[Yaml-core] Re: new-retro implicits via regex

[Oren Ben-K. <or...@be...>]

On Monday 13 September 2004 08:13, Clark C. Evans wrote:
> Here is the proposal...
>
>    %REGEX perl-compatible-regex  implied-tag

-1.

> So, if someone wanted the parser to distinguish between 12
> and '12', they could add:
>   %REGEX [-+]?(0|[1-9])[0-9,]*  tag:yaml.org,2002:int
> to their prologue, and bing, all unquoted integers will be
> reported as an integer.

Oy Vey :-) I think this example, by itself, is enough to kill the 
proposition here and now. YAMl is supposed to be human friendly, not 
require them to type REGEXPs into their documents!

> This proposal also gives pretty printers a bunch of flexibility,
> if a pretty-printer discovered that 20 percent of the tags in the
> document were integers, it could add this REGEX and then not
> bother with a !!int tag on those items.

Of course, it would have to very that none of the other 80% didn't match 
the regexp. And since its a pretty-printer and can't know the 
type-by-path rules used in the document, it can't do that. So, next 
thing you'll want is:

	%CONTEXT .../x	tag:yaml.org,2002:float

And then we'll have, in each YAML document, a mandatory embedded 
full-typing schema.

> That's it.  I'm not really for or against beacuse I have not
> thought-through the impliciations.  It's just an idea.

Putting aside the rusty-rezor-blade syntax of regexps, the sheer amount 
of directives it would take to type a simple config file, the fact that 
this replicates 80% of what a proper schema language would do, the 
pressure to move the whole sorry mess to an external document, and so 
on...

This essentially repeats the motivation for XML DTDs. Experience has 
shown that the use of DTDs for anything but validation is horribly 
wrong.

Let's not go there.

Have fun,

	Oren Ben-Kiki