Make MapWrapper.Entry's hashCode conform to the contract in java.util.Map.Entry's documentation #6233
Conversation
|
The following test is failing: https://github.com/scala/scala/blob/2.13.x/test/files/run/t5880.scala Which is actually a chi-square test to test that MapWrapper has a decent hash function, and was created as part of fixing https://issues.scala-lang.org/browse/SI-5880. Unfortunately, I think the right thing to do is to delete this test, and reverse the decision made in SI-5880 to "fix" this as this less than optimal hashing behavior for |
|
@SethTisue I can't seem to figure out how to remove the WIP label, but this PR is ready to be reviewed / discussed now. Thanks! |
|
If you squash your commits together, the Jenkins failure will go away. |
|
@hrhino squashed! Does anyone have any objections / feedback for this change? |
|
I couldn't comment on https://issues.scala-lang.org/browse/SI-5880, but I'd like to cc the folks on that ticket:@komamitsu and @axel22 (hope I found the right github accounts) |
|
the old JIRA tickets were migrated to scala/bug, so SI-5880 is here now: scala/bug#5880 |
|
Thanks! |
|
@isnotinvain Thanks for notifying me! |
|
Hi @SethTisue @hrhino, what's the next step for getting this reviewed? Should I also backport it to the the other 2.* branches? Thanks! |
|
Hello, what's the next step here? Thanks, |
we're no longer accepting 2.11.x PRs (unless they are exceptionally important) this looks like a change that could reasonably target 2.12.x. feel free to retarget the PR to 2.12.x as for 2.13.x, normally we merge everything forward from 2.12.x to 2.13.x. but for 2.13.x there is a special situation w/r/t the collections API; in 2.13 the collections code will be replaced entirely by https://github.com/scala/collection-strawman . so if this PR gets merged here, the next step will be to submit an equivalent PR in that repo. @Ichoran do you have time to review the content of this PR...? |
|
Happy to change the base to 2.12.x, maybe I can wait for some review feedback just in case that messes up the diff. |
|
@Ichoran sorry to bug you, have you had a chance to take a look at this? |
|
Sorry to keep bumping this, I just don't want to forget about it. Thanks! |
|
I could use this patch, anything I can do to help? |
|
@isnotinvain - Sorry, I'll get to this over the weekend--probably tonight. (I get so many notifications I didn't notice that I was on listed for this one until Seth told me personally yesterday.) |
|
Thanks @Ichoran ! |
| override def hashCode = byteswap32(k.##) + (byteswap32(v.##) << 16) | ||
|
|
||
| /** | ||
| * It's important that this implementation conform to the contract |
There was a problem hiding this comment.
This shouldn't be a scaladoc comment. It is to implementors. (Users don't have control over what hashCode is used.)
There was a problem hiding this comment.
OK -- what's preferred:
// this kind
// of comment
or
/* This
* kind
*/
There was a problem hiding this comment.
doesn't matter, I don't think we have a consistent style on that. personally, I tend to like // because it makes it easier to see that something definitely isn't a Scaladoc comment.
There was a problem hiding this comment.
OK, will do, thanks
|
Aside from the comment explaining things not working very well as a doc comment, the PR looks good. |
|
OK, I've updated the comments. Should I squash the two commits into 1? Do I need to backport these changes to any other branches? Thanks! |
|
Re-reading @SethTisue 's earlier comment, i can rebase this onto the 2.12 branch, and then send a similar PR to https://github.com/scala/collection-strawman. Should I squash this down to 1 commit or leave it as is? |
….Map.Entry's documentation
|
OK, squashed + rebased to 2.12.x -- assuming the tests etc pass I think this is ready to go. Thanks! |
|
The validate-main check actually passed if you click on "Details" not sure why it hasn't updated in the UI here. |
|
thank you Alex and Rex! |
See scala/bug#10663
I added a regression test, and I'd like to let the CI run the rest of the tests which I think I can do by opening this PR.