Typechecking within parcels to provide better error messages

[VeckoTheGecko]

I think it would be good to enforce types within the user level API (eg. if not isinstance(obj, class): raise ...). Some error messages are very uninformative otherwise.

Example

I wanted to combine a list of kernels down to a single combined kernel. Typing this:

import functools
combined_kern = functools.reduce(lambda x, y: pset.Kernel(x) + pset.Kernel(y), [test_kern, test_kern, test_kern])

raised this very cryptic traceback:

Exception has occurred: AttributeError       (note: full exception trace is shown but execution is paused at: _run_module_as_main)
'KernelSOA' object has no attribute '__name__'
  File "C:\Users\nickh\anaconda3\envs\parcels_hons\Lib\site-packages\parcels\kernel\basekernel.py", line 75, in __init__
    self.funcname = funcname or pyfunc.__name__
  File "C:\Users\nickh\anaconda3\envs\parcels_hons\Lib\site-packages\parcels\kernel\kernelsoa.py", line 47, in __init__
    super(KernelSOA, self).__init__(fieldset=fieldset, ptype=ptype, pyfunc=pyfunc, funcname=funcname, funccode=funccode, py_ast=py_ast, funcvars=funcvars, c_include=c_include, delete_cfiles=delete_cfiles)
  File "C:\Users\nickh\anaconda3\envs\parcels_hons\Lib\site-packages\parcels\particleset\particlesetsoa.py", line 650, in Kernel
    return Kernel(self.fieldset, self.collection.ptype, pyfunc=pyfunc, c_include=c_include,
  File "C:\Users\nickh\oceans\repos\honours-modelling\my_modelling\parcels_custom.py", line 213, in <lambda>
    combined_kern = functools.reduce(lambda x, y: pset.Kernel(x) + pset.Kernel(y), [test_kern, test_kern, test_kern])
  File "C:\Users\nickh\oceans\repos\honours-modelling\my_modelling\parcels_custom.py", line 213, in <module>
    combined_kern = functools.reduce(lambda x, y: pset.Kernel(x) + pset.Kernel(y), [test_kern, test_kern, test_kern])

The underlying error is that pset.Kernel is applied twice by the lambda (once on the left, and once on the right) and the Kernel method can't handle this as it does not verify that pyfunc is Callable:
https://github.com/OceanParcels/parcels/blob/7f6610417c948e85d65d98704db36e55e505b5a7/parcels/particleset/particlesetsoa.py#L644-L651

Clarifying this by having raising an error ("pyfunc must be Callable") or having pset.Kernel(kern) return kern would help here. In general, having this across the codebase would prevent similar errors.

Of course, my example can be solved by doing:

import functools
combined_kern = functools.reduce(lambda x, y: x+y, (pset.Kernel(i) for i in [test_kern, test_kern, test_kern]))

[erikvansebille]

Good catch, @VeckoTheGecko. Could you add this check in the code yourself and create a PR?