Devil's advocate here: (just about) everywhere else we are able to flip the order of the limits (particular favorite is pressure-as-height). Why shouldn't we be allowed to do something similar for polar plots?

@WeatherGod Fine with me! But someone has to rewrite it so it works, which it does not right now...

That makes sense. I’m fine w an error and if someone wants to implement inverted axes they can do so in future

I agree with@timhoffm here. You don't need to implement axis inversion here, but if you allow arbitrary order now, it'll be hard to allow for axis inversion viaset_*lim in the future.

The error might point towards a possible workaround for now. For the r axis: "To mimic an inverted r-axis plot data.max()-data and set the ticklabels accordingly.". For theta axis: "Useset_theta_direction instead."

Changed to a ValueError and changed the doc string to reflect thats what will happen.

I don't know that we need to point out all the possible work arounds in an error message - they tend to be hard to describe in a line or two. But if you had something pithy, that'd be fine.

Added a small test for ValueError....

squashed and rebased

It looks like this used to work with inverted axes, but we broke it (likely inadvertently) in the polar refactor. I think we should look into fixing it before we make it raise.

Fair enough. I didn’t know it used to work. I’ll clowe and we can reopen if we decide we can’t fix it.

I don’t see any reason inverting the axis is difficult mathematically. I just assumed it wasn’t working on purpose.

I just tested all they way back to 1.5 and this never actually worked (but silently kept going).

Sorry for the noise 🐑 .

No problem - but OTOH, I don't actually see why thiscan't work. Marking WIP while I take a quick look...

OK, this seems to work fine...

importnumpyasnpimportmatplotlib.pyplotaspltfig=plt.figure()ax=fig.add_axes([0.1,0.1,0.8,0.8],polar=True)ax.plot(np.arange(30),np.arange(30)/30.+1)ax.set_ylim(2,1)# out of order, and hence didn't draw properly.plt.show()

ax.set_rorigin(3.)

Rebased....

Something funny happens if you flip the radial limits after a draw:

#!/usr/bin/env pythonimportnumpyasnpimportmatplotlib.pyplotaspltfig, (ax0,ax1,ax2)=plt.subplots(1,3,subplot_kw=dict(polar=True))ax0.plot([0,1], [-1,1])ax0.scatter([0,1], [-1,1])ax0.set_rmin(-1.5)ax0.set_rmax(1.5)ax0.set_title('Forward')ax1.plot([0,1], [-1,1])ax1.scatter([0,1], [-1,1])ax1.set_rmin(1.5)ax1.set_rmax(-1.5)ax1.set_title('Backward')ax2.plot([0,1], [-1,1])ax2.scatter([0,1], [-1,1])ax2.set_rmin(-1.5)ax2.set_rmax(1.5)ax2.set_title('Backward with draw')fig.canvas.draw()ax2.set_rmin(1.5)ax2.set_rmax(-1.5)plt.show()

@QuLogic Agreed, but that predates this PR, and is a different issue. On master:

importnumpyasnpimportmatplotlib.pyplotaspltfig, (ax0,ax2)=plt.subplots(1,2,subplot_kw=dict(polar=True))ax0.plot([0,1], [-1,1])ax0.scatter([0,1], [-1,1])ax0.set_rmin(-1.5)ax0.set_rmax(1.5)ax0.set_title('Forward')ax2.plot([0,1], [-1,1])ax2.scatter([0,1], [-1,1])ax2.set_rmin(-1.5)ax2.set_rmax(1.5)ax2.set_title('draw; change rlims')fig.canvas.draw()ax2.set_rmin(0)ax2.set_rmax(1.5)plt.show()

Typos fixed - this has two approvals and could be merged if anyone has a minute...