The dependency on micromatch package should be removed

[prahladyeri]

I've observed that you have a dependency on micromatch npm package which pulls in additional trivial packages like nanomatch, is-odd, is-even and is-number. The only place you're using this module is in the /lib/optimize/SideEffectsFlagPlugin.js which should be trivial to implement in pure JS.

return mm.isMatch(moduleName, flagValue, {
				matchBase: true
});

I hope you'll do it and remove this dependency soon. I also hope you understand that adding dependencies for trivial functionality like this is a worst practice presently plaguing the javascript world and the idea of "don't repeat yourself" is being carried too far with packages like is-odd and is-even.

[montogeek]

That isMatch function is doing more than you think.
It also have a internal cache which is an optimization.

[davidtimovski]

I agree about the isMatch function. I'm waiting to see the justification of is-odd, is-even, and is-number.

[sokra]

Send a PR

[iamarkadyt]

Hmm, should globbing then support only a limited set of wildcards?
Plugin unittest tests with *, **, !, [a-z], {a, ..., n} and [[:lower:]] only.
Otherwise dep use seems quite justified...?

[sokra]

any limitation would be a breaking change. So only try to reduce dependencies without reducing functionality. You could also consider sending a PR to micromatch, or dependencies, to remove is-odd etc.

[jonschlinkert]

I'm the author of micromatch.

• is-odd was removed from the entire micromatch tree months ago.
• is-number checks string values that are parsed from bash-like ranges in brace and bracket patterns in globs. edit: This is not something that typeof val === 'number' will work for. Obviously, there are many ways to check for a number in JavaScript, but given that I'm the author of both libraries I trust them both and know what result should be expected. If you disagree with my choices in the actual code, create an issue and tell me how it should be done differently. It's silly to create issues about something like this as a broad sweeping generalization, like "It uses a dependency" without examining what the actual code does.

You could also consider sending a PR to micromatch, or dependencies, to remove is-odd etc.

This would be fantastic. We're actively working to improve micromatch. As long as all 36k unit tests continue passing and we keep parity with bash, we'll take any PR that improves the performance or reduces code. Reducing deps and leveraging new es syntax is a top priority.

[sokra]

Great, so I guess we can close this one.