts, ts[
for index in range(1, len(timesteps)):
# initialize values_to_aggregate
values_to_aggregate = []
# set timestamp and previous_timestamp
previous_timestamp = timesteps[index - 1]
timestamp = timesteps[index]
# if no point to process in [previous_timestamp, timestamp]
if points[i][0] > timestamp:
continue # go to the next iteration
# fill the values_to_aggregate array
while i < points_len: # while there are points to process
_timestamp, value = points[i]
i += 1
# leave the loop if _timestamp is for a future aggregation
if _timestamp > timestamp:
break
# if _timestamp is in timewindow and value is not None
# add value to list of values to aggregate
if value is not None:
values_to_aggregate.append(value)
# TODO: understand what it means :D
if self.aggregation == "DELTA" and last_point:
values_to_aggregate.insert(0, last_point)
# get the aggregated value related to values_to_aggregate
_aggregation_value = self._aggregation_value(
values_to_aggregate, fn)
# new point to add to result
aggregation_point = previous_timestamp, _aggregation_value
result.append(aggregation_point)
# save last_point for future DELTA checking
if len(values_to_aggregate) > 0:
last_point = values_to_aggregate[-1]
if i >= points_len:
break
else:
result = points
return result