1D scipy.interpolate.griddata using method=nearest produces nans outside boundaries

[revoltek]

When the coordinates are 1D the nearest method produces nans instead of the closest values when outside boundaries. An example:

import numpy as np
from scipy.interpolate import griddata

target_points = [1.,2.,3.,4.,5.,6.,7.]
points = np.random.rand(50)*2*np.pi
values = np.sin(points)

interp = griddata(points, values, target_points, method='nearest')
print interp

The last value printed is a NaN. Note that linear/cubic modes are expected to give NaNs outside the convex hull, but this should not be the case for the 'nearest' mode.

[maniteja123]

The documentation says that fill_value has no effect for nearest method here. Does that mean there won't be a fill_value for nearest method or that NaN isn't the default fill_value ?

In the above example, the function interp1d is called here, rather than NearestNDInterpolator since ndim is 1.

The value of interpolated y_new at _call_nearest stage is

[[ 0.85659341]
 [ 0.96631064]
 [ 0.17542803]
 [-0.78798655]
 [-0.98080843]
 [-0.25246329]
 [-0.05942813]] 

But the call in griddatahere performs _evaluate call, which does the _check_bounds that fills the last value as NaN.

[[ 0.85659341]
 [ 0.96631064]
 [ 0.17542803]
 [-0.78798655]
 [-0.98080843]
 [-0.25246329]
 [              nan]] 

In some instances, there were even more than one NaNs in the output.

Just pointing out that _evaluate doesn't take into account whether the method was linear, cubic or nearest. Hope it helps and correct me if I am wrong since I am not that familiar with this.