The set pattern for single range char is not working in glob

[tomlau10]

How are you using the lua-language-server?

Visual Studio Code Extension (sumneko.lua)

Which OS are you using?

Windows

What is the issue affecting?

Diagnostics/Syntax Checking

Expected Behaviour

I am testing the glob like pattern used in Lua.doc.<scope>Name, and the set pattern seems not working for single character.

• .luarc.jsonc

{
    "doc.privateName": [
        "_[aeiouA-Z]"   // underscore followed by any upper letter, or lower letters a|e|i|o|u
    ]
}

• test.lua

---@class A
local A = {}
A._A = 1
A._Z = 1
A._a = 1
A._e = 1
A._z = 1
A._AA = 1
A._aa = 1

---@type A
local t = {}
print(t._A)     -- warning
print(t._Z)     -- warning
print(t._a)     -- warning
print(t._e)     -- warning
print(t._z)     -- ok
print(t._AA)    -- ok
print(t._aa)    -- ok

Actual Behaviour

---@type A
local t = {}
print(t._A)     -- warning
print(t._Z)     -- warning
print(t._a)     -- ok (false negative)
print(t._e)     -- ok (false negative)
print(t._z)     -- ok
print(t._AA)    -- ok
print(t._aa)    -- ok

• the single range word aeiou is not working as expected

Reproduction steps

Use the provided snippet

Additional Notes

I know that the glob pattern syntax in defined using LPeg:

lua-language-server/script/glob/glob.lua

Lines 19 to 46 in ddc96bd

	local parser = m.P {
	'Main',
	['Sp'] = m.S(' \t')^0,
	['Slash'] = m.P('/')^1,
	['Main'] = m.Ct(m.V'Sp' * m.P'{' * m.V'Pattern' * (',' * expect(m.V'Pattern', 'Miss exp after ","'))^0 * m.P'}')
	+ m.Ct(m.V'Pattern')
	+ m.T'Main Failed'
	,
	['Pattern'] = m.Ct(m.V'Sp' * prop('neg', m.P'!') * expect(m.V'Unit', 'Miss exp after "!"'))
	+ m.Ct(m.V'Unit')
	,
	['NeedRoot'] = prop('root', (m.P'.' * m.V'Slash' + m.V'Slash')),
	['Unit'] = m.V'Sp' * m.V'NeedRoot'^-1 * expect(m.V'Exp', 'Miss exp') * m.V'Sp',
	['Exp'] = m.V'Sp' * (m.V'FSymbol' + object('/', m.V'Slash') + m.V'Word')^0 * m.V'Sp',
	['Word'] = object('word', m.Ct((m.V'CSymbol' + m.V'Char' - m.V'FSymbol')^1)),
	['CSymbol'] = object('*', m.P'*')
	+ object('?', m.P'?')
	+ object('[]', m.V'Range')
	,
	['SimpleChar'] = m.P(1) - m.S',{}[]*?/',
	['EscChar'] = m.P'\\' / '' * m.P(1),
	['Char'] = object('char', m.Cs((m.V'EscChar' + m.V'SimpleChar')^1)),
	['FSymbol'] = object('**', m.P'**'),
	['RangeWord'] = 1 - m.P']',
	['Range'] = m.P'[' * m.Ct(m.V'RangeUnit'^0) * m.P']'^-1,
	['RangeUnit'] = m.Ct(m.C(m.V'RangeWord') * m.P'-' * m.C(m.V'RangeWord'))
	+ m.V'RangeWord',
	}

I am not familiar with LPeg, but by adding a print(#range, range[1], range[2]) inside mt:range() here, those single range word seems don't even get parsed. 😕

lua-language-server/script/glob/matcher.lua

Lines 99 to 100 in ddc96bd

	for _, range in ipairs(exp.value) do
	if #range == 1 then

With a bit of testing, the RangeUnit definition seems should be changed from:

    ['RangeUnit']   = m.Ct(m.C(m.V'RangeWord') * m.P'-' * m.C(m.V'RangeWord'))
                    + m.V'RangeWord',

to =>

    ['RangeUnit']   = m.Ct(m.C(m.V'RangeWord') * m.P'-' * m.C(m.V'RangeWord'))
                    + m.Ct(m.C(m.V'RangeWord')),

Then the set logic for single range character starts to work 🎉
But I don't know why it works this way 🙈

---

Can anyone comment on my above suggested change?
If this is correct, I am going to open a PR. 🙂

Log File

No response

[tomlau10]

More issues found 😕

{
    "doc.privateName": [
        "_*[a-zA-Z0-9]"     // prefix with at least 1 _, followed by anything, then end with a char
    ]
}

• this is expected to match _id / _id_1 / __id but no, it gives false negative

---@class A
local A = {}
A._a = 0
A._a1 = 0
A.__a = 0
A.__a1 = 0

---@type A
local t = {}
print(t._a)     -- warning
print(t._a1)    -- ok (false negative)
print(t.__a)    -- warning
print(t.__a1)   -- ok (false negative)

---

By testing the glob pattern here, all of the above should be matched:
https://www.digitalocean.com/community/tools/glob?comments=true&glob=_%2A%5Ba-zA-Z0-9%5D&matches=false&tests=_a&tests=_a1&tests=__a&tests=__a1

[sumneko]

Perhaps I should go and find a third-party library.