def force_reference(series, ref, func=ma.mean):
"""
Force a timeseries to match a reference, in terms of frequency and first
and last dates.
Parameters
----------
series : TimeSeries
Input time series
ref : TimeSeries
Reference time series.
func : function
Function used for the conversion of series to ref.freq, if necessary.
"""
assert(isinstance(ref, TimeSeries) and isinstance(series, TimeSeries),
"Inputs should be valid TimeSeries object ! (got '%s' and '%s')" % \
(type(ref), type(series)))
(sfreq, rfreq) = (series.freq, ref.freq)
if sfreq > rfreq:
series = series.convert(rfreq, func=func)
elif sfreq < rfreq:
series = backward_fill(series.convert(rfreq))
if (series.dates[0] != ref.dates[0]) or (series.dates[-1] !=
ref.dates[-1]):
return ts.adjust_endpoints(series, ref.dates[0], ref.dates[-1])
return series
def set_ensoindicator(self, indicator):
if indicator is not None:
if not isinstance(indicator, ENSOIndicator):
raise TypeError("The input is not a valid ENSOIndicator ! " "(got '%s' instead)" % type(indicator))
# Reset the frequency
(sfreq, ifreq) = (self._dates.freq, indicator.freq)
if ifreq != sfreq:
if sfreq < ifreq:
indicator = indicator.convert(sfreq, func=ma.mean)
elif sfreq > ifreq:
indicator = backward_fill(indicator.convert(sfreq))
# Set the ENSO indices to the default (if we don't have any...)
if indicator._cachedcurrent is None:
indicator.set_indices()
# Reset the dates
dates = self._dates
if dates.isfull():
if len(dates) > 0:
(start, end) = dates.flat[[0, -1]]
if tuple(indicator._dates.flat[[0, -1]]) != (start, end):
indicator = ts.adjust_endpoints(indicator, start, end)
else:
indicator = None
else:
indicator = indicator[self._dates]
self._optinfo["ensoindicator"] = indicator
def _fix_cachedcurrent(self, ensoindices):
"""
Private function to convert cached ENSO indices.
This function is used to ensure that:
* the frequency of the indices matches the frequency of the indicator.
* the date range of the indices matches the date range of the indicator.
* the shape of the indices matches the shape of the indicator.
"""
_currentfreq = ensoindices._dates.freq
_freq = self._dates.freq
# Check the frequency, and convert if needed
if _currentfreq > _freq:
# To a lower frequency ('Q'/'A')
if self.ndim == 2:
conversionfunc = None
else:
conversionfunc = lambda x: mode(x)[0].squeeze()
ensoindices = ensoindices.convert(_freq, func=conversionfunc)
elif _currentfreq < _freq:
# To a higher frequency (eg, 'D')
ensoindices = backward_fill(ensoindices.convert(_freq))
# Reset to the original frequency if needed...
(start, end) = self._dates.flat[[0, -1]]
if tuple(ensoindices._dates.flat[[0, -1]]) != (start, end):
ensoindices = ts.adjust_endpoints(ensoindices, start_date=start, end_date=end)
# Reset the shape of the indices
if ensoindices.shape != self.shape:
ensoindices.shape = self.shape
return ensoindices
def force_reference(series, ref, func=ma.mean):
"""
Force a timeseries to match a reference, in terms of frequency and first
and last dates.
Parameters
----------
series : TimeSeries
Input time series
ref : TimeSeries
Reference time series.
func : function
Function used for the conversion of series to ref.freq, if necessary.
"""
assert(isinstance(ref, TimeSeries) and isinstance(series, TimeSeries),
"Inputs should be valid TimeSeries object ! (got '%s' and '%s')" % \
(type(ref), type(series)))
(sfreq, rfreq) = (series.freq, ref.freq)
if sfreq > rfreq:
series = series.convert(rfreq, func=func)
elif sfreq < rfreq:
series = backward_fill(series.convert(rfreq))
if (series.dates[0] != ref.dates[0]) or (series.dates[-1] != ref.dates[-1]):
return ts.adjust_endpoints(series, ref.dates[0], ref.dates[-1])
return series
def _fix_cachedcurrent(self, ensoindices):
"""
Private function to convert cached ENSO indices.
