Re: More about "extensible conversion framework"

[Charles Lowell]

On Monday, June 25, 2012 11:24:12 AM UTC-5, koops wrote:

In the 0.11 changelog, there's a line about a new value conversion framework. Is this demonstrated or documented anywhere? I'm wondering if the "unboxing" as used in https://github.com/sstephenson/execjs/blob/master/lib/execjs/ruby_racer_runtime.rb may now be redundant.

I have not written any documentation for it yet, but the source can be found here. https://github.com/cowboyd/therubyracer/blob/master/lib/v8/conversion.rb

I wouldn't say that it's redundant in that the standard v8 conversions are not sufficient for what they want which is apparently to nil out functions, but it could be written more elegantly now. i.e.

conversion = @v8_context.conversion

def conversion.to_ruby(v8_object)

  v8_object.is_a?(V8::C::Function) ? nil : super

end

or in a perhaps more module manner:

module NoFunctions

  def to_ruby(v8_object)

    v8_object.is_a?(V8::C::Function) ? nil : super

  end

end

module ForceUtf8

  def to_ruby(v8_object)

    super.tap do |object|

      object.force_endoding('UTF-8') if object.respond_to?(:force_encoding)

    end

  end

end

@v8_context.conversion.extend NoFunctions

@v8_context.conversion.extend ForceUtf8

As far as forcing the string encoding, this is just for show since as of 0.11, therubyracer will always properly encode utf8 strings when fetched from the underlying C++ object: https://github.com/cowboyd/therubyracer/blob/master/ext/v8/string.cc#L22-29

--koops