Which language is preferable when developing with Typhoon framework, Swift or Objective-C?
I can't answer that question directly, however I can give you the information with which to come to your own decision:
Typhoon is a reflective, dynamic dependency injection container and takes advantage of the full Objective-C runtime. Pure Swift classes - ones that do not derive from NSObject - are Objective-C classes "under the hood" but have the following features:
The Objective-C runtime's reflection methods will not work.
They use C++ style inline, static or vtable dispatch which essentially prevents dynamic runtime augmentation of classes. Swift is a static language not just in regards to its time system, but also in this regard.
Answer:
You can use whichever language you prefer.
But if you're working in Swift you should be aware:
Classes being injected by Typhoon should derive from NSObject.
When it comes to Cocoa or Cocoa-touch you'll be working with UIViewControllers, UIViews and so forth, so we don't see this as a problem in the near future. However, eventually we'd like to provide an alternative version of Typhoon that works for "pure" Swift using clojures or perhaps compile-time instrumentation.
Related
I want to import a Swift framework called "Beethoven" into Objective-C project. I import this framework via cocoapods.
The problem is that framework is written in pure swift. Since the classes are not subclass of NSObject they can't be used directly in my objC classes.
I am newbie in Swift and intermediate in objC, but I think there may be 2 solutions for that:
1-Modifying the whole library: which is probably not an optimal solution.
2-Using a new class which behaves as an interface between swift framework and my objC code. I think second solution will be a better alternative in terms of time and effort needed.
Actually below given post explains modifying the pure swift classes but I don't know how to apply this in my case.
How to use Swift non-NSObject subclass in Objective-C
As I mentioned I am not a very capable programmer and there may be better solutions for that problem. I would be grateful if someone can help me and suggest the best solution for integrating pure Swift framework into my objC project.
NSObject is usually not the only problem. Swift has many features that are not supported by Objective-C. To name a few: swift value types (struct's), swift enums with attached values. If a library uses any of those, it will not be possible to auto-generate an Objective-C header (.h) for it.
Adapting the code of the library on the spot might work, but it is likely more work long term in case if you ever need to update that library again.
Your 2nd approach sounds better: create a layer that is compatible with Objective-C on top of the library which exports methods that need to be exported with #objc and adapts the types. If you go that way consider making a PR contribution to the original library so that everyone could use it in Objective-C projects, and you share responsibility of updating it when the swift code changes.
The 3rd approach would be actually to rewrite some parts of you app to Swift. That might or might not be easier depending on the size of the part of the app that is using the library and how well it is isolated from the rest of the app.
I am building a new framework. The project is to be coded into Swift language, however the clients using this framework have the freedom of using either swift or Objective-C framework.
The question is how do I start. There could be numerous issues like
using structs in swift code but it cannot be made available in
objective C framework.
optionals are missing objective c
Even if I
write different set of files for Swift and Objective C, how will I
map them onto different frameworks under the same project.
Enums with other than Int as rawValue can't be used.
Tuples would not work
I know there have been a few questions around this but none have any satisfactory answer.
PS - Any link to a tutorial or blog would be super helpful too
I did this and got some unexpected results: I have trouble integrating the framework in Swift application. Objective-C works just fine.
You mentioned some of the caveats here.
I suggest doing this iteratively while writing test application in Objective-C which uses all the features. This way if there is some feature that does not cross Swift to Objective-C boundary well, it will be discovered as early as possible.
Your remarks about issues are generally correct with one small exception: optionals are not missing from Objective-C, they appear as nullable/nonnull modifiers on variables and method parameters. Although this does not replace optionals fully, it helps detecting issues early in the process.
Here is a random list of some other issues I discovered:
Bridging between Swift Error and NSError used in Objective-C. The conversion is not always as smooth at it could be, so better use NSError in exported code.
If you mix Objective-C and Swift in your framework, you cannot use bridging header, instead using modulemap files which tend to turn pretty large and complex.
