[Class modifiers] Linter implementation

[pq]

New class modifiers

This issue tracks the linter implementation of the new class modifiers feature.

• implement implicit_reopen #3920
• ensure one_member_abstracts does not lint on sealed classes #3928

Specs

• Sealed types
• Class modifiers

[pq]

The spec says sealed classes are implicitly abstract classes:

A class marked sealed is implicitly an abstract class with all of the existing restrictions and capabilities that implies.

and so we should look at lints that treat abstract classes specially.

One such lint is one_member_abstracts. That said, my initial reaction is that we do not want to extend it to lint one member sealed classes since sealed classes can't just be replaced with functions (as the lint advice suggests).

@munificent?

[srawlins]

The spec says @sealed classes are implicitly abstract classes:

You mean sealed here, right? Not @sealed?

[pq]

Ha! Yes, absolutely. 😬

EDIT: fixed. Thanks!

[munificent]

That said, my initial reaction is that we do not want to extend it to lint one member sealed classes since sealed classes can't just be replaced with functions (as the lint advice suggests).

👍

I wonder if one_member_abstracts still carries its weight. I added that corresponding rule to "Effective Dart" back in the days when many programmers still wrote Java in whatever language they were using and weren't comfortable with closures. These days, it seems like everyone is pretty fluent with lambdas and aren't likely to think they need to write an entire interface just for a single operation any more.

[pq]

I wonder if one_member_abstracts still carries its weight.

I wonder this too. I'm not sure it merits deprecation quite but just shy maybe?

In any event, the initial issue seems settled, we will not upgrade it to be sealed class aware.

Thanks for the feedback!

[bwilkerson]

Actually, we'll need to have a test in order to not produce the diagnostic when the sealed modifier exists. It's not so much a case of making it work with the new syntax as it is a matter of preventing it from working when the new syntax is being used.

[pq]

Closing in favor of individual tracking issues.