As of Xcode 12, Apple has matured Swift Bundle Manger to a level the place it is smart so as to add assist for Swift packages to your libraries. There are nonetheless just a few stumbling stones on the trail which haven’t any apparent resolution. So I determine, I’d share with you ways I obtained round them once I lately added SPM assist to DTCoreText, DTFoundation and Kvitto.
Earlier than SwiftPM, my basic strategy for a library can be to have all library code in a `Core` subfolder, with a `Supply` folder containing code which will get compiled and a Sources folder for every kind of sources, like for instance asset catalogs or XIB recordsdata.
A Little bit of Historical past
For the primary 7 iOS variations the product of this product may solely be a static library, Apple solely launched the flexibility to create dynamic frameworks for Goal-C as of iOS 8. With Swift it was the opposite approach round: you could possibly solely have dynamic frameworks with Swift code. For the primary 4 variations of Swift the ABI (Software Binary Interface) was an excessive amount of in flux to permit a statically linked product. With Swift 5, in 2019, we lastly obtained the required stability and thus Xcode gained the flexibility to provide static libraries containing Swift code. That is additionally the principle motive why Xcode all the time added a bunch of dylibs to your apps, containing Swift wrappers to all of the frameworks your app is perhaps interfacing. These dynamic libraries are the third type of libraries we now have encountered thus far.
Oh boy, I keep in mind all of the hackery we needed to do to provide a „faux“ framework that was primarily a fats static library (with slices for all supported processors) and all public headers. We might that in order that anyone utilizing our library may drop it simply into their mission and have all uncovered interfaces be seen. In Goal-C you would want to have the header recordsdata out there for public features and lessons contained within the library. These `.framework` bundles offered a pleasant encapsulation of that, in order that it was virtually like dealing with a single package deal including a third-party framework to your app.
Dynamic frameworks – in actual life, on machine – really don’t comprise any headers any extra as these turn out to be ineffective after compiling. The primary good thing about first-party dynamic frameworks is that Apple can have their APIs and code shared between all apps put in on the machine. The one and solely UIKit framework – put in as a part of iOS – is being accessed by and dynamically linked to all put in iOS apps. Solely a single occasion is current in RAM at any time. Customized frameworks can’t be shared between a number of apps because of all apps being contained in their very own little sandbox. Each iOS app containing DTCoreText for instance has to have its distinctive copy of it inside its app bundle. If an app has quite a lot of third-party frameworks that strategy of loading all frameworks into reminiscence and dynamically linking can noticeably decelerate app launch.
Swift By no means Had Headers
With the improvements introduced with Swift additionally added the idea of modules to Xcode. The Swift Programming Language Web site gives this definition of modules.
A module is a single unit of code distribution—a framework or utility that’s constructed and shipped as a single unit and that may be imported by one other module with Swift’s import key phrase. Every construct goal (similar to an app bundle or framework) in Xcode is handled as a separate module in Swift.
Once you import a module in your code, then Xcode by some means magically is aware of all concerning the public interfaces contained in it, with out ever having to have a separate header file. I don’t understand how precisely that works, however I’m glad that it does!
It was the issue of discovering and integrating third-party libraries into your codebase, that Cocoapods was invented to unravel. The primary public launch of it was virtually precisely 9 years in the past, in September 2011. With the default settings – not utilizing frameworks – Cocoapods would compile the third-party code and merge it with your personal, leading to a single monolithic app binary. And naturally it will handle all these Goal-C headers for you. If you happen to added
use_frameworks! to your Podfile then the technique would change to as a substitute create a framework/module per pod/library. And that may be the requirement for once you had been utilizing exterior libraries written in Swift, or so I assumed …
I’ve all the time used that in apps I’m engaged on which use Cocoapods for dependencies. Think about me rambling on to a consumer of mine concerning the disadvantages of dynamic frameworks, making an attempt to persuade him of the advantages of Swift Bundle Supervisor. Think about my shock after we inspected his app’s bundle, solely to seek out however a single framework in there. All of the third get together code he had ended up fused with the app binary, my library – written in Swift and built-in by way of git submodule and Xcode sub mission – leading to the one dynamic framework in his app.
