Making a legend when some of the data sets could be empty - matlab

Question

For my project I have six sets of data to put on a scatter plot, like so:
plot(ax, ay, '.r', bx, by, '.b', cx, cy, '.m', dx, dy, '.c', ex, ey, '.y', fx, fy, '.k');
Sometimes these sets of data will be empty, so bx and by might have nothing in them, therefore getting skipped over.
Is there any way to build a legend that will match the right label to the right color piece of data? In other words, data in the [cx, cy] would always match the label 'c' on the legend next to a magenta dot, even when there is no 'b'. My current legend is as follows:
legend('a', 'b', 'c', 'd', 'e', 'f', -1);
Thanks!

Answer 1

You can get the results you want if you first replace any sets containing empty data with NaN values. For example:
x = [];
y = [];
if isempty(x) || isempty(y)
x = nan;
y = nan;
end
plot(1:10, rand(1,10), 'r', x, y, 'g', 1:10, rand(1,10), 'b');
legend('r', 'g', 'b');
Leaving x and y as empty would give a warning when creating the legend and result in an incorrect legend. When passing empty vectors ([]) to the plot command, it simply doesn't create a line object for that data. When passing a NaN value, it creates but doesn't render a line object, so it still exists and can have a legend entry generated for it.

Answer 2

To complete #gnovice answer with the advantage proposed in #Guto answer, if the order of the lines matters, you can still use NaN, and set their order after it:
x = 1:10;
% plot and get the handle to the lines:
p = plot(x,x,'--g',nan,nan,'r',x,x,'b','linewidth',2);
% set the order of the lines from front to back:
set(p(1).Parent,'Children',p(1).Parent.Children([3 2 1]))
% add the legend:
legend('data 1', 'data 2', 'data 3');
In the above example we bring the green line on top of the blue one:

Answer 3

One other possibility that does not use the x=NaN; is to use a 'dummy plot' out of the boundaries.
The drawback of this process is that you need to select the boundaries manually and if you made any changes in the main plots, you need to manually change the dummy plot as well. Bad idea for use in automated plots. Also If the legend is deleted and called again (either in the insert menu or by legend off; legend show)
The advantage is that you can plot it in a different order from the legend. This can be important when plotting multiple line types and thickness that overlay in some regions. In the figure below, for instance, if you plot the green first, it will disappear in the region x<5 under the blue line.
An example of code:
x_e = [];
y_e = [];
figure()
hold on
plot(100,100, '--g', 100,100, 'r', 100,100, 'b'); %dummy plot
x=1:10;
y=[1 3 3 3 3 4 5 6 9 10]/10;
y2=[1 3 3 3 3 7 6 6 4 3]/10;
plot(x,y2, 'b',x_e, y_e, 'r',x, y, '--g','linewidth',2);
set(gca,'box','on',... %box just to be prettier
'Xlim',[1 10], 'Ylim',[0 1]) % relevant set up!
legend('data 1', 'data 2', 'data 3');
and it give this graph: