Give more helpful error when n_findex = 0

[paulf81]

Give more helpful error when n_findex = 0

Working with FLORIS and FLASC, it can sometimes happen that FLORIS is set to have n_findex = 0 (for example there are no rows in the FlascDataframe that is being assigned to FLORIS. In this cases, it could be useful for FLORIS to give a more helpful error. Right now this code (run in the examples folder)

import numpy as np

from floris import FlorisModel



fmodel = FlorisModel("inputs/gch.yaml")
fmodel.set(
    wind_directions=np.array([]), wind_speeds=np.array([]), turbulence_intensities=np.array([])
)
print(f"NFINDEX: {fmodel.n_findex}\n")
fmodel.run()

Has output:

pfleming$ python 001_opening_floris_computing_power.py 
NFINDEX: 0

Traceback (most recent call last):
  File "/Users/pfleming/Projects/FLORIS/floris/examples/001_opening_floris_computing_power.py", line 14, in <module>
    fmodel.run()
  File "/Users/pfleming/Projects/FLORIS/floris/floris/floris_model.py", line 490, in run
    self.core.initialize_domain()
  File "/Users/pfleming/Projects/FLORIS/floris/floris/core/core.py", line 147, in initialize_domain
    self.farm.initialize(self.grid.sorted_indices)
  File "/Users/pfleming/Projects/FLORIS/floris/floris/core/farm.py", line 260, in initialize
    self.awc_modes_sorted = np.take_along_axis(
                            ^^^^^^^^^^^^^^^^^^^
  File "/Users/pfleming/opt/anaconda3/envs/floris/lib/python3.11/site-packages/numpy/lib/shape_base.py", line 166, in take_along_axis
    axis = normalize_axis_index(axis, arr.ndim)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
numpy.exceptions.AxisError: axis 1 is out of bounds for array of dimension 1

Proposed solution

Raise an error on run() when n_findex = 0?

[misi9170]

I was originally going to suggest a warning, rather than an error, might be most appropriate; but finally, anyone calling run() will then want to call on of the postprocessing calls, and those calls won't work properly (or worse, may extract a previous run()'s results) if an error is not raised. So, I'm in favor of raising an error (possibly a RuntimeError?) in this situation. Any external algorithm that calls run() (but possibly with n_findex = 0) can then have a try/except clause built in to catch the error and deal with it.

[misi9170]

How closely is this linked to #1007? Could it be addressed in that PR?

[paulf81]

They could be made to be related, but that one is focused on adding error checks to WindData objects and this one to FlorisModel. We could however make #1007 more general and include fixes to that as well since they are both about dealing with NaNs and then NaNs leading to 0-length inputs

[misi9170]

In my mind, having handling in the WindData objects is more in line with the ethos proposed in #1076.

The error raised is clear enough, I think, if the user passes something like fmodel.set(wind_speeds=[], ...); fmodel.run(); but perhaps a bit less clear if the user does something like

wind_rose = WindRose(wind_speeds=[], ...)
fmodel.set(wind_data=wind_rose)
fmodel.run()

[paulf81]

In my mind, having handling in the WindData objects is more in line with the ethos proposed in #1076.

The error raised is clear enough, I think, if the user passes something like fmodel.set(wind_speeds=[], ...); fmodel.run(); but perhaps a bit less clear if the user does something like

wind_rose = WindRose(wind_speeds=[], ...)
fmodel.set(wind_data=wind_rose)
fmodel.run()

Yes I agree, this will match the pattern described in #1076. I wonder if one PR should address this and those? Seems reasonable right? We could open a template PR with the checklist of fixed it should to include to address which issues and then tick through?