Remove numpy warnings in slope field

[kcrisman]

Having no arrows in an arrow plot (i.e., vector field that is a slope field) now causes warnings from matplotlib.

y = var('y') 
g = 1 
P=plot_slope_field(g,(x,3,4),(y,-1,1)) 
P 
<two sets of warnings> 

DSM has a diagnosis at this sage-support thread

P=plot_slope_field(g,(x,3,4),(y,-1,1),headlength=1e-8) 
works for me.  FYI, it's the following few lines in Quiver._h_arrows at fault: 
        minsh = self.minshaft * self.headlength 
        [....] 
        shrink = length/minsh 
        X0 = shrink * X0[np.newaxis,:] 
        Y0 = shrink * Y0[np.newaxis,:] 
Probably we should change the defaults and/or (if it's not done 
already) ask our matplotlib friends to special-case 0 for no 
arrowheads.

So maybe this should be reported upstream? See also #2922, which is also about the mpl quivers.

Depends on #10489

Upstream: Reported upstream. Little or no feedback.

CC: @jasongrout @novoselt

Component: graphics

Author: Douglas McNeil

Reviewer: David Loeffler, Karl-Dieter Crisman

Merged: sage-5.0.beta9

Issue created by migration from https://trac.sagemath.org/ticket/11208

[jasongrout]

Upstream: Not yet reported upstream; Will do shortly.

[jasongrout]

comment:1

This should be reported upstream, right?

[kcrisman]

comment:2

Probably, and I say so on sage-support, but I don't know how to make this pure matplotlib, and maybe they would say it's user error? But you are almost certainly right.

[kcrisman]

comment:3

I don't think this was ever reported upstream. And now it's going to cause a problem for our PREP calculus tutorial... grumble, grumble.

[sagetrac-mhampton]

comment:4

For the moment it seems OK to use Doug McNeil's hack of setting: headlength=1e-8
Its better than leaving this problem around any longer in my opinion.

[sagetrac-dsm]

comment:6

Patch attached. Wasn't sure of the idiom for doctesting the warnings; went with what I would have done in a non-Sage project.

[sagetrac-dsm]

headlength = epsilon hack

[loefflerd]

Dependencies: #10489

[loefflerd]

Reviewer: PatchBot

[loefflerd]

comment:7

Attachment: trac_11208_avoid_headless_quiver_warning.patch.gz

This doesn't apply to the current Sage dev version -- it seems to conflict with #10489. Can you rebase it?

[sagetrac-dsm]

Changed upstream from Not yet reported upstream; Will do shortly. to Reported upstream. Little or no feedback.

[kcrisman]

Changed reviewer from PatchBot to David Loeffler, Karl-Dieter Crisman

[kcrisman]

comment:9

On the plus side, it works! DSM, just be sure to not replace the previous patch so that we can compare the two, but the rebase should be nearly trivial (I almost did it just now, but decided you were probably in the midst). The only thing I could possibly say is that I would put the "(trac #11208)::" on the next line, but it's not really that important.

So positive review, modulo rebase.

[kcrisman]

Author: Douglas McNeil

[sagetrac-dsm]

headlength = epsilon hack, rebased to 5.0.beta7

[sagetrac-dsm]

comment:10

Attachment: trac_11208_avoid_headless_quiver_warning_v2.patch.gz

Dropped a message on matplotlib-devel; at least one dev thought special-casing zero was a good idea, so I'll work up a matplotlib patch if I get some time. Not much pressure to do it on our end now, though. [PS @kcrisman -- not only was I in the midst, I was typing things like 'hg export tip' as you were writing that..]

[kcrisman]

comment:11

Great, make sure to post a link to the discussion on matplotlib-devel. This applies to beta7 and fixes something we REALLY should have fixed a long time ago.

By the way, it turns out we now have this nice :trac: thing for documentation... not needed here, but just fyi for next time, I just found out about five minutes ago.

[loefflerd]

comment:12

Apply trac_11208_avoid_headless_quiver_warning_v2.patch

(for patchbot)

[jdemeyer]

Merged: sage-5.0.beta9

[kcrisman]

comment:14

Upstream report is here.