Poll: Supplemental graphical visualizations

[Volker Braun]

Since the last poll about graphics was immediately derailed by unrelated command line questions, let me try once more: Should selected objects display graphical visualizations when it can be done unobtrusively? We all agree that this would be disturbing on the commandline. Mathematica does precisely that in their GUI, e.g. https://reference.wolfram.com/language/ref/Graph.html. It can also easily switched off. Ticket at #18289

[ ] Yes, I want graphical user interfaces that make use of the graphical abilities

[ ] No, I do not want Sage to be a viable alternative to Mathematica (especially for novice users)

[Nils Bruin]

On Monday, May 4, 2015 at 9:54:15 AM UTC-7, Volker Braun wrote:

Since the last poll about graphics was immediately derailed by unrelated command line questions, let me try once more: Should selected objects display graphical visualizations when it can be done unobtrusively? We all agree that this would be disturbing on the commandline. Mathematica does precisely that in their GUI, e.g. https://reference.wolfram.com/language/ref/Graph.html. It can also easily switched off. Ticket at #18289

[x] Yes, I want graphical user interfaces that make use of the graphical abilities

In particular, in a graphics-capable environment, I'd be very unpleasantly surprised if

sage: plot(sin(x),x,-5,5)

does not result in a nice plot appearing. I'd be quite happy if

sage: G =  plot(sin(x),x,-5,5)

does not produce any output. A corollary would probably be that

sage: G

afterwards would display a plot. I'd be happy with that too, although at that point I wouldn't mind having to type G.view() or G.show() either.

Whether any object not asked for by plot(...) or <object>.plot(...) gets displayed graphically, I'd be hesitant about, but I imagine there are other objects of essentially graphical nature that should by default be displayed graphically whenever possible.

[Volker Braun]

Just to clarify, we are talking about the following:

[Andrey Novoseltsev]

[X ] Yes, I want graphical user interfaces that make use of the graphical abilities

[Nils Bruin]

On Monday, May 4, 2015 at 10:48:50 AM UTC-7, Volker Braun wrote:

Just to clarify, we are talking about the following:  <graph picture>

OK, I'm neutral on that (so delete my vote from the count).  I wouldn't normally think that a graph comes with a representation in the plane so I'd be a little surprised that it does, but since it's so easy to avoid to get the plot (and any other representation of a graph is rather uninstructive) I wouldn't particularly mind.

[Nathann Cohen]

Helloooooooooo,

 

and any other representation of a graph is rather uninstructive

Err... Well. Try "graphs.RandomGNP(30,.5).show()" to get an idea of how uninstructive a plot can be (compared to anything you might say on its, diameter, degree distribution, planarity, etc...) :-P

Just to complete Volker's question: it is not only about graphs, but in general about objects whose graphical representations "adds something" (there is a Matrix.plot, but that's not the subject).

 The ticket at #18289 does the same for Posets, for Polyhedra, and more will follow. His question is: "should it be the default behaviour, or only something that can be enabled manually if desired?"

Nathann

[John H Palmieri]

[X] Yes, I want graphical user interfaces that make use of the graphical abilities

On Monday, May 4, 2015 at 11:32:53 AM UTC-7, Nathann Cohen wrote:

Helloooooooooo,

 

and any other representation of a graph is rather uninstructive

Err... Well. Try "graphs.RandomGNP(30,.5).show()" to get an idea of how uninstructive a plot can be (compared to anything you might say on its, diameter, degree distribution, planarity, etc...) :-P

I think that enough objects will look good that it will outweigh bad examples like this one, especially for beginning users.
 

Just to complete Volker's question: it is not only about graphs, but in general about objects whose graphical representations "adds something" (there is a Matrix.plot, but that's not the subject).

 The ticket at #18289 does the same for Posets, for Polyhedra, and more will follow. His question is: "should it be the default behaviour, or only something that can be enabled manually if desired?"

I think it's important that the method for changing the behavior is clearly documented, in any case.

--
John

[Volker Braun]

On Monday, May 4, 2015 at 8:32:53 PM UTC+2, Nathann Cohen wrote:

Just to complete Volker's question: it is not only about graphs, but in general about objects whose graphical representations "adds something" (there is a Matrix.plot, but that's not the subject).

In fact, that is relevant. The matrix plot does not clarify at all what a matrix is. The ticket doesn't propose to use matrix plot, in fact, I'd be against adding it as a supplemental visualization. It surely is useful in certain situations, but it is just not helpful.

The important feature is that the code lets you add visualizations carefully. Each object can control when to do it and when not to do it.

[Volker Braun]

On Monday, May 4, 2015 at 8:40:24 PM UTC+2, John H Palmieri wrote:

Try "graphs.RandomGNP(30,.5).show()" to get an idea of how uninstructive a plot can be (compared to anything you might say on its, diameter, degree distribution, planarity, etc...) :-P

I think that enough objects will look good that it will outweigh bad examples like this one, especially for beginning users

Just to clarify, graphs.RandomGNP(30,.5) would not be plotted. The code only shows a supplemental plots if the number of vertices is < 20.

[Jan Keitel]

[X] Yes, I want graphical user interfaces that make use of the graphical abilities

[Ralf Stephan]

[X] Yes, I want graphical user interfaces that make use of the graphical abilities

Sounds like good design. The command that turns it off can then put into ~/.sage/init.sage 

[Eric Gourgoulhon]

[X ] Yes, I want graphical user interfaces that make use of the graphical abilities

[Nicolas M. Thiery]

In principle, I like the feature; I won't really use it myself being a
terminal guy, but my students will certainly benefit from it.

The only caveat I could see is if this would noticeably slow down the
output [1]. If that's not the case, +1.

Do you mind updating the ticket description with a short description
of the introduced protocol?

Cheers,
Nicolas

[1] In an ideal world, this feature would be multi-threaded, and
appear lazily when ready; I don't know whether the IPython interface
would support that though ...

--
Nicolas M. Thiéry "Isil" <nth...@users.sf.net>
http://Nicolas.Thiery.name/

[Volker Braun]

On Tuesday, May 5, 2015 at 10:53:04 AM UTC+2, Nicolas M. Thiéry wrote:

The only caveat I could see is if this would noticeably slow down the
output [1]. If that's not the case, +1.

The picture is only generated if the graph has < 20 vertices precisely to avoid long delays.

 

Do you mind updating the ticket description with a short description
of the introduced protocol? 

I'll try to write a section for the dev guide in the next couple of days.... 

[Franco Saliola]

[X] Yes, I want graphical user interfaces that make use of the graphical abilities and I want Sage to be superior to Mathematica (especially for expert users)

Franco

[Travis Scrimshaw]

[X] Yes, I want graphical user interfaces that make use of the graphical abilities

Despite the obvious bias in the question, I'm for it as the default in the notebook.

Best,
Travis