Clarification about newlines in multi-line strings and empty keys in quotes

[alexcrichton]

I updated some of the language around multi-line basic strings to be clear that they allow bare newlines (I think this is ok, right?).

I also didn't see specific discussion of an empty key in quotes after #283, so I added a commit as well which uses them as an example of bad key names. I could revert this, however, if it's not desired!

[ChristianSi]

I don't really get the point of the first commit: newlines are Unicode character just like any other, so of course they are allowed. Also, the name "multi-line strings" already implies that they can contain newlines.

As for the second commit, I'd think that any string within quotation marks is valid, including the empty string. Certainly not a good idea to use that as part of a key but I don't see a reason why it should be illegal.

[mojombo]

@alexcrichton I'm not sure this change is necessary. I agree with @ChristianSi that "Any Unicode character may be used except those that must be escaped" is quite clear about what's allowed.

As for empty quoted keys, I don't think they should be allowed. I can't think of a single good reason a config file would need a blank key, and allowing them could lead to accidentally incorrect parser behavior and confused consumers of config files. The ABNF I'm working on already disallows it:

key = unquoted-key / quoted-key
unquoted-key = 1*( ALPHA / DIGIT / %x2D / %x5F ) ; A-Z / a-z / 0-9 / - / _
quoted-key = quotation-mark 1*basic-char quotation-mark ; See Basic Strings

I'm planning on including the ABNF in the actual spec soon, and that should be sufficient to clarify this case without requiring additional examples. How's that sound?

[alexcrichton]

Hm, it appears my late-night self isn't parsing that sentence as easily as my early-morning self. I agree with @ChristianSi as well about the first change.

Also, an ABNF sounds fantastic, so I'll just close this in favor of that, thanks @mojombo!

[mojombo]

@alexcrichton Cool. Feel free to check out the full proposed ABNF in #236.

[ChristianSi]

I'm not so sure it was a good idea to close this. If empty quoted keys are forbidden, then the following sentence in the spec is wrong:

Quoted keys follow the exact same rules as basic strings and allow you to use a much broader set of key names.

Since "" is certainly a valid basic string, isn't it?

I'd propose to change to sentence to something like:

Quoted keys follow the same rules as basic strings, except that the empty string ("") is forbidden. They allow you to use a much broader set of key names.

Otherwise there would be a contraction between the ABNF and the written spec which would be a bad thing.