Any idea, why this fails? <https://pl.kotl.in/EW-w...

# announcements

[thana]

Any idea, why this fails?
https://pl.kotl.in/EW-w9FlLK
it says the general contract of Comparable is violated but io cannot spot the violation

[diesieben07]

Your - can overflow. For example Int.MIN_VALUE is clearly smaller than 1, so compareTo must return a negative number. Your method overflows and returns Int.MAX_VALUE, which is positive.

[thana]

ahhh eagle eyed you 😄 thanks 🙂

[diesieben07]

What got me curious what that it didn't happen if I reduced the size of the list. But then it did happen even with the smaller size. So it must only happen for some numbers and then I looked into the edge cases 😄

[thana]

makes totally sense. it's a shame i didn't had this thought although i realized it happens only seldomly with a smaller size 🙈

[Matteo Mirk]

Yes that’s the cause, the subtraction trick only works for small numbers. The weird part is that I don’t have that exception in IntelliJ, but a test run shows the ordering is inconsistent.
Your implementation should be corrected to:

override fun compareTo(other: Data) = compareValues(value, other.value)

[diesieben07]

The weird part is that I don’t have that exception in IntelliJ

[diesieben07]

Run it a few more times... it'll happen (stupid new slack editor, mumble mumble)

[thana]

the exeception is thrown by an implementation of TimSort in the JVm - maybe your runtime does use a different algorithm

[Matteo Mirk]

nope it doesn’t happen, depends on the runtime, probably the one used by Kotlin Playground is different

[diesieben07]

TimSort is in OpenJDK, so I'd be very surprised if you have something else

[diesieben07]

Maybe they changed it in versions after 8 though

[Matteo Mirk]

I’m using AdoptOpenJdk 11, looking at its source code right now and I can see where the exception is thrown, but it’s not happening for me. Maybe because I reduced the list size to 10

[diesieben07]

Yes, that makes the bug much less likely to materialize because you have a lot less elements

[Matteo Mirk]

ok got it

[Matteo Mirk]

found the point: is the array is smaller than 32 it will do a binary sort without merges

[Matteo Mirk]

the exception is raised inside the private mergeLo/mergeHi methods…. oh well it was a nice trip inside the algorithm 🙂

[stephanmg]

just out of curiosity:

[stephanmg]

would this be rather an edge case or corner case? 🙂

[thana]

sure - still it happens 😄

[stephanmg]

what you mean? 🙂

[thana]

edgecases happen

[stephanmg]

oh i see

[stephanmg]

i wanted to see which term would be more appropriate, since one could also say that this is a corner case

[stephanmg]

I’d opt for corner case :D

[diesieben07]

Edge case = within operating parameters, but at their extreme values
corner case = outside operating parameters

so this is an edge case ;)

[stephanmg]

well I think the terminology is a little bit fluent anyway

[stephanmg]

but I can accept this answer

[stephanmg]

corner would be at the contact region of two edges

[stephanmg]

so 🙂

[stephanmg]

I really don’t advocate for this terminology

[stephanmg]

if we talking about 2d