Linkaxes of Heterogenous Matrixes in Square Control of Matlab? - matlab

Question

I am trying to have the linked control of two matrices which have different xlims and ylims by linkaxes() where there exists an inverse function between the two matrices.
The axes of the matrices (signal D / square matrix D_square) are ax2/ax4, respectively.
The dimensions of the axes are
The limits of ax2 are different than those of ax4, since linkaxes() just makes all input axes have identical limits.
A selection of Fig. ax2 (= zooming an area in the figure) should bring forward appropriate selection in Fig. ax4.
I think you should tell Matlab how to do it.
The only difference between the two datasets is that data of ax2 is the matrix D, while the data of ax4 is the matrix D_square=squareform(D, 'tomatrix') which has an inverse transform, D=squareform(D_square, 'tomatrix') so it is possible to have a control between the two figures.
Code
data=randi(513,513);
D=mat2gray(pdist(data, 'correlation'));
% Set normalized outer position (x,y,width,height)
ax2 = axes('OuterPosition', [0.51 0.5 0.5 0.5]);
plot(D, 'Parent', ax2);
axis(ax2, 'square');
title('Corr pdist');
cbar2 = colorbar();
set(ax2, 'XLim', [0 size(D,2)]);
set(cbar2, 'Visible', 'off')
grid minor;
% Force a draw event to have the axes determine where the
labelconverter = #(x)sprintf('%.2g', x); % http://stackoverflow.com/a/35780915/54964
callback = #(varargin)set(ax2, 'xticklabels', arrayfun(labelconverter, get(ax2, 'xtick'), 'uniform', 0));
set(hFig, 'SizeChangedFcn', callback);
callback(); % necessary for the original small window and its scientific numbering
%% Problem here!
D_square=squareform(D, 'tomatrix');
ax4 = axes('OuterPosition', [0.51 0 0.5 0.5]);
set(ax4, 'XLim', [0 size(D_square,2)]);
image( D_square, 'Parent', ax4 ); % TODO problem here!
set(gca,'YDir','normal');
colormap('parula'); colorbar;
axis(ax4, 'square');
title('Square Corr pdist');
linkaxes([ax2,ax4], 'x');
Output
where Fig. ax4 gets substituted because lims of ax4 are substituted by ax2, vice versa for linkaxes([ax4,ax2], 'x'); in the last line.
Some fixes in Suever's answer
Some changes
Use imagesc() in him=imagesc( D_square, 'Parent', ax4 );
initiate/let things be square by axis(ax4, 'square'); just after imagesc(); not sure why
unknown about if linkaxes([ax2,ax4], 'xy'); or just one axis linkaxes([ax2,ax4], 'x');
Code
data=randi(513,513);
D=mat2gray(pdist(data, 'correlation'));
% Figure out the xrange of your first plot
xrange = [1, numel(D)];
% Set normalized outer position (x,y,width,height)
ax2 = axes('OuterPosition', [0.51 0.5 0.5 0.5]);
plot(D, 'Parent', ax2);
axis(ax2, 'square');
title('Corr pdist');
cbar2 = colorbar();
set(ax2, 'XLim', [0 size(D,2)]);
set(cbar2, 'Visible', 'off')
grid minor;
% Force a draw event to have the axes determine where the
labelconverter = #(x)sprintf('%.2g', x); % http://stackoverflow.com/a/35780915/54964
callback = #(varargin)set(ax2, 'xticklabels', arrayfun(labelconverter, get(ax2, 'xtick'), 'uniform', 0));
set(hFig, 'SizeChangedFcn', callback);
callback(); % necessary for the original small window and its scientific numbering
D_square=squareform(D, 'tomatrix');
ax4 = axes('OuterPosition', [0.51 0 0.5 0.5]);
set(ax4, 'XLim', [0 size(D_square,2)]);
him=imagesc( D_square, 'Parent', ax4 );
axis(ax4, 'square'); % To let things be square!
set(gca,'YDir','normal');
colormap('parula'); colorbar;
title('Square Corr pdist');
% Set XData AND YData (to keep things square)
set(him, 'XData', xrange, 'YData', xrange);
linkaxes([ax2,ax4], 'xy');
where the problem is
to keep things square by set(him, 'XData', xrange, 'YData', xrange);
and to initiate/let them be square by axis(ax4, 'square'); just after imagesc(); code not working as expected having initiation of square shape too far from plotting for some reason.
I think personally that it may not be possible to have strict control (both x,y axes by linkaxes([ax2,ax4], 'xy')) between the two figures.
I think only one axis (by linkaxes([ax2,ax4], 'x') or linkaxes([ax2,ax4], 'y')) can be controlled at least because there exists an inverse function. Some empirical evidence
Output in linkaxes([ax2,ax4], 'xy'); is two empty figures.
Output in linkaxes([ax2,ax4], 'x'); is one figure with some graph and full matrix. Similar with y control.
Active control of square shape
Here active control of square shape by set(him, 'XData', xrange, 'YData', xrange);.
Outputs with keeping square in original view
where one full matrix.
Outputs without keeping square in original view
Disable set(him, 'XData', xrange, 'YData', xrange);
where Fig.ax4 is right. No
Successful Output from Suever's 2nd edit
Original view in x control by linkaxes([ax2,ax4], 'x'); and other changes about labels as described in Suever's answer
where the medium picture is also functional.
How can you tell Matlab the relation between the figures in using linkaxes()?
How can you have a linked control between the two figures in Matlab?

