def merge_variable(fills, var):
return merge_variable_interpolated(fills, var)
xmin = sys.maxsize
xmax = -xmin
for fill in fills:
xmin = min(fill.data[var].x.min(), xmin)
xmax = max(fill.data[var].x.max(), xmax)
# The x-series
newx = np.arange(
xmin, xmax + 1, 1,
dtype=np.int) # using int's to maybe make comparisons more safe
# Will store the average values of y
avgy = np.zeros(newx.size, dtype=np.float)
norm = np.zeros(
newx.size, dtype=np.float
) # we use this to basically take the average value at each timestamp
# Will store maximum and minimum values of y
maxy = np.zeros(newx.size, dtype=np.float)
miny = 1.0e10 * np.ones(newx.size, dtype=np.float)
for fill in fills:
x_i = fill.data[var].x.astype(np.int)
index_map = np.searchsorted(newx, x_i)
np.add.at(avgy, index_map, fill.data[var].y)
np.add.at(norm, index_map, 1.0)
np.maximum.at(maxy, index_map, fill.data[var].y)
np.minimum.at(miny, index_map, fill.data[var].y)
# delete empty values (where no data was written)
to_delete = np.where(norm < 0.5)
newx = np.delete(newx, to_delete)
maxy = np.delete(maxy, to_delete)
miny = np.delete(miny, to_delete)
norm = np.delete(norm, to_delete)
avgy = np.delete(avgy, to_delete)
avgy /= norm
newx = newx.astype(dtype=float)
return {
"average": Fill.Variable((newx, avgy)),
"max": Fill.Variable((newx, maxy)),
"min": Fill.Variable((newx, miny))
}
def merge_variable_interpolated(fills, var, resolution=0.1):
xmin = -sys.maxsize
xmax = -xmin
for fill in fills:
xmin = max(fill.data[var].x.min(), xmin)
xmax = min(fill.data[var].x.max(), xmax)
newx = np.arange(math.ceil(xmin), math.floor(xmax), resolution)
avg = np.zeros(newx.size)
maxy = np.zeros(newx.size)
miny = np.full(maxy.shape, sys.maxsize, dtype=float)
for fill in fills:
ip = interpolate.interp1d(*fill.data[var], assume_sorted=True)(newx)
avg += ip
maxy = np.maximum(maxy, ip)
miny = np.minimum(miny, ip)
avg /= len(fills)
return {
"average": Fill.Variable((newx, avg)),
"max": Fill.Variable((newx, maxy)),
"min": Fill.Variable((newx, miny))
}