Stream.== and Stream.equals are not reflexive

[scabug]

val s = Stream.from(0)
s.equals(s) // does not terminate

But the contract states: "It is reflexive: for any non-null reference value x, x.equals(x) should return true."

See http://docs.oracle.com/javase/7/docs/api/java/lang/Object.html#equals%28java.lang.Object%29

[scabug]

Imported From: https://issues.scala-lang.org/browse/SI-6881?orig=1
Reporter: Kim Stebel (kimstebel)
Affected Versions: 2.10.0-RC5

[scabug]

@paulp said:
I believe that's still reflexive -- if it hasn't yet returned false or thrown an exception, then you can't yet say it didn't return true. But, I think it should be fixed anyway.

[scabug]

Kim Stebel (kimstebel) said (edited on Dec 31, 2012 2:46:54 PM UTC):
It does throw an error when it runs out of heap space.

[scabug]

@paulp said:
That's OK, because that means it returned Nothing and Nothing is a subtype of true.type.

You can't generally exclude "throw an exception" from "Foo should return a value Bar" directives. You can only exclude returning a not-Bar.

[scabug]

Kim Stebel (kimstebel) said:
By that reasoning, an equals method that always throws an exception complies with the contract? I find that hard to believe. It's certainly not what one would expect. If you want I can send a pull request...

[scabug]

@paulp said:
"By that reasoning, an equals method that always throws an exception complies with the contract?"

That's correct. Non-termination is part of life.

It's harder than it looks to fix - see https://groups.google.com/forum/?fromgroups=#!topic/scala-internals/cwEyQCmjX24

Of course it would work if you added it directly to Stream, but it should be fixed everywhere.

[scabug]

Kim Stebel (kimstebel) said:
Oh god, and I thought this would be simple. Adding it directly to Stream might not be such a bad idea. First, for classes other than Stream I think it is at best a performance improvement. Second, I wouldn't know how to implement equals in a way such that it also terminates and returns true for cases like

val s = Stream.from(1)
(0 #:: s) == (0 #:: s)

in the general case. From a cursory reading of your current implementation, I don't think it would pass that test case. But then again, I don't know whether you would want it to. I discovered this issue because I had a function converting lists to infinite streams by appending a stream of zeros. Then I tried to call equals on the resulting streams.

[scabug]

@paulp said (edited on Dec 31, 2012 7:13:11 PM UTC):
My ambition only extends to reference equality short-circuiting all equals calls.

(x eq x) should always be true (and it is desirable that it terminate, even if it is not mandatory)
(x eq null) should always be false unless of course x is null (and this too should terminate)

[scabug]

@paulp said:
I can't pull the right symbols today, but I mean x == x, and so by implication x.equals(x).

[scabug]

@adriaanm said:
Unassigning and rescheduling to M6 as previous deadline was missed.

[scabug]

@Ichoran said:
Here is a PR that would fix it for 2.11.6.

https://github.com/Ichoran/scala/compare/issue/6881

However, it slows down non-identical comparisons by up to about 15%, which I think is unacceptable. Thus, a proper fix should come in 2.12.x where we can introduce a new equality method for Stream. (And then it can catch loops also.)

[scabug]

@Ichoran said:
Seems partially fixed now, though I can't find the commit that did it.

However, the following is still broken:

val s = Stream.from(1)
(0 #:: s) == (0 #:: s)

Not sure whether we want to allow this to work.

[scabug]

@szeiger said:
"equals" is also still broken. Only "==" short-circuits on reference equality.

[scabug]

@szeiger said:
PR for 2.12.x: scala/scala#5259

[scabug]

@SethTisue said:
note that Stefan's PR addresses "equals" too (by overriding sameElements)