Strange behaviour of plot

When I do the following plot in R2025a, I do not see the line. (Here the code runs under R2024b)

plot([0,eps,1],[0,eps,1])

Answer 1

[Ray]

This is very likely due to a particular rendering algorithm being used to render the line that is more than 1 pixel wide or on a high-DPI machine. This is a known issue, and we are actively working on resolving this.

Currently, there are a couple of workarounds you can try:

1. Set the line's 'LineWidth' property to a small value such as 0.1 to force the line to be 1 pixel wide. For example:

plot([0,eps,1],[0,eps,1],LineWidth=0.1)

2. If the decreased line width isn't acceptable, then you can try setting the 'LineJoin' property to 'chamfer'. (Caveat: this may result in some performance degradation in rendering and interactions)

plot([0,eps,1],[0,eps,1],LineJoin="chamfer")

[Paul]

Is there a link to a Bug Report that discusses in more detail when to expect a problem?

[Mariano]

Thanks. Indeed, the resolution I have is 2880x1920.

[Ray]

Currently, we don't have an existing publicly facing bug report link.

For the plot issue here, we would expect any line plot with a LineWidth > 1 or a line plot on a high-DPI monitor may encounter this issue. The suggested workarounds are as above.

Answer 2

[John D'Errico]

Cute. If you try this in R2025a:

small = 1e-7;plot([0,small,1],[0,small,1])
small = 1e-8;plot([0,small,1],[0,small,1])

(Sorry that Answers is still seemingly broken, not executing code.)

When done using my own copy of R2025a, the first plot shows the expected diagonal line, whereas the second is a blank figure.

Even more interesting, is if I try this:

small = 1e-8;
plot([0 small 1 1+small 2],[0 small 1 1+small 2])

the plot does contain a diagonal line, but it starts at 1, not 0.

Its worth reporting to TMW as a bug.

[John D'Errico]

I've posted this as a bug report, so you do not need to do so also.

[Adam Danz]

Thanks for reporting!

[Walter Roberson]

I do not get a blank figure; I get one that is partly drawn.

[Paul]

@Rui Li,

Can you explain what the issue is for @John D'Errico's and @Walter Roberson's examples? I don't see how they relate to LineWidth or monitor resolution.

[Walter Roberson]

@Paul

The default line width is 0.5, and the default units is "points", which are 1/72 of an inch. But the behaviour is triggered if the line width is more than one physical pixel.

My display happens to be 5120 x 2880, and MATLAB is mapping it to 2560 x 1440.

On my system, the problem shows up when LineWidth of (strictly greater than) 0.375 is used; 0.375 or smaller LineWidth shows up as a single solid line.

0.375 / 0.5 * (5120/2560) = 1.5 pixels. speculation it might be rounding "half towards zero" so getting 1 pixel, whereas for anything greater than 0.375 would result in a value greater than 1.5 which is being rounded to 2 physical pixels or more.

[Paul]

I'm still not getting it.

small = 1e-7;plot([0,small,1],[0,small,1])
small = 1e-8;plot([0,small,1],[0,small,1])

The first one worked and the second didn't. But shouldn't they both be plotting with the same DefaultLineWidth? So either both should work or neither?

And these commands

small = 1e-8;
plot([0 small 1 1+small 2],[0 small 1 1+small 2])

resulted in half of a line? Does that imply that different parts of the line use different LineWidth?

My sense from the start of this thread is that somehow the issue was related to the use of eps (as with the OP) or "small" as in this Answer thread, but I don't see how the values in the input data to plot() can influence the LineWidth.

[Ray]

I think the issue here is due to the precision. For the wide line rendering, the generated line join geometry (see Line - Chart line appearance and behavior - MATLAB: LineJoin for details) might be affected/degenerated under a certain precision. This explains why two values can have different results.