By default, CocoaPods had been doing all alongside what we all know to be the smarter selection: if third get together code is on the market, to merge the article code it into the app binary. In fact closed-source frameworks that are solely out there as dynamic framework binaries depart you with out this feature. Personally I attempt to keep away from these, just like the satan avoids holy water.
Oh and I additionally would be the first to confess that I may by no means heat myself to Carthage. I’ve by no means checked out it. So far as I perceive, the distinction in strategy versus CocoaPods is that Carthage solely wants a repo URL so as to add a part, whereas CocoaPods wants a Podspec and can generate an Xcode workspace for you the place all dependencies are arrange in a Pods mission. I imagine it is perhaps this workspace wizardry which may put some folks off Cocoapods.
Resourceful Swift Packages
Earlier than the present model 5.3 of SPM the 2 huge remaining ache factors have been the shortage of dealing with of sources and no assist for distributing binaries as packages. These have now been remedied and what’s one of the best half is that Swift packages now have correct integration in Xcode 12.
One other huge benefit that CocoaPods had over different dependency managers was the existence of the “trunk”, a centralised repository of accessible pods. There you could possibly search and discover libraries that may fulfil sure wants of yours. One other necessary side can be that for a model to be launched on the CocoaPods trunk, you would need to “lint” your pod spec which might validate the syntax and guarantee that the library builds with out errors or warnings.
Apple (and the SwiftPM open supply group) have labored on sharpening the software itself. However the central repository with validation side of package deal administration was unfilled. Till Dave Verver stepped and established the Swift Bundle Index. In his personal phrases:
The Swift Bundle Index is a search engine for packages that assist the Swift Bundle Supervisor.
However this website isn’t merely a search software. Selecting the best dependencies is about extra than simply discovering code that does what you want. Are the libraries you’re selecting nicely maintained? How lengthy have they been in improvement? Are they nicely examined? Selecting high-quality packages is tough, and the Swift Bundle Index helps you make higher selections about your dependencies.
Dave launched the SwiftPM Library within the fall of 2019 which in June 2020 obtained re-engineered because the Swift Bundle Index which we use at the moment.
It was this implementation of a central index, focussing on package deal high quality, that pushed me over the sting to lastly begin embracing SPM. With CocoaPods it has been a tedium to arrange a CI server to maintain constructing your libraries for each change to guarantee that nothing breaks. Against this, SPI builds your package deal with Swift variations 4.0, 5.0, 5.1, 5.2, 5.3 for iOS, macOS Intel, macOS ARM, Linux, tvOS and watchOS and can then present on the package deal’s web page the place that labored.
This web page offers a really good overview by which builders can acquire an concept as to the standard of this library. And for us mission house owners it supplies an incentive to attempt to maximise the variety of inexperienced checkmarks you see.
SPI nonetheless tracks 5.3 as “beta” though Xcode 12 has gone gold a month in the past. The reason is that Apple has rushed out Xcode 12 and the finalised assist for constructing common apps that may additionally run on Apple Silicon can be in Xcode 12.2 – out there later this 12 months.
I additionally like how SPI tracks each the most recent secure launch (by way of tag on grasp) in addition to the progress on the develop department. I wanted for these builds to be coming sooner, ideally proper after pushing adjustments to the GitHub repo, however typically it could possibly take a very long time for the builds to be scheduled. Additionally a solution to retry a failed construct can be very good, as we’re used to from Travis-CI or GitLab-CI.
At this level I wished to enter the issues I discovered so removed from including SPM to a few of my libraries, however I’m nonetheless preventing with SPI over a few of these coveted checkmarks. Additionally this text has already turned out longer than I wished it to, that I’ll try this within the subsequent one.
Let me know if that’s of curiosity to you, by dropping me a tweet. Are you contemplating including SPM your self? Which half did you battle with?
Half 2 is right here.
Additionally revealed on Medium.