Function for plotting large data sets (e.g. time-series with more than 1e6 data points).
The main benefit of this function can be seen in the fact that zooming and paning will work much smoother,
since the data is automatically downsampled to 1e4 points (=> change "n_points" for different number).
The usage is exactly identical to the regular plot-command, i.e.:
HINTS & LIMITATIONS
Will only work in Matlab 7.3 or higher.
"DeleteFcn" & "CreateFcn" are used for plotted line and should not be overwritten.
The "ActionPostCallback" (=> zoom & pan) is also used and cannot be used otherwise in the according plots / figures.
Downsampling is simply performed by plotting only every n-th data point. Therefore, aliasing may occur !!!
Data is clipped outside current axislimits and is updated to the current axis-limits after panning / zooming is finished.
Using double-clicking in "zoom-in" modus might not work as expected, because the dispayed data-set
is clipped. It is possible to return to the original data-set by using the scroll-wheel or the "zoom-out" tool.
The x- & y-data is stored as application-data to the plot-handle.
To get the true x- and y-data for the line with handle h, use ...
x = getappdata(h,'LDS_xdata');
y = getappdata(h,'LDS_ydata');
EXAMPLES (Use zoom & pan; compare to regular plot-command !!!
phi = linspace(0,2*pi,1e6); plotLDS(sin(x),cos(x),'.')
ax(1) = subplot(2,1,1); plotLDS(rand(1e6,1))
ax(2) = subplot(2,1,2); plotLDS(rand(1e6,1))
plotLDS(rand(1e6,1)); ax(1)=gca; figure; plotLDS(rand(1e6,1)); linkaxes(ax,'x'); ax(2)=gca; linkaxes(ax,'x')
Jiro Doke informed me that his submission "DSPLOT", which was chosen as "Pick of the Week", performs a similar task as this submission. I did not notice before and of course he should be acknowledged.
I did a quick comparison on the two submissions and noticed the following differences:
1) plotLDS is not restriced to single figures / axes but will work for arbitrary combinations. Most important: it will also work for linked axes !!!
2) plotLDS is not restricted to plots with monotonically increasing x-data, but will also handle plots with monotonically increasing y-data and plots with unstructured data (e.g. point-clouds).
3) plotLDS uses exactly the same syntax as the regular "plot"-command. This facilitates the use of this command by simply replacing "plot" by "plotLDS" without having to care for input-arguments.
3) Jiro's submission (DSPLOT) includes a more advanced way of investigating the data to make sure it doesn't miss any "outliers".
thanks for this very useful contribution. You say it is completely compatible with the plot Syntax. However, I receive an error message if I use it with the 'XScale' and related properties:
It would be nice if you could implement similar downsampling for logarithmic axes, with equidistant data points on the log scale. In this case, you will probably have to interpolate. E.g., plotLDS(log10(x),y) produces poor results since too many of the small x values are left out with respect to those at large x.
as the documentation says ... : "The usage is identical to the regular plot-command for VECTORS ...". So, if you have matrices, you will just have to break them apart and plot each column separately.
my data has multiple columns (in a matrix), which the regular plot() command or Jiro's dsplot() will draw multiple lines correctly. plotLDS() gave me this error message:
Error in ==> plotLDS at 177
h = plot([min(x) min(x) max(x) max(x)],[min(y) max(y) min(y) max(y)], ...
I'd like to use this feature in subplots, but it seems like Jiro's version doesn't support it.
Just wanted to let you know I also created a similar function:
V1.1 (22.02.08): Works for all plotting-syntaxes, monotonically increasing y-data, unstructured data, and for saving and reloading a figure.
V1.0 (20.02.2008): First release