Fix _repr_ of graphics objects

[vbraun]

Graphical output of plots (or other graphics objects) is hooked into _repr_(). Obviously, Python doesn't expect repr to have side effects so we get fun stuff like

sage: g = Graphics()
sage: g?

showing a plot in addition to the help. Or (g,g) opening two plot output windows if used on the command line. You don't want to try [g]*100

This patch moves the decision logic into the displayhook, and makes repr always return a string representation as it should.

Also, deprecate the show_output() function. Wat?

Depends on #14203
Depends on #14266
Depends on #14471
Depends on #15016

CC: @ppurka

Component: graphics

Author: Volker Braun

Reviewer: Travis Scrimshaw

Merged: sage-5.12.beta1

Issue created by migration from https://trac.sagemath.org/ticket/14469

[vbraun]

Dependencies: #14203

[vbraun]

Author: Volker Braun

[kcrisman]

comment:5

I've totally done something analogous to [g]*100. Making lists of plots is useful, but the previous behavior was not good, and no one should have been relying on it.

I don't have time to learn about displayhook now, but the rest looks good, assuming it works properly in the notebook.

Unless... you don't suppose someone would actually want to be able to change the default behavior of (say) [g]*100 to show all these graphics in addition to printing the list? Just a crazy thought.

[vbraun]

comment:6

The IMHO only remaining problem is that the patch triggers what looks like a Python bug, see #14471.

[tscrim]

comment:7

Two things on the current patch:

• Is there a reason why you're importing DOCTEST_MODE in displayhook.py? You don't seem to be using it.
• The method _graphics_() in sage_input.py will need a doctest.

Also I think if someone wants [g]*100 to display all plots, we can just tell them to use map(lambda x: x.show(), L) where L is the above list since (to me) this is not likely to be very desirable...

[vbraun]

Changed dependencies from #14203 to #14203, #14266

[vbraun]

comment:8

Rebased for

[vbraun]

Changed dependencies from #14203, #14266 to #14203, #14266, #14471

[vbraun]

comment:10

Changed patch to not catch any exceptions that might have come from generating the plot

[ppurka]

comment:11

Just a heads up -- it is failing some doctests according to the patchbot.

[vbraun]

Updated patch

[vbraun]

comment:12

Attachment: trac_14469_repr_graphics.patch.gz

Should be fixed now.

[tscrim]

Reviewer: Travis Scrimshaw

[tscrim]

comment:13

Looks good to me.

[vbraun]

Changed dependencies from #14203, #14266, #14471 to #14203, #14266, #14471, #15016

[vbraun]

comment:15

This requires a version of the notebook that includes the commit vbraun/sagenb@89c6b6c. I've tentatively made a ticket for the upgrade to have something to put into the dependency fied.

[vbraun]

comment:17

Fixed with the sagenb spkg from #15016

[jdemeyer]

Merged: sage-5.12.beta1

[vbraun]

comment:19

Followup at #15066

[novoselt]

comment:20

I am not sure if this is the reason, but in a server-worker setup graphics is not shown in 5.12.beta4, while it does in 5.11 + sagenb-0.10.7.1.

[novoselt]

comment:21

Made a ticket #15168 for this.