fix issue740: parse colon in plain scalar correctly

[dota17]

According to the yaml specification, a colon can be seen as a part of a plain scalar if it is not followed by white space in most cases.

yaml-cpp/src/exp.h

Line 167 in 3f381f1

(RegEx(':') + (BlankOrBreak() | RegEx() | RegEx(",]}", REGEX_OR))) |

In function EndScalarInFlow() , RegEx(':')+RegEx(",]}", REGEX_OR) means if a ":" is followed by a "," or " ] " or " }", the ":" will not be a part of a plain scalar. It is against the spec, so I removed it, and this solved the problem of #740 and #899.

*ex7_3 and *ex7_17 in specexamples.h have a little difference with the spec content, This difference will cause the corresponding testcases can't work as expected because of the new rule. I changed them to be the same as spec.

[perlpunk]

The spec ns-plain-char says:

[129] ns-plain-safe-in ::= ns-char - c-flow-indicator
[130] ns-plain-char(c) ::= ( ns-plain-safe(c) - “:” - “#” )
                         | ( /* An ns-char preceding */ “#” )
                         | ( “:” /* Followed by an ns-plain-safe(c) */ )

That means, I believe, that for example a colon followed by a comma or another flow-indicator will end a plain scalar in flow.
So, as far as I understand, this PR will make this example (7.17) invalid (you added spaces in test/specexamples.h):

{
unquoted : "separate",
http://foo.com,
omitted value:,  # colon + comma
: omitted key,
}

But this is valid (it's an example in the spec after all), and part of the YAML Test Suite:
https://github.com/yaml/yaml-test-suite/blob/master/test/4ABK.tml

[dota17]

Ok, I see. @perlpunk ,thanks for your explaination. I have a misunderstanding about this and confuse the “°” with “·”. I will close this PR.

[perlpunk]

If you have a closer look at the spec example 7.17, you'll notice that it contains

omitted value:°,

where the ° stands for the empty node, so it should not be replaced with a space.

[dota17]

ha-ha, yes. And if one is not careless, it's easy to confuse them, just like me.

[perlpunk]

It would be cool to fix #899 though.
When I ran that recently in yaml-runtimes (generating the https://matrix.yaml.io/) it froze my computer, and I had to hard reset it, so I'm seriously considering taking out yaml-cpp from the test matrix :-/

[dota17]

ok, I will solve this problem as soon as possible

[dota17]

@perlpunk Should processor throw an exception when it parses the [:]? Or, parse it as [~: ~]. I know that pyyaml dose throw an exception.

[perlpunk]

@dota17 it is actually an edge case.
The reference parser and HsYAML and a couple of other libraries parse that the same as [ null : null ], but since it is an edge case and noone should write YAML like that, I think it's not that important what the result is, as long as it does not hang.

Edit: and as an additional note, empty implicit keys will very probably be removed in the next YAML version, so this case will become invalid anyway.

[dota17]

Update the solution and testcases for the bug in issue741/899. The previous solution has been deleted.