Faster Ruby through Smalltalk

Wed, 15 Nov 2006 10:15:11 +0000

There has been occasional talk over the last couple of years about the performance advantages to be gained by executing Ruby on top of a Smalltalk virtual machine (VM) but nothing's ever come of it.

I just read about Avi Bryant's experiment where he uses JRuby' Ruby parser and visitor architecture to translate a (very) small Ruby program into Smalltalk source that can be executed by Squeak to show that 3+4 does, in fact, equal 7! Earlier posts have a lot of talk of invokedynamic dispatch byte-codes and VM comparisons.

Avi acknowledges that it's a small first step and I'm not sure where the next step would lead unless one were considering an out-and-out Ruby to Smalltalk translator. But it shines a light on what seems to be one of the key roadblocks in front of general experimentation: parsing Ruby source code.

As I understand it the Ruby parser can only be understood by hacking into the rather gnarly parse.y (122K of bisonic joy). There is no canonical specification of Ruby's grammar that enables you to quickly produce a Ruby parser in your own tool or language.

Seems to me that would be a good, if painful, step. Maybe some of that Rails Doc project money could be diverted to get an expert on the case?

Curiouser and Curiouser! on jruby

Faster Ruby through Smalltalk