feat: Add new negatedExtglob state to result of scan

[danez]

This allows consumers to see if the glob starts with a negated extglob similar to parse.
At first I wondered if this should maybe set negated to true, but that might be a breaking change and also is then completely different from parse.

One example is jest which uses scan and matcher to mimic micromatchs main api. https://github.com/facebook/jest/blob/d49cb30fb2540efa8c451fb6cb24e8997284bc67/packages/jest-util/src/globsToMatcher.ts#L49-L52

This makes it possible to solve this issue micromatch/micromatch#179 in jest as well as basically support all negated extglobs as non of them work correctly atm.

[coveralls]

Coverage increased (+0.03%) to 92.303% when pulling c1950a5 on extglob-scan into 8839c01 on master.

[jonschlinkert]

Nice work! As you noted this does not catch negated extglobs when they are not the first thing in the glob. Perhaps we should also detect non-starting negated extglobs as well? Or we could change the flag name to negatedStartingExtglob or something like that. Otherwise this might be misleading. Thoughts?

[danez]

Nice work! As you noted this does not catch negated extglobs when they are not the first thing in the glob. Perhaps we should also detect non-starting negated extglobs as well? Or we could change the flag name to negatedStartingExtglob or something like that. Otherwise this might be misleading. Thoughts?

I tried to align the property with the state of parse for consistency and in parse it is also only set if it is at the beginning.
I can rename it to negatedStartingExtglob, but then I would guess it should be also added to parse and deprecate negatedExtglob there?

[jonschlinkert]

I tried to align the property with the state of parse

Good point, and that reminds me that the point of this is feature detection, so it's okay to simply return true if that feature exists. Thanks for discussing, it's helping me remember finer details about the code.

Really nice work! Feel free to merge and publish.

[danez]

I will first try to propose a different change to jest, swapping mm.scan() + mm.matcher() for picomatch(globs, {}, returnState: true). This eliminates scan completely as picomatch can return the state of parse anyway.

[danez]

Okay we went with picomatch directly in jest, but I guess this is still a valuable feature, so merging.