Do not use previous value for the slash token

[mrmlnc]

For some patterns the .parse method returns tokens with correct type but wrong value.

// a/b/**/
[
  [ 'bos', '' ],
  [ 'text', 'a' ],
  [ 'slash', '/' ],
  [ 'text', 'b' ],
  [ 'slash', '/' ],
  [ 'globstar', '**' ],
  [ 'slash', '*' ] // <------------- * — before, / — after
]

// **/a
[
  [ 'bos', '' ],
  [ 'globstar', '**' ],
  [ 'slash', '*' ], // <------------ * — before, / — after
  [ 'text', 'a' ]
]

I don't know where I must write tests for it :)

[jonschlinkert]

Thanks! Small nitpicks, would you mind making those changes?

Also, how did you find this bug? I appreciate you looking into it. What was the glob pattern you used for your example?

[jonschlinkert]

Good catch!

Instead of reassigning value, let's move that to the token:

push({ type: 'slash', value: '/', output: '' });

[mrmlnc]

JFYI: I use reassigning value only for local consistency within this module (in most cases we use { value } without property value assigning).

[jonschlinkert]

Thanks for pointing that out, and I appreciate you paying attention to those details!

[mrmlnc]

Also, how did you find this bug?

I'm trying to split pattern into segments. Now these are just experiments:

• Optimization of algorithms for complex patterns mrmlnc/fast-glob#156 (comment)

What was the glob pattern you used for your example?

One of the test patterns in this repository:

• picomatch/test/braces.js

  Line 124 in 606e35f

  assert(!isMatch('a/b/c/xyz.md', 'a/b/**/c{d,e}/**/xyz.md'));

[jonschlinkert]

Impressive! Thank you!

…

Sent from my iPhone

On Dec 24, 2019, at 3:22 AM, Denis Malinochkin ***@***.***> wrote: Also, how did you find this bug? I'm trying to split pattern into segments. Now these are just experiments: mrmlnc/fast-glob#156 (comment) What was the glob pattern you used for your example? One of the test patterns in this repository: https://github.com/micromatch/picomatch/blob/606e35fb2037fc1c000f6528820fb4df28bf2dac/test/braces.js#L124 — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.