Is it possible to use gscatter in matlab app designer?

Hi! :)

I am watching some videos of matlab on the internet. In the videos a man is applying just x and y axis plots and I think they dont look that interesting .

Does anybody knows how or whether it is possible to use gscatter plots in matlab app designer. I have matlab 2018b release.

Answer 1

[dpb]

Works with R2021b; I do not know at what release the support for uiaxes was introduced; used to be able to get to the doc for earlier releases, now I don't see any way to get there any more...nothing was mentioned in R2018b release notes but that doesn't mean too much for small incremental improvement.

Current doc shows the syntax and explicitly mentions ax or uiaxes object handle; if your release doesn't show that, then it won't work and nothing you can do about it other than either upgrade to a release that does support it or generate the scatter plot by groups a group at a time with regular scatter() plot and grouping variables. One could probably get clever and do it in a one-liner with varfun if your data are in a table or certainly with findgroups and splitapply. "More than one way to skin a cat..."

load carsmall
whos

  Name                Size            Bytes  Class     Attributes

  Acceleration      100x1               800  double              
  Cylinders         100x1               800  double              
  Displacement      100x1               800  double              
  Horsepower        100x1               800  double              
  MPG               100x1               800  double              
  Mfg               100x13             2600  char                
  Model             100x33             6600  char                
  Model_Year        100x1               800  double              
  Origin            100x7              1400  char                
  Weight            100x1               800  double              
  cmdout              1x33               66  char                

[g,id]=findgroups(Model_Year);
hAx=axes; hold(hAx,'on')
splitapply(@(x,y)scatter(x,y,'filled'),Displacement,Horsepower,g)
legend(hAx,"Model Year "+string(1900+id),'location','northwest')
xlabel('Displacement'), ylabel('Horsepower')

I wasn't able to get varfun() to work with scatter() for some reason...not at all clear why it shouldn't work similarly as does splitapply() above.

[Dyuman Joshi]

The documentation of uiaxes mentions it was introduced in R2016a.

"used to be able to get to the doc for earlier releases, now I don't see any way to get there any more"

I think there's a trick to get the documentation of a function corresponding to a release. If I find it, I'll update you here as well.

[dpb]

That's when the uiaxes itself was released; the question here is when the support for uiaxes as the target for gscatter() was introduced (in addition to the default axes) The doc says gscatter (in Stat TB) was introduced prior to R2006a so it predates the introduction of the uiaxes by some 10 releases and so wouldn't have known anything about them initially...but when TMW got around to it is the question.

[Dyuman Joshi]

Okay, got it.

I found a way to get documentation of functions w.r.t a specific MATLAB release -

Current link

https://in.mathworks.com/help/stats/gscatter.html

R2019b link

https://in.mathworks.com/help/releases/R2019b/stats/gscatter.html

Insert /releases/MATLAB_Version/ between /help/ and /Toolbox_Name/function_name

Replace Toolbox_Name with MATLAB for built-in functions.

[Walter Roberson]

R2018a, which was being asked about, did not support passing in an axes.

[dpb]

Actually, OP wrote "I have matlab 2018b release." so could be lucky and it was in between the two.

Do you know of a way to get to the doc for older releases, Walter? I am sure I used to be able to find them, but I can't seem to find any way to do so now; I tried when answering this before...

[Dyuman Joshi]

@dpb, please check this comment - https://in.mathworks.com/matlabcentral/answers/2036886-is-it-possible-to-use-gscatter-in-matlab-app-designer#comment_2937766

[Walter Roberson]

Mathworks only keeps the latest 5 years on public-facing sites. We discussed this with them, and they said that it was taking significant resources to keep updating the old documentation to match changes in their search engine.

They had offered to make older releases available as PDFs, but they have not done that.

Answer 2

[Walter Roberson]

In your release, R2018a, gscatter hard-codes calls to gcf and to newplot without any parameters.

gcf() and newplot() can only "see" figures and axes that have their HandleVisibility set 'on'

uifigures default to HandleVisibility off -- and in your release, the HandleVisibility cannot be turned on for uifigure() so it is not possible to "fool" gcf into looking for uifigures by turning the handle visibility on.

R2019b is the first release that supported passing an axes handle to gscatter

In your R2018b release what you need to do is allow gscatter to plot into a traditional figure with traditional axes, and then copyobj() the contents into your uiaxes and then delete the traditional figure . This will cause the traditional figure to flash into existence and be drawn into before disappearing again with the contents showing up in the uiaxes.

This was not a great situation, but it is what was available 5 years ago.

[dpb]

"This will cause the traditional figure to flash into existence and be drawn into before disappearing again..."

Could one not set 'Visible','off' to not have that side effect, Walter?

Above I presumed one could use scatter() with a uiaxes handle in R2018b; is that not the case, either? This thing of hiding prior release doc on the web site is a real pain...

[Walter Roberson]

Good point about the possibility of turning visbility off.

There are some cases where you must let an axes render before it has correct / consistent information, but I think in this particular case it would not be a problem.

You can use scatter, but it takes more work -- either calling scatter() once for each group or else fudging the legend() entries

gscatter(1:10, rand(10,5), randi(3,1,10))

[dpb]

I'm pretty sure given the limitations I'd prefer the above illustration with @splitapply to do the grouping and plotting directly into the uiaxes over the workaround of plotting to a normal axis and then copyobj(). As the example shows, there's no "fudging of legend entries" required...

That does presume, of course, that scatter and legend had been updated to allow a uiaxes handle by the given release.