plot(x, xmin=1, xmax=-1) comes up empty

[ppurka]

The following two plots

plot(x, xmin=1, xmax=-1)
plot(x).show(xmin=1, xmax=-1)

give different outputs. In particular, the plot comes up empty in the
former one. This seems to be a conscious decision taken in #5651. The main
line seems to be this line in sage/plot/plot.py and similar lines in
several other files.

    G_kwds = Graphics._extract_kwds_for_show(kwds, ignore=['xmin', 'xmax'])

Is there a reason for this? Should this be "fixed?"

CC: @kcrisman @sagetrac-jakobkroeker @paulmasson @sagetrac-jhonrubia6 @novoselt @alauve @tscrim @jm58660

Component: graphics

Author: Paul Masson

Branch/Commit: u/paulmasson/13368 @ f0eebbb

Reviewer: Javier Honrubia González

Issue created by migration from https://trac.sagemath.org/ticket/13368

[kcrisman]

comment:1

One data point is an email I got today (unrelated to Sage development, just use) asking why

plot(1/x,xmin=-5,xmax=5)

didn't result as expected. So maybe we should do more than ignore them, but raise an error or something...

---

NO! My belief is that this is a typo. Notice that in #5651, the places where this is ignored are ALL some kind of contour or "rectangular" plot, except for plot.py, but then plot_field.py does not have this, even though it's this kind of plot. In fact, we probably should support things like

sage: plot(x,(x,1,5),xmin=2)

since we might want to keep the data but not show it.

[kcrisman]

comment:2

Though we'd probably have to deal with

sage: plot(x,(x,1,5),xmin=2).show(xmin=3)

somehow...

[ppurka]

comment:3

with ymin, the parameter in show overrides the parameter in plot.

[pjbruin]

comment:6

An example from #5124 (duplicate of this ticket; note that this was created before #5651):

E = EllipticCurve('37a')
E.plot(xmin=-10, xmax=10)

[ppurka]

comment:7

I think I had some idea how to fix this. If I get some time I will look at it again. It will not be any time soon, unfortunately. Essentially, we just need to mimic how ymin and ymax are handled.

[kcrisman]

comment:11

Just came across this again. Really, all relevant options should be passed to show!

[paulmasson]

comment:12

As of 7.3.beta8 evaluating plot(x, xmin=-1, xmax=1) and plot(1/x,xmin=-5,xmax=5,ymin=-5,ymax=5) both produce the expected graphs.

Evaluating plot(x, xmin=1, xmax=-1) with a reversed domain does not produce a plot. Evaluatingplot(1/x,xmin=-5,xmax=5,ymin=5,ymax=-5) with an inverted range flips the plot upside down.

Shouldn't this ticket be first about a sanity check on domain and range? Is there a good reason to allow reversal and inversion?

[ppurka]

comment:13

Replying to @paulmasson:

Evaluating plot(x, xmin=1, xmax=-1) with a reversed domain does not produce a plot. Evaluatingplot(1/x,xmin=-5,xmax=5,ymin=5,ymax=-5) with an inverted range flips the plot upside down.

Second command should be the desired output.

Shouldn't this ticket be first about a sanity check on domain and range? Is there a good reason to allow reversal and inversion?

A good reason to allow flipped axis is to have axes that are decreasing instead of increasing. It is already possible by using plot(...).show(xmax=-1,xmin=1). This ticket was about making the behavior of plot consistent with what plot(...).show() does.

[paulmasson]

Branch: u/paulmasson/13368

[paulmasson]

Commit: 0f674f2

[paulmasson]

comment:15

He's a solution to this problem. Part of the reason the plots were empty is that the function setup_for_eval_on_grid() in plot/misc.py outputs plotting steps as an absolute value, not a signed value. I've removed that and modified the doctests accordingly, but six other plotting files use that function in addition to plot.py.

All plotting doctests pass and documentation all builds. I scanned all seven plotting documentation HTML files that use setup_for_eval_on_grid() and didn't see anything strange, but other eyes need to look at them closely.

Adding Javier to cc list since he's been working on plotting examples and might spot something wrong.

---

New commits:

46d5778	Pass xmin and xmax to show()
0f674f2	Fix doctests after removing 'abs'

[paulmasson]

comment:19

Andrey, I haven't changed any keywords and don't plan to. All I did was remove abs so that range_steps can take negative as well as positive values. This is necessary for creating plots in plot.py when xmin > xmax.

The question is whether the change of allowing range_steps to be negative will have an unintended impact on other plotting routines. I don't expect you to read all of the plotting code, but if anything else comes to mind let me know. Thanks!

[sagetrac-git]

Changed commit from 0f674f2 to c3a7f71

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

c3a7f71

Rebase for consistency with #20924

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

2b85a57	Pass xmin and xmax to show()
f0eebbb	Fix doctests after removing 'abs'

[sagetrac-git]

Changed commit from c3a7f71 to f0eebbb

[paulmasson]

comment:22

Code rewritten for consistency with #20924

[sagetrac-jhonrubia6]

comment:23

everything looks good to me. As I see no negative vote, I'll make it a positive review

[tscrim]

comment:24

Reviewer name is needed.

[sagetrac-jhonrubia6]

Reviewer: Javier Honrubia González

[vbraun]

comment:26

sage -t --long src/sage/schemes/elliptic_curves/ell_generic.py  # 1 doctest failed

see patchbot

[paulmasson]

comment:27

The changes on this ticket produce unexpected results for elliptic curves. Needs to be investigated.

[jm58660]

comment:30

Maybe not related, but why "e" in "2.717e" in

plot(x/log_b(x, 2), (2.717, 2.72))

[kcrisman]

comment:31

Maybe not related, but why "e" in "2.717e" in

Indeed unrelated, but also worth a ticket. We don't do a good job dealing with scientific notation that ends up in the tick labels, and you can see that if you try plotting things with really high output values but in a narrow range. For example:

plot(x^2,(x,10^6,10^6+1))