Add signature of FileUtils

[ybiquitous]

This change adds signatures for the FileUtils module.

Note: FileUtils has some sub-modules like FileUtils::Verbose, but this change is the first step and does not include such sub-modules.

See also:

• https://ruby-doc.org/stdlib-2.7.2/libdoc/fileutils/rdoc/FileUtils.html
• https://github.com/ruby/fileutils

[pocke]

I guess it also accepts _ToPath and string.

rbs/core/builtin.rbs

Lines 25 to 27 in e7504e6

	interface _ToPath
	def to_path: () -> String
	end

rbs/core/builtin.rbs

Line 64 in e7504e6

type string = String | _ToStr

Because most File methods accept them.

[ybiquitous]

Thanks, it is a nice idea!

Also, it seems good to add the following type to core/builtin.rbs, what do you think?

type path = string | _ToPath

Anyway, I will improve the signatures via new following types: 👍

module FileUtils
  type path = string | _ToPath
  type pathlist = path | Array[path]
end

[ybiquitous]

Fixed via 929ecc8

[pocke]

Also, it seems good to add the following type to core/builtin.rbs, what do you think?

Personally I think the path type alias should not be defined to the top level. path alias is a domain-specific type, unlike string and so on. So I think it should be defined under a namespace, and I think File::path is appropriate.

And we need to reconsider the other type aliases and interfaces, such as _ToPath, _Reader and so on.

So I'd like to keep them in this pull request, and move them under a namespace by another pull request.

[ybiquitous]

That makes sense. 👍

[pocke]

Most (or all?) FileUtils methods are module_function. So let use self? instead of self.

And we need to write aliases for both instance and singleton methods, like the following.

rbs/core/kernel.rbs

Lines 234 to 235 in e7504e6

	alias raise fail
	alias self.raise self.fail

[ybiquitous]

Thanks! I'll improve the signatures and tests! 💪

[ybiquitous]

BTW, it might be nice if we could write it as follows... 🤔

-alias self.chdir self.cd
-alias chdir cd
+alias self?.chdir self?.cd

[ybiquitous]

Fixed via ca1d180

[pocke]

BTW, it might be nice if we could write it as follows... thinking

True. I heard the unexistence of self? alias is intentional by @soutaro because of rare.

But maybe we can add self? alias now. Previously I didn't know there are many singleton-instance aliases like that, so I thought self? is unnecessary. But we found demands for singleton-instance aliases.

[ybiquitous]

we found demands for singleton-instance aliases.

Yes, I think so. We could reduce duplications in a case like FileUtils.

[soutaro]

@ybiquitous Thanks! I don't think adding new syntax for aliasing module functions makes much sense for now. Let me know if you find this is a frequently seen pattern.

[ybiquitous]

OK, thanks for your comment 👍

[pocke]

Great improvement!

I wrote a few comments, but the other parts look good.

[ybiquitous]

@pocke Thank you so much for your advice! I've addressed your comments. 🙆‍♂️

[soutaro]

Thank you @ybiquitous ! 👏