Gentlefolk,
Swift v2 try-catch mechanisms maps onto Objective-C quite well. Except when there is a shadowing problem.
For example:
- (nullable NSDictionary *) dataForKey: (NSString *) key error: (NSError * __nullable __autoreleasing *)error;which maps into Swift v2 as:
func dataForKey(key: String) throws -> [NSObject: AnyObject]The Objective-C method Is also commonly paired with this shortened form:
- (nullable NSDictionary *) dataForKey: (NSString *) key;This short method just calls the main one with a NULL for the NSError**. The signature for the short method is remarkably similar to the Swift throwing method. If I leave the short method in my Objective-C header, then errors ensue:
warning: no calls to throwing functions occur within 'try' expression
do { let data = try self.account.dataForKey(kPrivateKey)
^Obviously, the compiler has chosen the wrong signature. If one comments out the short method, then the app compiles just fine. In code I control, this is obviously not that much of a problem. I comment out the short method and add NULLs to all the places it was called in my legacy code. The problem comes from commercial Frameworks I link against that adopt the short method pattern. There is no way to fix their code.
Or is there a workaround to the above problem? I have tried using the NS_SWIFT_NOTHROWS annotation. It allows me to retain my Swift v1.2 code without change.
I argue this is a compiler problem and should be fixed. In the meantime, I will raise an issue against the commercial frameworks I use. I assume folks would like a radar to the above effect?
