Hi Kaivlya,
It is my understanding that the performance issue occurs because the code creates a separate plot object for each data point using a loop with individual plot calls. When using ‘datacursormode’, MATLAB attempts to manage a data tip for every single plot object, which can lead to significant slowdowns, high memory usage, and can cause MATLAB to become unresponsive, especially with hundreds of thousands or millions of points.
To improve performance, you can plot all points in a single call instead of looping, store any additional data (such as architecture details) in the ‘UserData’ property of the plot object and customize the data tip using a custom ‘UpdateFcn’ that accesses this data.
Below is an example how you can modify your code:
n_arch = 100000;
archs_cost = rand(1, n_arch) * 100;
archs_util = rand(1, n_arch) * 100;
archs = string(randi(9, n_arch, 5));
% Plot all points at once
h = plot(archs_cost, archs_util, 'b.');
h.UserData = archs;
datacursormode on
dcm = datacursormode(gcf);
set(dcm, 'UpdateFcn', @myupdatefcn);
function txt = myupdatefcn(~, event_obj)
pos = event_obj.Position;
idx = event_obj.DataIndex;
archs = event_obj.Target.UserData;
arch_str = char(join(archs(idx, :), " "));
txt = {['X: ', num2str(pos(1))], ...
['Y: ', num2str(pos(2))], ...
['Arch: ', arch_str]};
end
This approach creates only one plot object, which is much more efficient, the data cursor will display the correct information for each point and should work smoothly even for very large datasets as shown below:
