Feature #5578: Embedded YAML for Ruby 2.0

• Status: Rejected
• Assignee: -
• Target version: 2.0.0

[trans (Thomas Sawyer)]

Way cool would be support for embedded YAML.

data = ---
a: 1
b: 2
c: 3
...

[kernigh (George Koehler)]

=begin
Current Ruby can do

require 'yaml'

data = YAML.load <<EOF
a: 1
b: 2
c: 3
EOF

p data

Suppose that Ruby adds embedded YAML, and I change my program to

data = ---
a: 1
b: 2
c: 3

p data

Where does YAML end? Is "p data" part of YAML, and not Ruby code? Is it decided by indentation?

=end

[trans (Thomas Sawyer)]

YAML defines ... as end of document marker.

data = ---
a: 1
b: 2
c: 3
... <= this

Yes, it is something like HERE docs, but more concise. I also have this gut feeling that it could evolve into in even more useful ways in the future, though I can't quite put my finger on it yet. Some interesting "hints" though include:

data = eval(source)

Where source could be Ruby or YAML.

load 'foo.yml'

You'd have to use a callback of some sort to get the data, but it's interesting nonetheless that it would work.

And I wonder if a "Ruby code escape" could be worked-in, e.g.

class A; end
foo = "foo"

data = ---
a: ->{ A }
b: ->{ ->{ foo } }
c: ->{ 1 + 1 }
...

data #=> {'a'=>A, 'b'=>#<Proc ...>, 'c'=>2}

You get the idea.

[trans (Thomas Sawyer)]

Hey, btw, how do you get

 section on this thing?

[kstephens (Kurt Stephens)]

Why does YAML need first class status? Define a constructor on the YAML module, and use interpolating heredoc.

require 'pp'
require 'yaml'
def YAML.[]str; YAML.load(str); end

x = YAML[<<"..."]
a: 1
b: 2
c: #{ 1 + 1 }
...

pp x

And it's generic enough for other encodings:

xml = XML[<<"end"]
text
end

etc.

[judofyr (Magnus Holm)]

On Mon, Nov 7, 2011 at 16:30, Kurt Stephens [e-mail] wrote:

Why does YAML need first class status?  Define a constructor on the YAML module, and use interpolating heredoc.

require 'pp'
require 'yaml'
def YAML.[]str; YAML.load(str); end

x = YAML[<<"..."]
a: 1
b: 2
c: #{ 1 + 1 }
...

pp x

Or you can simply define a method called YAML:

require 'yaml'
def YAML(str); YAML.load(str); end

x = YAML <<"..."
a: 1
b: 2
...
p x

On Tue, Nov 08, 2011 at 04:14:02AM +0900, Magnus Holm wrote:

On Mon, Nov 7, 2011 at 16:30, Kurt Stephens [e-mail] wrote:

Why does YAML need first class status?  Define a constructor on the YAML module, and use interpolating heredoc.

require 'pp'
require 'yaml'
def YAML.[]str; YAML.load(str); end

x = YAML[<<"..."]
a: 1
b: 2
c: #{ 1 + 1 }
...

pp x

Or you can simply define a method called YAML:

require 'yaml'
def YAML(str); YAML.load(str); end

x = YAML <<"..."
a: 1
b: 2
...
p x

I will add this.

--
Aaron Patterson
http://tenderlovemaking.com/

[trans (Thomas Sawyer)]

@kurt "Why does YAML need first class status?"

On the surface it's just more elegant. e.g.

x = ---
    a: 1
    b: 2
    ...

vs.

x = YAML %{
      a: 1
      b: 2
    }

But see my last post about the potential for deeper integration. In short, the later can only ever handle a string. Integrating YAML more deeply could allow for Ruby objects to be inserted directly into the YAML structure.

---
time: ->{ Time.now }
...

Would be equivalent too

{
  'time' => Time.now
}

Why would you want to do that? For the very reasons we like YAML. It is an easy to read and concise format, especially when the data structure gets a little bit more nested than just a couple of elements (unlike the examples above).

Note also that JSON is a subset of YAML. So if Ruby supported YAML natively, it would also support JSON natively --just like Javascript.

[drbrain (Eric Hodel)]

=begin
This is valid ruby code:

x = ---
a

This is valid YAML:

p YAML.load "---\na: -> { Time.now }"

=> {"a"=>"-> { Time.now }"}

You seem to be proposing a new thing which collides with valid Ruby code and valid YAML. How do you propose to solve these problems? Can you show a concrete specification?

I think it will be easier to read and write YAML as a String where there is no ambiguity about the rules.

=end

[rkh (Konstantin Haase)]

On Nov 8, 2011, at 15:29 , Thomas Sawyer wrote:

Issue #5578 has been updated by Thomas Sawyer.

@kurt "Why does YAML need first class status?"

On the surface it's just more elegant. e.g.

x

[now (Nikolai Weibull)]

On Tue, Nov 8, 2011 at 19:29, Thomas Sawyer [e-mail] wrote:

@kurt "Why does YAML need first class status?"

But see my last post about the potential for deeper integration. In short, the later can only ever handle a string. Integrating YAML more deeply could allow for Ruby objects to be inserted directly into the YAML structure.

Why is deeper integration necessary? What is it about YAML that’s
solves your problems? Whenever I’ve come across it, it makes me ask
“why aren’t you using Ruby instead?”.

[alexeymuranov (Alexey Muranov)]

I think that when two different languages are used in a same source file, especially a whitespace sensitive one and a non whitespace sensitive, it is good to have them clearly separated.

[angdraug (Dmitry Borodaenko)]

Count me in on the "Nay" side: I strongly agree with Nikolai's and Alexey's points. Ruby as a language is already quote cool, and certainly cool enough for me. Ruby as a platform, on the other hand, has lots of bugs. Can we please do more of fixing bugs and less of adding new cool stuff, at least for a while?

[technohippy (Yasushi ANDO)]

• File prelude.rb.yaml.patch prelude.rb.yaml.patch added
• File parse.y.yaml.patch parse.y.yaml.patch added

Hi Thomas, this is a stunning idea for me! I attached my patches of bad quality to add the yaml literal to ruby1.9.3. I wish they could help you.

[nobu (Nobuyoshi Nakada)]

Does YAML end by "..." in middle of a line?

[trans (Thomas Sawyer)]

Does YAML end by "..." in middle of a line?

No. It only ends when flush left.

[tenderlovemaking (Aaron Patterson)]

• Status changed from Open to Rejected