Since you cannot embed frameworks inside a framework, you have to make sure that the application sets ALWAYS_EMBED_SWIFT_STANDARD_LIBRARIES flag for its target. This has to be clearly indicated in the documentation. Also, when creating fat library for your framework, you have to strip these files from the distributed SDK.
And, as I said in the beginning, I still have no success using the resulting mixed language framework in Swift application.
Hope, this will add to your list of things to take into account when developing the library.
Is there like an official workaround with protocols and types that are not supported in Object C
As an example, I have a protocol that returns a reactive cocoa signal producer
public protocol PLoginService {
func login(username:String,password:String) -> SignalProducer<Bool,NSError>
}
If I use #objc on the protocol, then I will get the compiler error that the return type is not supported in Objective C. I will have a lot of constructions like this, so I was wondering what I can do in order to benefit from the Typhoon magic
The current version of Typhoon is a reflective and dynamic dependency injection container, and therefore relies on the full Objective-C run-time. This means that, unfortunately, all protocols that will pass through Typhoon must have the #objc directive. Also, anything that Typhoon injects must be an Objective-C compatible type.
We're currently doing R&D on a pure Swift version of Typhoon, and you'd be most welcome to participate in the development of that project on Github.
Xcode 7 contains an Objective-C variant, where you can define a type-hint for (homogenous) NSArray return values and properties, defined such as NSArray<UIImage*>.
I'd like to use this feature to rewrite my JSON-deserializer class (which needs such kind of type hints – previously I have solved this by adding a -(Class)jsonHintForKey:(NSString*)key to my classes that have homogenous arrays as properties.)
Do you know whether (and if so, how) I can use the Objective-C runtime to get the class of this new type hint at runtime?
The lightweight generics introduced in Xcode 7 are just compile time hints to help the compiler raise warnings, but at run time you get the same old behavior with your variable being just NSArrays of ids.
Source: WWDC '15 "Swift and Objective-C Interoperability" session
See the transcript of the talk:
So the entire lightweight generics feature is based on a type erasure model. Which means that the compiler has all of this rich static type information but it erases that information when generating code.
It is not possible to do that.
Generics were introduced in objective-c to improve the bridge between swift and objective-c. The advantage it gives to objective-c is only useful at compile time, and I that information is lost at runtime.
To provide cross compatibility, Swift allows for the generation of a bridging header so that Objective-C can communicate with Swift classes.
Due to Swift's wonderful namespacing we no longer need to worry about prefixing our Swift files as they are namespaced by their containing framework. A UIView for instance is implicitly namespaces as UIKit.UIView.
Now that Apple are pushing frameworks, I was wondering what the best practices are to avoid header collision when there exists two swift bridging headers with the same symbols.
An example: Say we have two frameworks that have declared a Swift class called Downloader. The Downloader provides the interface: downloadWithURL(url: NSURL)
Generating a bridging header will yield a Downloader-Swift.h file for both of these frameworks. Thus causing a collision. What are the best practices to avoid this?
According to Apple's engineers the <#Module Name#>-Swift.h header uses a macro that mangles the name so as to avoid conflicts (see WWDC video Swift Interoperability In Depth, beginning at 45 min, 40 sec). They gave an example of a Document Swift class:
SWIFT_CLASS("_TtC5MyApp10Document")
#interface Document : UIDocument
// rest of the interface...
To your Swift code, the class will be available as MyApp.Document, if MyApp is the module it is in. So, if you have two Swift classes of the same name coming from different modules – say, one is your own and the other from an open source Swift framework SomeFramework – they will be available to your Swift code as MyApp.Document and SomeFramework.Document...
On the Obj-C side, however, importing these two classes into the same lexical scope leads to Duplicate interface definition for class 'Document' compiler error. That's just Obj-C... In the vast majority of cases, though, this will not be an issue since you can still import these two classes across the app as long as they do not trespass each other's territory. Indeed, how often would you want to use MyApp.Document and SomeFramework.Document in the same module of your app? As we are moving into swifter times, I'm not sure this particular issue warrants a particular strategy, compared to so many urgent issues, such as, multicore, distributed, functional, haptic, anticipative, wearable, autonomous, etc...