Re: [Yaml-core] Flow indentation

[Kirill S. <xi...@ga...>]

On Friday 14 April 2006 08:54, Oren Ben-Kiki wrote:
> The simple thing is to allow flow content to be non-indented. This leads 
> to all sort of problems, ranging from readability to breaking all type 
> of heuristics (folding, error bounding, etc.). The worst is:
> 
>     ---
>     foo: "string
>     containing
>     ...
>     ---
>     oops!"
> 
> Today, we _know_ that a ... line ends a document and a --- line starts a 
> new one. Allowing flow content to be non-indented breaks that. YAML was 
> designed to allow efficient streaming, and being able to break a stream 
> into documents without parsing is important to us.

I don't see any problems with this example. '...' still ends a document.
PyYAML raises an error for it:

yaml.scanner.ScannerError: while scanning a quoted scalar
  in "<string>", line 2, column 6:
    foo: "string
         ^
found unexpected document separator
  in "<string>", line 4, column 1:
    ...
    ^

Similar errors are raised for unclosed { or [:

>>> data = """
... sequence: [1,2,3
... mapping: {1: 2, 3: 4}
... """
>>> yaml.load(data)
[...]
yaml.parser.ParserError: while scanning a flow sequence
  in "<string>", line 2, column 11:
    sequence: [1,2,3
              ^
expected ',' or ']', but got ':'
  in "<string>", line 3, column 8:
    mapping: {1: 2, 3: 4}
           ^

> Here is an alternative. Today the rule is that all the flow content must 
> be more-indented than the parent node. Instead, we can change the rule 
> so that all scalar content must be more indented. The { [ , ] } 
> indicators only need to be "as indented" as the parent node. This would 
> mean:

Actually I like this rule and think that both human and automatic
emitters should adhere it. But then I don't see a compelling reason to
enforce this rule in a parser. After all, parsers are expected to be
more permissive than the spec requires.


-- 
xi