[Bugfix] Cast yaw angles to np.ndarray on set

[paulf81]

[Bugfix] Fix yaw setting

In the course of some other work, I noticed that a few conditions that I would've expected to be ok throw an error:

• Passing a list to yaw_angles in fi.set
• If the values of yaw_angles are integers instead of floats

This is not so hard to fix but I wasn't positive the preferred solution (attrs?). So I thought I'd start just by adding the failing tests and then we can agree on how we want to fix it.

Other things to consider:

• Does this also impact power setpoints? No, tested
• Does this also impact disable turbines? No, tested
• Does fixing this in FlorisInterface intrinsically fix the problem in Parallel and Uncertain interface? It seems to, yes. I tried this on the Update uncertainty interface to 4d, new API #821 (implemented the fixed, and then altered the test to take yaw angles as a list), and it worked.

Note this issue is a little bit being worked-around in #821

[rafmudaf]

See https://github.com/NREL/floris/blob/v4/floris/simulation/farm.py#L332

[misi9170]

The fix was straightforward here---in _set_operation(), I've now wrapped the assignment of the yaw_angles onto self.floris.farm.yaw_angles in a np.array() statement.

Incidentally, this is something that @rafmudaf warned me about---that reaching in and altering the attributes (rather than instantiating them properly, which would pass through __attrs__) can lead to unintended consequences. For now, this fix works; but we should possibly consider adding setter methods to Farm to avoid this happening again.

[misi9170]

@paulf81 I can't add you as a reviewer, since you opened the PR, but please look through and see what you think.

[paulf81]

This works great for me, thank you @misi9170 I "approve"

[rafmudaf]

Is this the only place where yaw angles are going to be set? If not, you might consider adding a setter method on Farm to do this always:

class Farm:
    def set_yaw_angles(yaw_angles):
        # Do lots of checks
        ...
        # Cast to Numpy array
        self.yaw_angles = np.array(yaw_angles)

You could also use the @property decorator the way we used to do, if you think maintaining the attribute-style use is important.

[rafmudaf]

In fact, there's already a function with this name at https://github.com/NREL/floris/blob/v4/floris/simulation/farm.py#L332

[rafmudaf]

Ah and I just read #828 (comment), sorry to say the same thing @misi9170

[misi9170]

No problem---I'm going to try to do this (use a setter method), it seems like a better approach