Is the Mirah design still the same as the 2009 RubyConf presentation?

[Hamlet D'Arcy]

Hi all,

I watched the 2009 RubyConf presentation that is listed on the home
page. It is about 20 months old now... is the design that Charles
describes in the video still the same? Specifically, he said this:


Phase 1: Parse Ruby
a. Ruby + Type Annotations
b. Parsed by JRubyParser
c. Creates Ruby AST + type nodes

Phase 2: Transform to Duby AST
a. Ruby AST + type nodes
b. Transformer (with plugins)
c. Untyped Duby AST

Phase 3: Infer Types
a. Untyped Duby AST
b. Typer (with plugins)
c. Typed Duby AST

Phase 4: Generate Native Code
a. Typed Duby AST
b. Compilers (with plugins)
c. Platform native code


Thanks,
Hamlet

[Rib Rdb]

This is mostly correct, except that we don't use JRubyParser any more.  MirahParser outputs generic mmeta AST, which is transformed to Mirah AST.

In the newast branch, MirahParser directly generates a new Mirah AST and phase 2 is eliminated.