inpolygon, different results from MATLAB R2022a and R2023a?

aoi1_x = [0, 960]; 
aoi1_y = [0, 1080]; 
inpolygon(2, 2, aoi1_x, aoi1_y)

ans = logical
   0

R2022a returns 1, but R2023a returns 0.

I would like to know why these results differ.

[Matt J]

Undoubtedly, because of a bug that has now been fixed.

[Dyuman Joshi]

Even R2021a Update 8 gives output as 1.

I couldn't find any Release notes or Bug fixes report on this.

[Matt J]

Well, even if it's not in the Release Notes, we should be glad that it has been fixed, right?

[DGM]

I can't find it in bug reports or release notes, but this behavior appears somewhere between R2009b and R2015b, so it should be present over a wide range of versions. There were other bugs from R2016b which were addressed in R2022b, but it's hard to say if they were related.

Bug reports would be a lot more useful if you could trust that known bugs would actually show up there.

[Dyuman Joshi]

I disagree. It's frustrating to know that a well known* bug is not even acknowleged by TMW, nor is there any acknowlegment of the bug being fixed for the later versions.

And the fact that it is still not fixed for the versions mentioned, just adds onto the frustration.

* - I say well known because it appears across multiple versions.

[Adam Danz]

To confirm what others have mentioned above, this behavior was updated in R2022b.

A bug report was not published.

For FAQs on bug reports, see https://www.mathworks.com/support/bug_reports/faq.html

Answer 1

[Cris LaPierre]

This is not an official MathWorks answer.

My observation is the following. You used to be able to define a rectangular ROI by specifying just 2 diagonal vertices. However, this behavior was undocumented. All doc examples and wording indicate you should specify all vertices. Apparently around R2022b, changes were made to the function that no longer allow you to specify a rectangular ROI with just 2 points.

This example highlights the risk of using undocumented behavior - it can change at any time without notice.

[Junichiro Kawahara]

Great. Your observation makes sense.

Thank you,

[Matt J]

@Junichiro Kawahara Please Accept-click Cris' answer, seeing as you are happy with it.

[DGM]

Ah. I didn't even expect that to be a thing. That kind of makes the other bug report make more sense now.