From: Gavin Sinclair <gsinclair@...>
Date: 2003-09-09T13:25:58+09:00
Subject: Re: ANN: MetaTags 1.0

> Hi,
>
> In message "Re: ANN: MetaTags 1.0"
>     on 03/09/09, Hal Fulton <hal9000@hypermetrics.com> writes:
>
> |Much easier than I thought. I'll bet there are some gotchas,
> |though.
> |
> |Matz??
>
> Yes?  It is hard for me to merge this, because
>
>   * it's no better than e.g. "y %{foo: bar}".
>   * it's against my policy of "syntax should be stable".
>
> I agree with usefulness of user-defined literals, and this is much
> disciplined than generic reader macros, but still we need to discuss
> pros and cons.
>
> 							matz.

A thought occurred to me.  If %y{foo: bar} is seen as desirable as a short
and simple way of declaring a YAML literal, what about this instead?

  YAML{foo: bar}

Which is implemented as

  class << YAML
    def {(str)
      # whatever ... "YAML.new(str)??
    end
  end

Notes:

 - no syntax clashes as everyone tries to claim %y{...}
   (I know, this is FUD, but it addresses the known namespace problem)

 - allowing '{' as a method is a nice general change, maintaining a
   unified syntax

 - syntax-wise, Ruby treats YAML{...} as YAML.{%q{...}}

 - this proposal is, if anything, clearer than %y{foo: bar} without
   being burdensome

 - all of the current %_{...} tricks could be implemented in this way
   (not suggesting they should be, rather, pointing out consistency)
   e.g. Regexp{...}, Split{...}, Execute{...}

I make no attempt at pros and cons here, but it seems like a pretty good
idea to me.

Gavin