YAML module issue

[devnul]

Hello,

I'm having an issue with YAML module. Hopefully someone can help.

Let me be brief:

perl -e 'use YAML;$data{'zip'} = "07102";print YAML::Dump(\%data)' Outputs:

--- #YAML:1.0
zip: 07102

My read of the spec (and experience with the parsers so far) is that a properly functioning parser will treat this as octal and return the value 3650. What I need is:

zip: '07102'

Any ideas?

Devnul

[dragonchild]

• Do you have a failing test? If so, submit it to Ingy.
• Use YAML::Syck

[blazar]

You made me curious, especially since here at work I'm investigating the possibility of using YAML to serialize some data. So I installed YAML::Syck and tried it on the same data as the OP:

$ perl -MYAML::Syck -le 'print Dump { zip => "07102" }'
---
zip: 07102

So the behaviour is just the same. But I also get

$ perl -MData::Dumper -MYAML::Syck -le 'print Dumper Load Dump { zip =
+> "07102" }'
$VAR1 = {
          'zip' => '07102'
        };

which is not what he fears. That only happens if $YAML::Syck::ImplicitTyping is set to a true value (by default, it's false):

$ perl -MData::Dumper -MYAML::Syck -le '$YAML::Syck::ImplicitTyping=1;
print Dumper Load Dump { zip => "07102" }'
$VAR1 = {
          'zip' => 3650
        };

[xdg]

Hmm... how to write a dependency-installation failure test that goes inside the distribution... if I can figure that out, the recursive nature sounds very Ingy-ish.

# t/yaml_install.t
use Test::Install 'YAML';

[devnul]

Thanks for your comments.

The spec seems pretty clear that something like:



zip: 07102

would be implicitly typed as an octal integer. I also tried YAML::Syck, but can't get it to produce a "valid" YAML file that can be read with other parsers which force implicit casting of data types.

- Devnul

[ingy]

Well it's like this. In YAML, every document has a "schema" which determines exactly what type or tag every node in the document has. For two applications to share a document, they need to agree on a "schema".

I quote "schema" because it isn't really a tangible thing like XML schema language. In YAML, there is not yet an official schema language.

A YAML schema can come in one of the following forms:

• Every scalar is a string. This is (or at least should be) the default schema for your general purpose YAML Dumper/Loader.
• YAML implicit typing is honored. (Implicit typing is what turns 07102 into 3650) This is usually not the default, but some processors might choose this.
• The YAML Loader/Dumper chooses a schema of its own. This is in effect what all loader/dumpers should do. Basically it means that it can load whatever it dumped into the same internal structure.
• The application chooses the schema. This is used when every scalar is a string, usually. The program says, "oh the string at $y->{event}{date} should be a date object".
• An external schema document can be used to declare the type of each node. Nobody does this yet.

YAML.pm's strategy is to load all unquoted scalars as strings except for ~ which becomes undef.

If you want to use YAML.pm with other yaml processors you need to agree on a schema, which really means you need to find out what they do and make sure everything is in sync for your needs.

Cheers, Ingy

[devnul]

I agree with everything you've said. Is there some way to make YAML.pm quote all scalars so that I can make it agree with the "schema" of a Loader that doesn't allow much adjustment?

dEvNuL