1D quiver in non-PlateCarree projections

[ocefpaf]

Using quiver with 1D arrays of x, y, u, v works just fine with ccrs.PlateCarree(), but raises a ValueError with any other projection:

ValueError: x, y, u and v arrays must be the same shape

Using np.atleast_2d fixes this, and the quiver docs do say: Plot a 2-D field of arrows.

However, I wonder if the atleast2d should be inside Cartopy, or if the 1D case should raise an error in PlateCarree too.

(See http://nbviewer.ipython.org/gist/ocefpaf/b2f79fe0f96fa1776b97 for a SSCCE.)

[ocefpaf]

BTW: This bug was found by a student @rafacsantana during our Python course here. Hopefully the students will be sending PRs of their own soon 😉

[pelson]

👍 - let's fix it.

[pelson]

Ping @scmc72 / @Jozhogg if you're interested.

[JosephHogg]

Taking a look now 👍

[ocefpaf]

@Jozhogg take a look at:

#674

[JosephHogg]

Thanks @ocefpaf . Didn't spot your PR at first but I just made that same change myself! It certainly makes sense to me - transform_vectors is ok with everything being 1D so I feel like it makes sense to have that capability available in quiver. Opinions on that fix @pelson ?

[pelson]

Apologies to @ocefpaf and @Jozhogg. I hadn't seen the PR (or at least, hadn't linked the two).
Looks like #674 will do the trick. 👍

[ocefpaf]

Apologies to @ocefpaf and @Jozhogg.

My bad. I sent this after a long day and forgot to cross-reference issue/PR.