Well, example 9.5 is pretty explicit in what the spec says. Note that the auto-detect does not return 0, it returns at least 1, but since we start with -1 this results with an overall 0 indentation (since production 207 uses -1 and not 0).

The same logic holds for both of your samples; example 9.3 shows an unindented plain at the top level.

Oren.



On Sat, Apr 13, 2013 at 2:10 PM, KOSEKI Kengo <koseki@gmail.com> wrote:
Hi,

I noticed that LibYAML doesn't parse block scalar without indentation.

  Sample: 1
  ---
  XXX
  YYY
  ZZZ

  Sample: 2
  --- |
  XXX
  YYY
  ZZZ

I expected that these samples would be parsed like this:

  [{"Sample": 1}, "XXX YYY ZZZ"]
  [{"Sample": 2}, "XXX\nYYY\nZZZ\n"]

LibYAML can parse Sample 1 but not Sample 2.

  $ ./run-parser Sample1.yaml
  [1] Parsing 'Sample1.yaml': SUCCESS (11 events)

  $ ./run-parser Sample2.yaml
  [1] Parsing 'Sample2.yaml': FAILURE (10 events)

PyYAML returned error.

  expected '<document start>', but found '<scalar>'


I checked some implementations. The results are:

  Syck (YAML 1.0) - Ruby
    Sample1 ... PASS
    Sample2 ... PASS

  YAML.pm (YAML 1.0) - Perl
    Sample1 ... FAIL
    Sample2 ... PASS

  PyYAML (YAML 1.1) - Python / LibYAML
    Sample1 ... PASS
    Sample2 ... FAIL

  Psych     (YAML 1.1) - Ruby / LibYAML
    Sample1 ... PASS
    Sample2 ... FAIL

  SnakeYAML (YAML 1.1) - Java / based on LibYAML
    Sample1 ... PASS
    Sample2 ... FAIL

  JS-YAML (YAML 1.2) - JavaScript
    Sample1 ... PASS
    Sample2 ... PASS

  Yamerl (YAML 1.2) - Erlang
    Sample1 ... PASS
    Sample2 ... PASS


My questions are:

  (1) Which parsers are implemented properly, at this point?
  (2) Is there possibility that LibYAML support Sample2?
      At YAML 1.1 trunk or future 1.2 branch?


In YAML 1.0 spec, Example 4.21 says:

  --- |
  Usually top level nodes are not indented.

In YAML 1.2 spec, Example 9.5 says:

  %YAML 1.2
  --- |
  %!PS-Adobe-2.0
  ...

  %YAML 1.2
  ---
  !!str "%!PS-Adobe-2.0\n"
  ...

I couldn't find an example like these in the YAML 1.1 spec.
It may be determined by [161] detect(m) being able to return 0 or not.


I found this message.

 * http://osdir.com/ml/text.yaml.general/2008-07/msg00001.html

> The spec requires that continuation lines of a flow scalar are indented
> with at least one space. LibYAML relaxes this rule by allowing
> multi-line flow scalars not to be indented,

So I realized that the Sample1 result was came from relaxing
the spec. But how about the block scalar, Sample 2?


I think this notation is important, because this enable to append
metadata to the plain text.

An email will be represented like this.

  Date: 2013-04-13 20:10:53+09:00
  Subject: Top level block scalar without indentation
  From: koseki@gmail.com
  To: yaml-core@lists.sourceforge.net
  --- |
  Hi,
  I noticed that ...


Using template engine to generate HTML.

  title: YAML block scalar
  template: blog
  published: 2013-04-13
  --- |
  <article>
    <h1>xxxxxx</h1>
    <pre>
      Here must
      be written in
      multi lines.
    </pre>
  </article>


Best regards,

--
koseki

------------------------------------------------------------------------------
Precog is a next-generation analytics platform capable of advanced
analytics on semi-structured data. The platform includes APIs for building
apps and a phenomenal toolset for data science. Developers can use
our toolset for easy data analysis & visualization. Get a free account!
http://www2.precog.com/precogplatform/slashdotnewsletter
_______________________________________________
Yaml-core mailing list
Yaml-core@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/yaml-core