Add method for extracting licenses from expression

[ajhenry]

This adds a new method ExtractLicenses which pulls all of the licenses out of an expression string (without duplicates)

I added this into the Satisfies file where the rest of the public methods are but if there's a better spot let me know 🤓

[dangoor]

Looks good overall! I just wonder about the possibility of a panic. It's good that there are tests that show this working in a variety of cases.

[dangoor]

Huh, that brings up an interesting question about whether exceptions should be preserved in some fashion. It probably doesn't make a huge difference in what we need this for now, but it is relevant in general to know that a piece of code is licensed with exceptions.

I wonder if it's worth considering the license to be GPL-2.0-or-later WITH Bison-exception-2.2 instead of just the GPL part, but I'm also wary of complicating what we're working on. Probably something to think about in a followup.

[elrayle]

I'm wondering if reconstructedLicenseString() in node.go has a role in this. The current approach appears to use the simple license ignoring LicenseRed, DocumentRef, and Exceptions. That takes a node and returns the license string with license ref, document ref, and exceptions put back in the string.

case

{"WITH modifiers", "GPL-2.0-or-later WITH Bison-exception-2.2", []string{"GPL-2.0-or-later"}},

would become

{"WITH modifiers", "GPL-2.0-or-later WITH Bison-exception-2.2", []string{"GPL-2.0-or-later WITH Bison-exception-2.2"}},

[dangoor]

I think this is a better way to go. It's more accurate to think of the license+exceptions as a different license.

[ajhenry]

Updated it to include modifiers like above

[elrayle]

Since extract isn't used by the satisfies process, it would make more since to me that it lives in extracts.go. I like the new helpers.go. A refactor might pull out other functions (if any) that are used by both satisfies() and extract().

[elrayle]

I'm wondering if reconstructedLicenseString() in node.go has a role in this. The current approach appears to use the simple license ignoring LicenseRed, DocumentRef, and Exceptions. That takes a node and returns the license string with license ref, document ref, and exceptions put back in the string.

case

{"WITH modifiers", "GPL-2.0-or-later WITH Bison-exception-2.2", []string{"GPL-2.0-or-later"}},

would become

{"WITH modifiers", "GPL-2.0-or-later WITH Bison-exception-2.2", []string{"GPL-2.0-or-later WITH Bison-exception-2.2"}},

[dangoor]

Looks great with the changes! Just one suggestion to remove that new nil check.

[dangoor]

Looks good to me!