Fix for ruby 1.8 bug with ActiveSupport::OrderedHash #1359
Conversation
ActiveSupport::OrderedHash did not behave identically to Hash when given a block with a splat.
Fix for ruby 1.8 bug with ActiveSupport::OrderedHash
|
This now makes each_pair different for 1.8 and 1.9 - is this what we want? Anybody using each_pair on ordered hashes is now going to have implement a version check - developers writing apps probably won't be affected that much but what about gem/plugin authors using Active Support? If you've gone to the trouble of using an AS::OrderedHash in a gem it's likely you'll want the same behaviour on 1.8 and 1.9. |
|
@pixeltrix Does each_pair behaves differently on 1.8 and 1.9 for Hash? If so, it is ok for us to be differently here, if not, we should revert this. |
|
Yes it behaves differently - my concern is that a gem/plugin may be using AS:OrderedHash for consistent behavior between 1.8 and 1.9. It at least needs to be noted in the CHANGLOG. :-) |
|
Agreed. |
|
I added an entry to the CHANGELOG on my fork: https://github.com/AndrewRadev/rails/commit/07a353dde1aab7eabf082f4e07cf41f9dc3b2664. I can't tell if this information is enough, though. Should I explain it in a bit more detail? Um, I also don't actually know how to add it to this pull request, or even if it's possible considering it's closed :). Sorry, I'm afraid I'm not very experienced with pull requests. |
|
@AndrewRadev it looks good. However this fix is not going into 3-1-stable. So you just need to put it above. As per the pull request, you have to send a new one and then just link here. Thanks for your work! |
Sorry about that, moved it under 3.2.0. Here's the link to the pull request, hope it's okay now: https://github.com/rails/rails/pull/1378/commits
Glad I could help :). |
Add changelog entry for pull request #1359
…methods, and reducing routes creation by 50% [rails#1359 state:committed] Signed-off-by: David Heinemeier Hansson <david@loudthinking.com>
An example of the problem can be seen in this gist: https://gist.github.com/994919.
Basically,
ActiveSupport::OrderedHashbehaves differently than the built-inHashwhen yielding to a block that accepts its parameters with a splat. This is only an issue on ruby 1.8 because 1.9 uses the built-inHashat all times. It's a very minor inconsistency, but it causes some problems withpp, for example.Tested with rvm-installed ruby 1.8.7-p330, on Mac OS X 10.6.7. At least on this machine, tests are all running green.