In YAML, treat a:b (without space) as text, not key-value pair

[sjohannes]

Currently a:b in YAML is parsed as if it's a key:value pair in a mapping, but instead it should be a text scalar.

Example:

colon:in:key: value

Current styling:

{2}colon{9}:{0}in:key: value

Expected styling:

{2}colon:in:key{9}:{0} value

There are two commits in this merge request. The first commit adds a YAML test directory containing some simple test cases for the sake of having that structure there. The second commit is the actual bugfix, including the test case mentioned above. I haven't added an entry in LexillaHistory because I'm not sure if you prefer to write it yourself, but let me know if I should supply it as well.

[zufuliu]

It's likely this will break (rarely used) JSON-like quoted keys:
http://ben-kiki.org/ypaste/data/3051/index.html

'key':value
"key":value

[sjohannes]

Is that really allowed? I've tried a few parsers (PyYAML, JS-YAML, ruamel.yaml, and yaml (npm)) and all of them fail on "key":value, so I think at least in practice it's ok to ignore that particular syntax....

[sjohannes]

Now that I've looked it up on the spec, I don't think your example is valid YAML.

Mappings use a colon and space (“: ”) to mark each key: value pair.
— §2.1. Collections

Note however that in block mappings the value must never be adjacent to the “:”, as this greatly reduces readability and is not required for JSON compatibility (unlike the case in flow mappings).
— §8.2.2. Block Mappings

That kind of syntax is is only allowed within flow syntax (e.g. {"key":value}), which LexYAML doesn't really support correctly anyway.

[nyamatongwe]

Shouldn't the test for isspacechar be for an explicit (space) as isspacechar matches other whitespace (HT, LF, VT, FF, CR):

constexpr bool isspacechar(int ch) noexcept {
    return (ch == ' ') || ((ch >= 0x09) && (ch <= 0x0d));
}

[zufuliu]

https://yaml.org/spec/1.2/spec.html#Characters
5.4. Line Break Characters and 5.5. White Space Characters
at least space, tab and new lines can follows colon :.

[sjohannes]

Yeah, it should just be space, tab, newlines, or nothing (end of file). I've modified the patch to check for

IsWhiteSpaceOrEOL(lineBuffer[i + 1]) || i == lengthLine - 1

where IsWhiteSpaceOrEOL checks for SP/HT/LF/CR, and i == lengthLine - 1 checks for EOF.

[nyamatongwe]

Committed the changes. There were some problems with committing the first patch so the change to .gitattributes was extracted as a separate commit.

[nyamatongwe]

Fix included in 5.1.0 release.