1110 join single point error

[DLafayetteII]

Changes

Allowing join to succeed with data of shape (1,). Previously failed due to specific handling of data where np.ndim(data.variable) = 0. See issue #1110 for more info and the example failure case. Unsure of the motivation behind this specific handling of (1,) shape data, but this fix makes the join method work as expected.

Checklist

• added tests, if applicable
• updated documentation, if applicable
• updated CHANGELOG.md
• tests pass

[kameyer226]

Code seems good, only move transpose.append statement in front of if...else since it is needed for both (all) cases, and remove the "no transpose needed" phrase

[kameyer226]

Just a moment, a new issue may appear with this fix. What if the variable of shape (1,) has a different value for each of the arrays being joined? If they are different, the join should fail, shouldn't it?

[DLafayetteII]

Just a moment, a new issue may appear with this fix. What if the variable of shape (1,) has a different value for each of the arrays being joined? If they are different, the join should fail, shouldn't it?

I am realizing that I have thus far only considered when the variable of shape (1,) is the transformed axis along which you are joining. In this case, it is fine if the values are different. If the shape (1,) variable is not your transformed axis along which you join, I want to say there's still no problem, but I have not tested and do not have a strong intuition.

[kameyer226]

Yes, I should have stated the (1,) variable is the transformed axis along which to join. But why would it be fine if they are different? If they are different, isn't that saying they are not to be joined on top of each other?

[kameyer226]

It looks good.

[ksunden]

I'm a little confused as to how you got a scalar in the first place... I don't think we officially have supported such data objects, as their behavior is just different enough in all of these situations...

(Certainly you can have a size 1 data object with shape [1,1,1] for however many ndim you have, just thought scalar was not something we officially supported...)

[ksunden]

note that scalar and shape (1,) are two different ideas, the latter will not have np.ndim(var) == 0

[ksunden]

I would tend to want a test to show the failing behavior (and prevent it from recurring in the future... if we missed an edge case once, we could do so again)

[kameyer226]

The variable was originally shaped (1,) prior to join, so it was never a scalar. I tried joining data objects consisting of scalar variables and the attempt failed.

[DLafayetteII]

note that scalar and shape (1,) are two different ideas, the latter will not have np.ndim(var) == 0

This misconception is the heart of the issue. We do np.ndim(var.points), and np.ndim(var.points)==0, so scalar and shape(1,) get treated the same in the code. Maybe we should change the code to do np.ndim(var) instead? @ksunden