Answer 1

You're having issues because the image in your bottom axes is 513 x 513 (xlims = [0.5 513.5]) whereas the xlimits of your top axes [0 131328]. If you need them to be the same, you can simply alter the XData and YData of your image to be the same as your XLims for your top plot.
% Figure out the xrange of your first plot
xrange = [1, numel(D)];
% Set XData AND YData (to keep things square)
him = image( D_square, 'Parent', ax4 );
set(him, 'XData', xrange, 'YData', xrange);
Now when you call linkaxes (which links XLims) they should move together. You will also need to call axis(ax4, 'tight') to refit the axes to the new image data.
I have included a modified version of your code below that also keeps the xticklabels the same between the two plots.
hFig = figure();
data = randi(513,513);
D = mat2gray(pdist(data, 'correlation'));
D_square = squareform(D, 'tomatrix');
% Set normalized outer position (x,y,width,height)
ax2 = axes('OuterPosition', [0.51 0.5 0.5 0.5]);
plot(D, 'Parent', ax2);
axis(ax2, 'square');
title('Corr pdist');
cbar2 = colorbar();
set(ax2, 'XLim', [0 size(D,2)]);
set(cbar2, 'Visible', 'off')
grid minor;
% Force a draw event to have the axes determine where the
labelconverter = #(x)sprintf('%.2g', x); % http://stackoverflow.com/a/35780915/54964
callback = #(varargin)set(ax2, 'xticklabels', arrayfun(labelconverter, get(ax2, 'xtick'), 'uniform', 0));
set(hFig, 'SizeChangedFcn', callback);
callback(); % necessary for the original small window and its scientific numbering
ax4 = axes('OuterPosition', [0.51 0 0.5 0.5]);
him = imagesc( D_square, 'Parent', ax4 ); % TODO problem here!
% Set the XData and YData of the image
set(him, 'xdata', [1, size(D, 2)], 'ydata', [1, size(D,2)])
set(ax4,'YDir','normal');
colormap('parula');
colorbar;
axis(ax4, 'square');
% Fit the axes to the new XData and YData
axis(ax4, 'tight')
title('Square Corr pdist');
% Link the two together
linkaxes([ax2,ax4], 'x');
% Ensure that the labels ALSO remain the same
linkprop([ax2,ax4], 'XTickLabel')