Question

I’m trying to interpolate with the following code

    self.indeces = np.arange( tmp_idx[len(tmp_idx) -1] )
    self.samples = np.interp(self.indeces, tmp_idx, tmp_s)

where tmp_idx and tmp_s are numpy arrays. I get the following error:

array cannot be safely cast to
required type

Do you know how to fix this?

UPDATE:

   class myClass
    def myfunction(self, in_array, in_indeces = None):
        if(in_indeces is None):
            self.indeces = np.arange(len(in_array))
        else:
            self.indeces = in_indeces       
        # clean data
        tmp_s = np.array; tmp_idx = np.array;
        for i in range(len(in_indeces)):
            if( math.isnan(in_array[i]) == False and in_array[i] != float('Inf') ):
                tmp_s = np.append(tmp_s, in_array[i])
                tmp_idx = np.append(tmp_idx, in_indeces[i])
        self.indeces = np.arange( tmp_idx[len(tmp_idx) -1] )
        self.samples = np.interp(self.indeces, tmp_idx, tmp_s)

Answer 1

One of your possible issues is that when you have the following line:

tmp_s = np.array; tmp_idx = np.array;

You are setting tmp_s and tmp_idx to the built-in function np.array. Then when you append, you have have object type arrays, which np.interp has no idea how to deal with. I think you probably thought that you were creating empty arrays of zero length, but that isn’t how numpy or python works.

Try something like the following instead:

class myClass
    def myfunction(self, in_array, in_indeces = None):
        if(in_indeces is None):
            self.indeces = np.arange(len(in_array))
            # NOTE: Use in_array.size or in_array.shape[0], etc instead of len()
        else:
            self.indeces = in_indeces       
        # clean data
        # set ii to the indices of in_array that are neither nan or inf
        ii = ~np.isnan(in_array) & ~np.isinf(in_array)
        # assuming in_indeces and in_array are the same shape
        tmp_s = in_array[ii]
        tmp_idx = in_indeces[ii] 
        self.indeces = np.arange(tmp_idx.size)
        self.samples = np.interp(self.indeces, tmp_idx, tmp_s)

No guarantees that this will work perfectly, since I don’t know your inputs or desired outputs, but this should get you started. As a note, in numpy, you are generally discouraged from looping through array elements and operating on them one at a time, if there is a method that performs the desired operation on the entire array. Using built-in numpy methods are always much faster. Definitely look through the numpy docs to see what methods are available. You shouldn’t treat numpy arrays the same way you would treat a regular python list.