Fix highlight for atoms containing @ character

[w-sanches]

Fixes #510

In eb06df6 I used a regexp that was matching too many things. I completely forgot that spaces are also printable characters 🤦‍♂

This PR comes with a new version using [^\d0-\d127] to match those special characters. This will:

• Not match spaces, so atoms will be matched correctly;
• Match accented characters like á ó é;
• Match least common characters like ಠ which are valid in an atom like :ಠ‿ಠ;
• Match some invalid characters;

Elixir actually accepts any UTF-8 letter as atoms and Vim does not support Unicode modifiers AFAIK, so using something like \p{L} is not possible 😞
I went with the route of adding any non-ASCII on top of words and @s as it gave us more positive results and the false-positives will raise errors when compiled, so they are easily fixable. I still can't decide by myself if we should add those false-positives or not.
WDYT?

The other option would be to remove support for accents and other UTF-8 characters and only highlight those atoms with \w and @ matched chars.

@nifoc @jbodah please take at look at this. Any inputs are welcome.

[sodapopcan]

Is the only thing stopping this from getting merged the conflicts, or is there something wrong with the commit? I'd be happy to fix the conflicts.

[ghost]

Vim does not support Unicode modifiers AFAIK

I think [[:lower:][:upper:]] should work with Unicode.

:ಠ‿ಠ

I got this error:

** (SyntaxError) iex:1:2: unexpected token: "‿" (column 2, code point U+203F)
    |
  1 | :ಠ‿ಠ
    |  ^
    (iex 1.15.6) lib/iex/evaluator.ex:294: IEx.Evaluator.parse_eval_inspect/4
    (iex 1.15.6) lib/iex/evaluator.ex:187: IEx.Evaluator.loop/1
    (iex 1.15.6) lib/iex/evaluator.ex:32: IEx.Evaluator.init/5
    (stdlib 5.0.2) proc_lib.erl:241: :proc_lib.init_p_do_apply/3

Elixir's syntax reference for atoms mentions that unquoted atoms should start with unicode letters or an underscore and then it supports numbers and at symbols (@).

I think the following regex can match unquoted atoms: :[_[:lower:][:upper:]][_[:lower:][:upper:][:digit:]@]*

I would use a syn-region for quoted atoms, although, I would need to think a bit more about this approach.

[ghost]

I would defined it as:

syn region qatoms start=/:"/ skip=/\\"/ end=/"/
syn region qatoms start=/:'/ skip=/\\'/ end=/'/
syn match  uatoms /:[_[:lower:][:upper:]][_[:lower:][:upper:][:digit:]@]*[?!]\?/

syn cluster atoms contains=qatoms,uatoms

hi link qatoms identifier
hi link uatoms identifier

This is just an idea, I might send a patch later 🙂 .

EDIT: Here's a screenshot with some highlights, it misses the :++ but it can be extended once the operators are defined: