Furthermore for ease of documentation, i.e. same grid as all other examples, this nan-handling
plot(np.nan_to_num(d), selection, secondary_source)
helps to plot. NaN presumably occurs since grid and secondary source position coincide. Is there a more elegant way to avoid this. Anyway, I'd vote to change the grid for these examples.

I didn't check all cases, but nan is already in the driving signals.
Presumably this is due to the Hankel function which returns nan for argument 0.

The reason for the first problem is that np.ceil() returns floating point values instead of integers.

There are numerous issues about that, e.g.:

• On Python3, round with one argument should return an integer type. numpy/numpy#5700
• np.ceil and np.floor are inconsistent with math.ceil and math.floor numpy/numpy#9068
• numpy.float64.__round__ is incompatible with float.__round__ in python3 numpy/numpy#11557
• round() returns floating point, not int, for some numpy floats when no second arg numpy/numpy#11810

I see two options:

import math
Nc = math.ceil(2 * k * max(r) * alpha/np.pi)

or

Nc = np.ceil(2 * k * np.max(r) * alpha/np.pi).astype(int)

Regarding NaN values, I don't think we should convert them to zero, right?

Since now the ESA driving functions are in their own module, it's no problem to choose a different grid and/or array for the examples.

Regarding NaN values, I don't think we should convert them to zero, right?

Yeah, that would be inappropriate. We rather should check what precisely happens. I have it on my list.

Since now the ESA driving functions are in their own module, it's no problem to choose a different grid and/or array for the examples.

Yeah, that's nice about the new handling. I will cover this as well ASAP.

I see two options:

import math
Nc = math.ceil(2 * k * max(r) * alpha/np.pi)

or

Nc = np.ceil(2 * k * np.max(r) * alpha/np.pi).astype(int)

I'd vote for the second one, then we give a more clearer hint, that something is happening and we explicitly take care of it. The first solution is blurry, because it might only raise wondering why the coders did not use np.ceil...which is good, but we don't tell then.

I see two options:

import math
Nc = math.ceil(2 * k * max(r) * alpha/np.pi)

or

Nc = np.ceil(2 * k * np.max(r) * alpha/np.pi).astype(int)

I'd vote for the second one, then we give a more clearer hint, that something is happening and we explicitly take care of it. The first solution is blurry, because it might only raise wondering why the coders did not use np.ceil...which is good, but we don't tell then.

I also like the second variant!