[Scilab-users] ?==?utf-8?q? Axis break with Scilab ?

[Antoine Monmayrant]

Le Mardi, Novembre 22, 2016 17:18 CET, sgougeon at free.fr a écrit: 
 
> Hello Antoine,
> 
> This is an interesting user case. It is quite easy to get an equivalent, even smarter,
> but it depends on what is really required. here,
>  * There is no overlay between the 4 curves. The technical solution would be quite specific.

Well this example is not really representative of a typical use case (it's the only one I found after 5 minutes of searching).
Usually, one have breaks only along one axis: either x or y.
Example along x: you have the evolution of humidity over several years and you want do quickly compare side by side several months of august to see local changes and general trends.
Example along y: you plot energy diagrams for electrons inside a semiconductor and you need to look a the relative curvatures of the conduction and valence bands that are really far apart on each side of the band gap.

>  * The vertical scale is different from one y section to another one.

Well, usually it's not the case.
The key for usability is to treat each subgraphs as parts of the very same graph: you should be able to zoom on the graph in one action.
For example with a vertical break along the y axis: the zoom should give you the same range along x in one action.

>    IMO this is quite embarrassing (not from a technical point of view, but from a reader one. It can easily mislead the reader).

Yes, the example in the link I provided is terrible. However, when used wisely, it can help the reader, for some specific cases.
I don't really fancy axis breaks myself and believe that several subplots would be clearer.
The issue is that in some scientific communities, people are used to axis breaks and explicitly ask them for some figures in articles.
But anyway, with subplots, I haven't found a reliable way to get "zoom synchronization" across several  subplots...

>  * we don't know if the sampling step is constant for each curve, and whether it is the same from one curve to another one.

Sampling might not be constant, but IMO, the scale should be the same.

> 
> In practical, xsetech(), drawaxis(), and/or the fact that data with %nan values are skipped in plots could be used to achieve such a rendering.

I didn't know about xstetech, I am still not sure how to use it. As for %nan, it can be used to hide a certain part of the plot, but not skip the corresponding range.

Thank you anyway for sharing your scilab-fu!

Antoine

> 
> Best regards
> Samuel
> 
> ----- Mail original -----
> De: "Antoine Monmayrant"
> À: users at lists.scilab.org
> Envoyé: Mardi 22 Novembre 2016 13:31:07
> Objet: [Scilab-users] Axis break with Scilab ?
> 
> Hi all,
> 
> I just got a question from one of my colleagues: "Is this possible to get axis breaks when plotting with Scilab?"
> (ie a plot where a bit of the scale in x, y or even both is missing, see : http://www.originlab.com/doc/Tutorials/Multiple-Axis-Breaks ).
> As far as I know, I don't see any direct way to do it.
> 
> Am I wrong?
> 
> As for workaround, I think one could achieve similar results with subplots and visible/hidden axis, but it would require a bit of overhead ...
> 
> Cheers,
> 
> Antoine
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> http://lists.scilab.org/mailman/listinfo/users
>