Word brackets should be detected with word boundaries

[ndac-todoroki]

Problem

When there are single lined do...end block inside Elixir code, bracket-lens cannot parse brackets correctly.

Examples

In the valid Elixir code below, brackets aren't correctly detected:

What is happening

def zip is judged to be inside def bar(null)'s do block.

Configuration

"[elixir]": {
    "bracketLens.languageConfiguration": {
        "ignoreCase": false,
        "comments": {
            "block": [
                {
                    "opening": "@moduledoc \"\"\"",
                    "closing": "\"\"\""
                },
                {
                    "opening": "@doc \"\"\"",
                    "closing": "\"\"\""
                },
            ],
            "line": [
                "#"
            ]
        },
        "brackets": {
            "symbol": [
                {
                    "opening": "(",
                    "closing": ")",
                    "headerMode": "smart",
                    "inters": []
                },
                {
                    "opening": "[",
                    "closing": "]",
                    "headerMode": "smart",
                    "inters": []
                },
                {
                    "opening": "{",
                    "closing": "}",
                    "headerMode": "smart",
                    "inters": []
                },
            ],
            "word": [
                {
                    "opening": "do",
                    "closing": "end",
                    "headerMode": "smart",
                    "inters": []
                }
                {
                    "opening": "fn",
                    "closing": "end",
                    "headerMode": "smart",
                    "inters": []
                }
            ]
        },
        "strings": {
            "inline": [
                {
                    "opening": "\"",
                    "closing": "\"",
                    "escape": "\\\""
                },
                {
                    "opening": "'",
                    "closing": "'",
                    "escape": "\\'"
                }
            ],
            "multiline": [
                    {
                        "opening": "\"\"\"",
                        "closing": "\"\"\"",
                        "escape": "\\\"\\\"\\\""
                    }
                ]
        }
    }
},

Thoughts

In Elixir, do...end blocks are syntax sugar of [do: term] keyword lists, which could be written without brackets.

if some do
  true
else
  false
end

if some, do: true, else: false

if(some, [do: true, else: false])

if(some, [{:do, true}, {:else, false}])

These are all the same.

If my settings are not wrong, maybe we need a "ignore" option to ignore do: s.

{
    "opening": "do",
    "ignore": ["do:", ":do"],
}

[ndac-todoroki]

I rethought about this... Maybe we can just make the opening closing options to match strictly by the word boundary, like the \b anchor in regexp.

# given settings as `"word": [ { "opening": "do" } ]`
def foo do   # <- matches
  ...
end

def foo, do: ...   # <- does not match, because `do:` is not `do` when regexp is /\bdo\b/

This should not be enabled on symbol brackets, because then this kind of troubles occur:

public static main(int i;    //<- will not match
    return null;
)

But when using word brackets, I don't think anyone (any language) wants dos and ends to work without a word boundary.

[wraith13]

First, thank you for suggestions!

But I'm not sure about this.

This should not be enabled on symbol brackets, because then this kind of troubles occur:

public static main(int i;    //<- will not match
   return null;
)

Isn't it the behavior due to the bracketLens.minBracketScopeLines option?

[ndac-todoroki]

Ah no, I meant that if you force to match symbol brackets by word boundaries too, like /\b\(\b/, then that opening bracket in the example above won't match (which should be detected).

日本語OKだったことに気づいたので日本語で書きます(説明難しかったので) symbol bracketの場合に単語境界を必須にしていると、上の例のような変な改行位置ではあるが有効なコードである場合に正しくパースできなくなってしまう気がしまうので、word bracketだけ単語境界を見るようにしてほしいです、という話でした！