This function is used to ensure that:
* the frequency of the indices matches the frequency of the indicator.
* the date range of the indices matches the date range of the indicator.
* the shape of the indices matches the shape of the indicator.
"""
_currentfreq = ensoindices._dates.freq
_freq = self._dates.freq
# Check the frequency, and convert if needed
if _currentfreq > _freq:
# To a lower frequency ('Q'/'A')
if self.ndim == 2:
conversionfunc = None
else:
conversionfunc = lambda x: mode(x)[0].squeeze()
ensoindices = ensoindices.convert(_freq, func=conversionfunc)
elif _currentfreq < _freq:
# To a higher frequency (eg, 'D')
ensoindices = backward_fill(ensoindices.convert(_freq))
# Reset to the original frequency if needed...
(start, end) = self._dates.flat[[0, -1]]
if tuple(ensoindices._dates.flat[[0, -1]]) != (start, end):
ensoindices = ts.adjust_endpoints(ensoindices,
start_date=start, end_date=end)
# Reset the shape of the indices
if ensoindices.shape != self.shape:
ensoindices.shape = self.shape
return ensoindices
def set_ensoindicator(self, indicator):
if indicator is not None:
if not isinstance(indicator, ENSOIndicator):
raise TypeError("The input is not a valid ENSOIndicator ! "\
"(got '%s' instead)" % type(indicator))
# Reset the frequency
(sfreq, ifreq) = (self._dates.freq, indicator.freq)
if ifreq != sfreq:
if sfreq < ifreq:
indicator = indicator.convert(sfreq, func=ma.mean)
elif sfreq > ifreq:
indicator = backward_fill(indicator.convert(sfreq))
# Set the ENSO indices to the default (if we don't have any...)
if indicator._cachedcurrent is None:
indicator.set_indices()
# Reset the dates
dates = self._dates
if dates.isfull():
if len(dates) > 0:
(start, end) = dates.flat[[0, -1]]
if tuple(indicator._dates.flat[[0, -1]]) != (start, end):
indicator = ts.adjust_endpoints(indicator, start, end)
else:
indicator = None
else:
indicator = indicator[self._dates]
self._optinfo['ensoindicator'] = indicator
def same_tsrange_of(self,_ints,*_args):
args=_args[0]
logger.debug('{SAME_TSRANGE_OF} _ints=%s %s',_ints,args)
args=args[0]
if type(args) in (ets.Timeseries,):
freq = _ints._data.freqstr
start_date = args._data.start_date.asfreq(freq)
end_date = args._data.end_date.asfreq(freq)
e = min(_ints.data.end_date,end_date)
_outs = ts.adjust_endpoints(_ints.data,
start_date=start_date,
end_date=e,
copy=True)
_res = ets.Timeseries(data=_outs)
return _res
def _set_annual_indices(self, minimum_size=None, reference_season=None):
"""
Sets the ENSO indices per periods of 12 months, starting at the first
month of the reference season if any, otherwise at October.
The same steps are followed as for :meth:`set_monthly_indices`.
Parameters
----------
minimum_size : {None, int}, optional
Minimum size for the groups of consecutive values.
If None, defaults to :attr:`minimum_size`.
reference_season : {None, string or sequence}, optional
Reference season.
If None, defaults to :attr:`reference_season`.
See Also
--------
:meth:`set_monthly_indices`
Sets the ENSO indices for each month.
"""
# Get the monthly indices .....
_monthly = self.set_monthly_indices(minimum_size=minimum_size,
reference_season=reference_season)
# Make sure we reset the full_year flag to True (we lost it w/ set_monthly
self.full_year = True
# Get the annual indices
refseason = self.refseason
if refseason:
_annual = _monthly[self.months == refseason[0]]
refseason = months2code(refseason)
else:
_annual = _monthly[self.months == 10]
_annual = adjust_endpoints(forward_fill(fill_missing_dates(_annual)),
self._dates[0], self._dates[-1])
# Cache the results ...........
self._cachedmonthly['indices_annual'] = _annual
return _annual
def _set_annual_indices(self, minimum_size=None, reference_season=None):
"""
Sets the ENSO indices per periods of 12 months, starting at the first
month of the reference season if any, otherwise at October.
The same steps are followed as for :meth:`set_monthly_indices`.
Parameters
----------
minimum_size : {None, int}, optional
Minimum size for the groups of consecutive values.
If None, defaults to :attr:`minimum_size`.
reference_season : {None, string or sequence}, optional
Reference season.
If None, defaults to :attr:`reference_season`.
See Also
--------
:meth:`set_monthly_indices`
Sets the ENSO indices for each month.
"""
# Get the monthly indices .....
_monthly = self.set_monthly_indices(minimum_size=minimum_size, reference_season=reference_season)
# Make sure we reset the full_year flag to True (we lost it w/ set_monthly
self.full_year = True
# Get the annual indices
refseason = self.refseason
if refseason:
_annual = _monthly[self.months == refseason[0]]
refseason = months2code(refseason)
else:
_annual = _monthly[self.months == 10]
_annual = adjust_endpoints(forward_fill(fill_missing_dates(_annual)), self._dates[0], self._dates[-1])
# Cache the results ...........
self._cachedmonthly["indices_annual"] = _annual
return _annual
def weighted(self,*args):
R = None
(weights,[data,series]) = args
logger.debug("Params are (w=%s,d=%s,s=%s)" % (weights,data,series))
if check_type(weights,str,unicode) and check_type(data,str,unicode) and check_type(series,str,unicode):
logger.debug("Params are (w=%s,d=%s,s=%s)" % (weights,data,series))
Ws = weights
if ',' in series:
Ss = series.split(',')
else:
raise ValueError,"La stringa di definizione delle serie è errata"
if ',' in weights:
Ws = weights.split(',')
else:
if '$$' not in weights:
raise ValueError,"La stringa di definizione dei pesi non permette la generazione di un elenco"
Ws = [ weights.replace('$$',S) for S in Ss ]
if ',' in data:
Ds = data.split(',')
else:
if '$$' not in data:
raise ValueError,"La stringa di definizione dei dati non permette la generazione di un elenco"
Ds = [ data.replace('$$',S) for S in Ss ]
if len(Ws)==len(Ds):
W = 0.0
S = 0.0
for d,w in zip(Ds,Ws):
wb=IS.has_key(w)
db=IS.has_key(d)
if wb and db:
Wn = IS[w].data[0]
Dn = IS[d].data
try:
# if isinstance(Dn,ts.TimeSeries):
# Dn = np.nan_to_num(Dn)
# if isinstance(W,ts.TimeSeries):
# W = np.nan_to_num(W)
# if isinstance(W,ts.TimeSeries):
# ts.align_series(W,Dn)
if isinstance(Dn,ts.TimeSeries):
Dn = ts.time_series(Dn,copy=True)
if isinstance(W,ts.TimeSeries):
Dn = ts.convert(Dn,W.freq)
ts.fill_missing_dates(Dn,dates=W.dates,fill_value=np.nan)
Dn = ts.adjust_endpoints(Dn,
start_date=W.start_date,
end_date=W.end_date,
copy=True)
# _ts = np.ma.masked_invalid(Dn)
# Dn = _ts.filled(0.0)
# if isinstance(W,ts.TimeSeries):
# print "GOò"
# _report(W,Dn*Wn)
W += Dn*Wn
# if isinstance(W,ts.TimeSeries):
# print "="
# _report(W)
except Exception, exc:
# if isinstance(W,ts.TimeSeries):
# _report(W)
# print "DN"
# _report(Dn)
# print w,d,"W",type(W),W,W.shape,"D",type(Dn),Dn,Dn.shape,"Wn",type(Wn),Wn
raise
logger.error('Non posso comporre %s * %s | %s', d,w)
ValueError, "%s * %s" % (d,w)
S += Wn
else:
if not wb: logger.error(u'la serie %s non è presente nell\'IS. ATTENZIONE i risultati dell\'aggregazione possono non essere sono attendibili',w)
if not db: logger.error(u'la serie %s non è presente nell\'IS. ATTENZIONE i risultati dell\'aggregazione possono non essere sono attendibili',d)
try:
R = ets.Timeseries(data=W / S,name="WEIGHTED(\"%s\",\"%s\")" % (','.join(Ws),','.join(Ds)))
except ZeroDivisionError, exc:
logger.warn('ZeroDivisionError')
R = None
