Imported From: https://issues.scala-lang.org/browse/SI-5139?orig=1
Reporter: @rklaehn
Assignee: @rklaehn
Affected Versions: 2.9.1
See #4398
Attachments:

• ConservingMap.scala (created on Nov 4, 2011 6:55:19 PM UTC, 19991 bytes)
• ConservingMapTest.scala (created on Nov 4, 2011 6:55:19 PM UTC, 411 bytes)

@rklaehn said:
Here is a modified HashMap from scala.collections.immutable that fixes the behavior, and a small test.

I'm not sure what to do about the .asInstanceOf[AnyRef]. Since HashMap1 and HashTrieMap are not specialized, it should not matter since the values will be boxed anyway, right?

@rklaehn said:
Map.remove (- operator) also is affected. Fixing this should also be just a one line fix.

scala> val a=Map((1 to 10).map(x=>x->x):_*)
a: scala.collection.immutable.Map[Int,Int] = Map(5 -> 5, 10 -> 10, 1 -> 1, 6 -> 6, 9 -> 9, 2 -> 2, 7 -> 7, 3 -> 3, 8 -> 8, 4 -> 4)

scala> val b=a-10000
b: scala.collection.immutable.Map[Int,Int] = Map(5 -> 5, 10 -> 10, 1 -> 1, 6 -> 6, 9 -> 9, 2 -> 2, 7 -> 7, 3 -> 3, 8 -> 8, 4 -> 4)

scala> a eq b
res36: Boolean = false

scala> a == b
res37: Boolean = true

@dcsobral said:
So, if I assign an AST tree to an existing key in a map, it will traverse all AST just to check whether it is equal or not to the exiting value? That looks extremely suboptimal to me. Or are you just doing reference equality checks? I tried to check what your patch does, but you did not submit a patch -- you submitted the whole edited file. Don't do that, please -- diff was invented 40 years ago, so there's no reason not to use it.

@rklaehn said:
Hi Daniel. I am just doing a reference equality check. The changes are pretty similar to what was done on scala.collection.immutable.Set in response to #4398. Sorry about not submitting a patch.

@rklaehn said:
This was fixed in the process of fixing #5879 (which is not yet completely fixed, but anyway).

So I think this can be closed.

@rklaehn said:
Fixed by rework of updated0 in #5879