From: Yehuda Katz Date: 2009-08-23T15:55:34+09:00 Subject: [ruby-core:25050] Re: Proposal: Simpler block format --0016e64bde3250971c0471c9966e Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit On Sat, Aug 22, 2009 at 7:38 PM, Caleb Clausen wrote: > Yehuda Katz wrote: > >> I'd like to propose that we add the following syntax for procs in Ruby: >> >> x = { contents } >> x = { |y| contents } >> { contents }.should raise_error(Exception) >> call_method({ proc_1 }, { proc_2 }, { proc_3 }) >> Essentially, any use of {} that was not a Hash would become a proc. >> Effectively, this includes all use of {} that produces a SyntaxError at the >> moment ({} remains the equivalent of Hash.new). I suggested this to >> > > How do you disambiguate this case: > > { foo bar => baz } > > which could be parsed 2 ways: > { foo(bar) => baz } #a hash > or > { foo(bar=>baz) } #a block > > (Note, however, that the first snippet is currently a syntax error....) > > I think Rick DeNatale raises a good point too. > > Ruby 1.9 used to have a feature like this, but it was taken out, I think > out of fear of the ambiguities it creates. As a user, I like this feature a > lot. But as an implementor of ruby's syntax, I've learned to dread > ambiguities; they just make my life hard. That's not necessarily a reason > not to do it, but I would like the ambiguities to be very carefully > considered and explicitly mapped out. Right. That's why I asked @ujihisa to write the patch before I submitted the proposal -- so we could see just how difficult it would be. I think the patch demonstrates that the implementation isn't too bad, and I definitely would like to see the specification mapped out. Now that we have a working patch, we can see whether it works the way we'd expect by just trying it out :) -- Yehuda Katz Developer | Engine Yard (ph) 718.877.1325 --0016e64bde3250971c0471c9966e Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable

On Sat, Aug 22, 2009 at 7:38 PM, Caleb C= lausen <caleb= @inforadical.net> wrote:

Yehuda Katz wrote:

I'd like to propose that we add the following syntax for procs in Ruby:=

=A0x =3D { contents }
=A0x =3D { |y| contents }
=A0 =A0{ contents }.should raise_error(Exception)
=A0 =A0call_method({ proc_1 }, { proc_2 }, { proc_3 })
=A0Essentially, any use of {} that was not a Hash would become a proc. Eff= ectively, this includes all use of {} that produces a SyntaxError at the mo= ment ({} remains the equivalent of Hash.new). I suggested this to

How do you disambiguate this case:

=A0{ foo bar =3D> baz }

which could be parsed 2 ways:
=A0{ foo(bar) =3D> baz } #a hash
or
=A0{ foo(bar=3D>baz) } =A0 #a block

(Note, however, that the first snippet is currently a syntax error....)

I think Rick DeNatale raises a good point too.

Ruby 1.9 used to have a feature like this, but it was taken out, I think ou= t of fear of the ambiguities it creates. As a user, I like this feature a l= ot. But as an implementor of ruby's syntax, I've learned to dread a= mbiguities; they just make my life hard. That's not necessarily a reaso= n not to do it, but I would like the ambiguities to be very carefully consi= dered and explicitly mapped out.

Right. That's why I asked @ujihisa to write the pat= ch before I submitted the proposal -- so we could see just how difficult it= would be. I think the patch demonstrates that the implementation isn't= too bad, and I definitely would like to see the specification mapped out. = Now that we have a working patch, we can see whether it works the way we= 9;d expect by just trying it out :)

=A0

--
Yehuda Katz
Develop= er | Engine Yard
(ph) 718.877.1325
--0016e64bde3250971c0471c9966e--