Plots.jl: Quiver affected by another plot command

[dpsanders]

There's something strange going on with the last quiver here:

https://gist.github.com/dpsanders/76fdf876b098a29efbf3e447fe8a1837

The result of the quiver seems to be affected by what else is being plotted.

EDIT: Reported by a student.

[dpsanders]

The two quiver calls are (as far as we can tell) identical, but the results are different.

[mkborregaard]

I'm not sure, why are the two calls identical? S differs because tspan differs, so you're passing some different data. Am I reading this wrong?

[mkborregaard]

No I see what you mean, it's really quite strange. Especially as the arrows series of the first and last Plot object are still identical. Also, it appears to work in pyplot - the error for me comes when using GR. Can you confirm that it works on the other backends for you too?

[apalugniok]

The function GR.drawarrow seems to have issues with xflip = true and yflip too. I think that is what is causing the issue.

using Plots; gr()
quiver([1], [1], quiver = [1,1], xflip = true)

The expected arrow direction, I think, comes from just GR.polyline since I can't see an arrowhead. GR.drawarrow then draws the arrow ignoring the flip.

The line and arrow drawing happens here https://github.com/JuliaPlots/Plots.jl/blob/495753b5170e73dfc060cd7574aee74fbb6d3151/src/backends/gr.jl#L180-L192

I think this is a GR bug and not us?

[sdewaele]

I am also experiencing an issue when using yflip on GR, as @apalugniok reports. GR seems to igore the yflip.