[SR-10118] Warn when OptionSet static constants are initialized with a zero value

[beccadax]

Previous ID	SR-10118
Radar	rdar://problem/48689254
Original Reporter	@beccadax
Type	Improvement
Status	Resolved
Resolution	Done

Additional Detail from JIRA

Votes	0
Component/s	Compiler
Labels	Improvement, DiagnosticsQoI, StarterBug
Assignee	Vercantez (JIRA)
Priority	Medium

md5: dd3960163b3287c2517c33c41718adda

Issue Description:

An OptionSet-conforming instance with a value of 0 behaves very strangely; for instance, contains(_:) will always return true for it. Because of this, and because [] is the preferred syntax for an empty option set, ClangImporter removes none members when it imports an NS_OPTIONS enum.

However, if you write an OptionSet-conforming type in Swift, there's nothing stopping you from adding one yourself:

struct MyOptions: OptionSet {
  var rawValue: Int
  static let none = MyOptions(rawValue: 0)
}

It would be nice if the compiler emitted a warning when you did this. The criteria would probably be something like:

• A call to init(rawValue:)

• Passing the literal 0

• In an initial value expression (Initializer, in the compiler's parlance)

• For a static let variable

• In a type conforming to OptionSet

• When the variable belongs to the type it's declared within

[belkadan]

One problem is that there has to be a way to silence the warning. For the importer, that's providing an explicit NS_SWIFT_NAME, but that's less obvious here. Any ideas?

[beccadax]

There are a number of ways you could write zero that we could intentionally not detect, such as (0) or .zero (since the raw type must conform to FixedWidthInteger, which refines AdditiveArithmetic). We could also suggest that you write MyOptions([]), which explicitly uses the syntax we want to encourage.

[belkadan]

Ah, good idea with []. static let none: MyOptions = [] would also work. Okay!

[swift-ci]

Comment by Miguel Salinas (JIRA)

Just checking in. I got started with this after try! Swift SJ a couple weeks ago. Working slowly but steadily. One question I’ve had, however, is why the strange behavior isn’t changed instead of emitting a warning.

[belkadan]

The behavior is "correct" by the definition of OptionSet: any particular set contains all zero of the elements in the none set.

[swift-ci]

Comment by Miguel Salinas (JIRA)

Ah ok would return true for all x. A simple test proved my assumption wrong. Thanks)

[belkadan]

Did you end up putting up a PR for this, Vercantez (JIRA User)?

(I'm checking in on all the StarterBugs that haven't been touched in over a month; it's totally fine if you just haven't had time but still want to keep it.)

[swift-ci]

Comment by Miguel Salinas (JIRA)

Haven't put out a PR yet. Sorry for the late response.

[swift-ci]

Comment by Miguel Salinas (JIRA)

Almost done. I'll put out a PR pretty soon.

[swift-ci]

Comment by Miguel Salinas (JIRA)

PR: #27089

[swift-ci]

Comment by Miguel Salinas (JIRA)

#27089

[swift-ci]

Comment by Robert (JIRA)

Hello, why it doesn't work in case:

(rawValue: 1 << 999999)

?