[:call, [:lit, 1], :+, [:array, [:lit, 1]]]
A literal looks like this:
[:lit, 42]
To traverse the AST, ParseTree comes with the SexpProcessor library, which facilitates the creation of visitors. To analyze all node types of a Ruby AST, a subclass of SexpProcessor with process_XXX methods is created, where XXX is the name of the node. For instance, this handles the
:alias
node:def process_alias(node)
cur = node.shift
nw = node.shift
# ...
end
The Ruby to Rubinius bytecode compiler is built in this way. For instance, a Ruby
alias
call is parsed into [:alias, :old_name, :new_name]
, which the compiler handles as such:def process_alias(x)
cur = x.shift
nw = x.shift
add "push :#{cur}"
add "push :#{nw}"
add "push self"
add "send alias_method 2"
end
The compiler takes the old name (in
curr
) and the new name (in nw
), and creates the bytecode instructions (as strings) necessary to implement the functionality, which are then turned into the binary bytecodes executed by the Rubinius interpreter.Having the compiler in Ruby makes it easy to get insight into the inner workings and modify it for experiments. Useful scenarios could include instrumentation of the generated code or a low overhead way of collecting statistics about the compiled code.
To look at the Rubinius source code, either refer to InfoQ's article about getting started with Rubinius development or just take a peek at the Rubinius source code online, for instance the current version of the Rubinius bytecode compiler.
The compiler is not the only aspect necessary for Rubinius. A complete standard library is necessary too. Marcus Crafter, of Red Artisan, provides a tutorial on how to add library functionality to Rubinius. The tutorial shows to use the Rubinius foreign function interface (ffi) to access native library calls. This is used to implement some missing library functionality, in this tutorial, the POSIX call
link
.