when Hash#compare_by_identity is called, it should not copy the keys #1664
Description
It seems like we're all copying keys when inserting a value with a string key. I guess this is to avoid people mutating the string after inserting into a hash, and forgot to rehash the hash.
ruby -ve 'h={};s="";h[s]=s;p s.object_id, h.keys.first.object_id'
ruby 2.1.1p76 (2014-02-24 revision 45161) [x86_64-darwin12.0]
70146287427600
70146287428080
jruby -ve 'h={};s="";h[s]=s;p s.object_id, h.keys.first.object_id'
jruby 1.7.11 (1.9.3p392) 2014-02-24 86339bb on Java HotSpot(TM) 64-Bit Server VM 1.7.0_09-b05 [darwin-x86_64]
2000
2002
rbx -ve 'h={};s="";h[s]=s;p s.object_id, h.keys.first.object_id'
rubinius 2.2.5 (2.1.0 e543ba32 2014-02-08 JI) [x86_64-darwin12.5.0]
360
364
However, MRI would not copy the strings after calling compare_by_identity, or it's defeating the purpose:
ruby -ve 'h={}.compare_by_identity;s="";h[s]=s;h[s]=s;p s.object_id, h.keys.first.object_id, h.size'
ruby 2.1.1p76 (2014-02-24 revision 45161) [x86_64-darwin12.0]
70333940591460
70333940591460
1
Note that MRI is not copying the data, therefore even if we insert twice, it still maps to the same object. Both JRuby and Rubinius are still copying here, defeating the purpose of compare_by_identity.
jruby -ve 'h={}.compare_by_identity;s="";h[s]=s;h[s]=s;p s.object_id, h.keys.first.object_id, h.size'
jruby 1.7.11 (1.9.3p392) 2014-02-24 86339bb on Java HotSpot(TM) 64-Bit Server VM 1.7.0_09-b05 [darwin-x86_64]
2000
2002
2
rbx -ve 'h={}.compare_by_identity;s="";h[s]=s;h[s]=s;p s.object_id, h.keys.first.object_id, h.size'
rubinius 2.2.5 (2.1.0 e543ba32 2014-02-08 JI) [x86_64-darwin12.5.0]
360
364
2
Would also file this for Rubinius.
Done: rubinius/rubinius#3017