add unused_identifier and special_identifier

[ananthakumaran]

This introduces a couple of new issues. Earlier we used to alias
identifier as remote_identifier and function_identifier. I removed
remote_identifier as it was not actually used for any highlighting
(dot_call itself provides enough ways to highlight if
required). function_identifier on the other hand is tricky to remove,
as we use it for def* highlight case.

One way is to introduce a child node (function_identifier (special_identifier))
which keeps the info. Do we need this though?

fixes #19

[akash-akya]

There are some corner cases with unused_identifier

def func(_foo?, _bar!, <<_baz::binary>>) do
  :ok
end

(program [0, 0] - [3, 0]
  (call [0, 0] - [2, 3]
    function: (function_identifier [0, 0] - [0, 3])
    (call [0, 4] - [0, 40]
      function: (function_identifier [0, 4] - [0, 8])
      (arguments [0, 8] - [0, 40]
        (identifier [0, 9] - [0, 14])
        (identifier [0, 16] - [0, 21])
        (binary [0, 23] - [0, 39]
          (binary_op [0, 25] - [0, 37]
            left: (identifier [0, 25] - [0, 29])
            right: (identifier [0, 31] - [0, 37])))))
    (do_block [0, 41] - [2, 3]
      (atom [1, 2] - [1, 5]
        (atom_literal [1, 2] - [1, 5])))))

I think we have to check for unused_identifier at all the places wherever we might return identifier

[akash-akya]

We can remove (#match? @variable.parameter "^[^_]") query clauses from highlight.scm I think.

This introduces a couple of new issues .... function_identifier on the other hand is tricky to remove,
as we use it for def* highlight case.

btw, which issues are you referring to? 🤔
Do we need to remove function_identifier?

[ananthakumaran]

__using__(x)
_another(y)

(program [0, 0] - [3, 0]
  (call [0, 0] - [0, 12]
    function: (function_identifier [0, 0] - [0, 9])
    (arguments [0, 9] - [0, 12]
      (identifier [0, 10] - [0, 11])))
  (call [1, 0] - [1, 11]
    function: (function_identifier [1, 0] - [1, 8])
    (arguments [1, 8] - [1, 11]
      (identifier [1, 9] - [1, 10]))))

<program>
  <call>
    <function_identifier type="function">__using__</function_identifier>

    <arguments>(
      <identifier>x</identifier>
)</arguments>
</call>

  <call>
    <function_identifier type="function">_another</function_identifier>

    <arguments>(
      <identifier>y</identifier>
)</arguments>
</call>
</program>

in the above example, the identifier is represented function_identifier and we lose the info whether it's special or unused

[ananthakumaran]

We can remove (#match? @variable.parameter "^[^_]") query clauses from highlight.scm I think.

removed

[akash-akya]

in the above example, the identifier is represented function_identifier and we lose the info whether it's special or unused

Ok, it's about keeping the AST information. From query perspective we can still use #match? like before right?

(call function: (function_identifier) @special_function
      (#match? @special_function "^__.+__$"))

But if we are thinking of keeping AST consistent then we can use child-node as well like you said.

[ananthakumaran]

Ok, it's about keeping the AST information. From query perspective we can still use #match? like before right?

yes, #match? would still work.

But if we are thinking of keeping AST consistent then we can use child-node as well like you said.

child-node would expose the info, but not that consistent with AST nodes. Anyhow this is probably something we can solve later.

[akash-akya]

Thinking again about child-node, If we add child-node for function_identifier then highlighting might be tricky. Since we don't want to highlight _foo in def _foo(...) as comment right? We might come back to using negation query