def interpolate(self, step=1, kind=1):
"""Array interpolation for summary synchronization."""
vt = Reference.interpolate(self, step)
if vt is False:
return False
# Value sequence
# starting from the oldest time minus step
oldi = self.get_time(vt[0] - step)
# Check if we have enough points to interpolate
if len(self) - oldi < 5:
return False
# If possible, go back one more point, for interpolation safety
if oldi > 1:
oldi -= 1
# Decode values and separate time and value vectors
dat = self[oldi:]
# print 'Getting data',self.path,dat,vt
dat = np.array(dat)
dat = dat.transpose()
# Build a linear spline using vt points as knots
#f=LSQUnivariateSpline(dat[0],dat[1],vt, k=kind)
# Do a linear fitting
(slope, const), res, rank, sing, rcond = np.polyfit(
dat[0], dat[1], kind, full=True)
# Build a vectorized evaluator
f = np.vectorize(lambda x: slope * x + const)
while vt[0] < dat[0][0] and len(vt) > 1:
vt = vt[1:]
while vt[-1] > dat[0][-1] and len(vt) > 1:
vt = vt[:-1]
if len(vt) <= 1:
return False
try:
# Interpret time series
out = f(vt)
except:
print 'Array.interpolate', self.path, vt, dat
raise
# Encode in (t,v) append-able list
out = np.array([vt, out]).transpose()
self.summary.commit(out)
return True
