Dirk Gadsden

@delner et al: Please note that this is very much a work-in-progress that stems from some fun-seeking vacation explorations into Ruby C tracing. It'd be cool to pull this off,...

@delner: Sounds good! I'll dive back in this weekend, so y'all should have something much more polished to start reviewing and discussing come Monday.

@pawelchcki: Thank you for the comments and questions! I'll get to them when I get home tonight. 😄

Fooled around with this idea some (a lot) tonight. To really make any major changes to underlying class semantics on Rubinius will probably require a _ton_ of work and messing...

Please see [this comment](https://github.com/rails/jbuilder/issues/289#issuecomment-146000448) on the proposal pull request for performance results.

I will be implementing this proposal in [Everlane's fork](https://github.com/Everlane/jbuilder).

@vincentwo, @rwz, @dhh: This has been implemented in PR #298.

@vincentwoo: I'll put a side-by-side together today.

Here are the results from benchmarking with [this script](https://gist.github.com/dirk/4048df31641f5dab1c67): ``` $ ruby -v ruby 2.2.1p85 (2015-02-26 revision 49769) [x86_64-darwin14] $ git checkout master $ ruby benchmark.rb Calculating ------------------------------------- basic 3.000...

I ran into this problem last night as well. Thank you for doing some investigating and getting the conversation started, @jfirebaugh! I'm wondering if there's a flag that can be...