I'm currently writing some swift libraries to be included in an App that uses CocoaLumberjack to log.
So initially I've added CocoaLumberjack as a dependency to all of them and it works quite well.
Then I've seen this ticket where they say, that you should not add it as a dependency, but use if it is there.
Despite that I've already seen some projects on GitHub where they do exactly that in Objective-C, I haven't seen it yet in Swift.
Can somebody point me to a sample project or help me to find the right direction to take
THX
Your should add CocoaLumberjack/Swift as a dependency if your library uses it as a logger.
But your library code should not add any loggers (DDTTYLogger, DDFileLogger, etc.) to avoid log duplication.
Adding loggers should be done in final application that uses your library.
For library itself it could be test bundle with tests:
class YourKitTests: XCTestCase {
override func setUp() {
super.setUp()
DDLog.add(DDTTYLogger.sharedInstance(), with: .verbose)
}
}
Related
Pretty new to creating frameworks with SPM dependencies. So I made a new framework project, added some of my classes/files as well as a SPM dependency (CocoaLumberjack logger). Framework compiles fine.
When I look for my framework product that I'm planning on embedding into some other project I see that it is in my Products folder. Alongside with it I see CocoaLumberjack module. Inside of my framework there is not much beside the exec file.
When I try to embed my framework into some other projects. Nothing compiles because it says that CocoaLumberjack module is missing.
Does anyone know how to fix this? Am I missing an important step or soemthing?
Well, there are numerous isses you could have faced during importing framework itself. It also depends if you use framework as binary or source code. I assume you were using source code approach as you are the creator of framework. You can however check all approaches here: in this SO question . Lets look at all the steps you need to implement in order to successfully use framework with SPM dependencies in your swift project.
create SPM properly and also link all additional SPM dependencies tutorial here. Make sure all your classes, structs etc. and their coresponding initializer has correct access level property. If you plan to use them outside of the package, use public initializers..
2)Once you created you SPM package, link it to framework. For the sake of this answer I created testFramework and linked one of my custom SPM package called VodApiPackage . This package also contains dependency to another BaseTvApiServicePackage.
I also added TestPrinter file containing simple function for creating error declared in my SPM package. This function servers only for checking that everything is working properly and will be user later. It is also declared public.
import Foundation
import VodApiPackage
public struct TestPrinter {
public init () {}
public func makeTest() {
let x = VodApiError.customErr(msg: "testMsg")
print(x.localizedDescription)
}
}
Open your project and make link to framework, you can also check this nice tutorial. The most important step from tutorial is step 5 and 6. Where you drag .xcproj into your project and link libraries and framework
make sure your library and SPM dependencies are correctly linked in your project. Check sample project below.
Build and test using your framework and its packages:
import UIKit
import testFramework
class ViewController: UIViewController {
override func viewDidLoad() {
super.viewDidLoad()
testmodel()
TestPrinter().makeTest()
}
}
I've found another question which brings more details regarding the problem and possible solutions. It seems like there is a known bug which is a subject for future improvements.
Objective C classes within an iOS Swift-based dynamic framework
I'm developing a framework in Swift and I'm using some Objective-C code inside the framework. So far my module map looks like this:
framework module MyModule {
umbrella header "MyModule-umbrella.h"
export *
explicit module Private {
header "MyTools.h"
}
}
My concern is that all the APIs from MyTools.h are visible from outside the framework: for example, if you install the framework using Cocoapods, then you import MyModule into your application (not MyModule.Private), you are able to access MyTools.h which is not desirable and redundant. Is there any way to make MyTools invisible from outside the framework?
PS. I use Cocoapods to distribute the framework, here is my podspec (the most significant part):
s.module_map = 'Pod/MyModule.modulemap'
s.frameworks = 'CoreData', 'CoreTelephony', 'SystemConfiguration'
s.resources = 'Pod/Classes/MessageStorage/*.xcdatamodeld'
s.public_header_files = 'Pod/Classes/**/*.h'
s.private_header_files = 'Pod/Classes/MyTools/**/*.h'
s.source_files = 'Pod/Classes/**/*.{h,m,swift}'
PSS. My umbrella header does not import MyTools.h
PSSS. Just tried to exclude the header from the main module:
framework module MyModule {
umbrella header "MyModule-umbrella.h"
export *
exclude header "MyTools.h"
explicit module Private {
header "MyTools.h"
}
}
No luck.
I found another question which brings more details regarding the problem and possible solutions (which don't work though). It seems like there is a known bug which is a subject for future improvements.
Objective C classes within an iOS Swift-based dynamic framework
I had exactly the same problems recently. The quick answer is you can't :) Even if you declare "private" modulemap, it can be always imported by your framework users. Please note, that usually, it is not a concern, especially with open source. You just say "this is an internal module, don't use it".
But (there is always but) - you can have behavior, that effectively works the same - allows you to use your Objective-C classes withing same framework target, without making them public. It works in closed source setup, I'm not 100% sure how would it behave with pods.
The case a bit too complex to paste everything here. I'm adding a link to my article about the topic, maybe it will help you. But speaking honestly - it might be a bit of overhead in your setup.
Creating Swift framework with private Objective-C members. The Good, the Bad, and the Ugly
Github example project
For example, I never use the description of XCTestCase.expectation, so I'd like to use a function to provide a default for it, and make it clear via naming that I'm initializing the expectation, as you can't really use an initializer for XCTestExpectation. But if the extension is not in a test target, then it can't compile:
Cannot load underlying module for 'XCTest'
import XCTest
public extension XCTestCase {
func makeExpectation() -> XCTestExpectation {
return expectation(withDescription: "")
}
}
I've created an xcworkspace here (https://github.com/dtweston/TestFrameworkSample) that demonstrates a solution to your issue. There are two projects in this workspace:
SampleApp project with an iOS app target and a unit test target.
SharedTestFramework project that imports XCTest and declares the single extension you put in your question.
The SampleAppTests target links to the SharedTestFramework to be able to use the extension it defines. The single test file imports the SharedTestFramework.
With those steps, I also encounter the Cannot load underlying module for 'XCTest' when building the SharedTestFramework.
The fix for that is to update the Framework Search Paths to include "$(PLATFORM_DIR)/Developer/Library/Frameworks". Now the SharedTestFramework compiles correctly, and as you can see in the workspace I uploaded, the SampleAppTests target is able to use it successfully.
Old and busted answer
Are you building a separate framework that is designed to be imported into test targets? If that's the case then I think you just need to reference XCTest.framework from this custom framework you're building.
On the other hand, if you're trying to add this extension to a framework that is used by your app target, that seems like a bad idea, because it would mean linking XCTest.framework to the binary that goes to the store and runs on people's devices.
I'm not sure if that's possible. I'm more confident that it's not a scenario Apple expects or supports.
I'm giving a MVVMCross a spin, to see if it will be of use in some bigger projects coming up, and it's great. I like the navigation, viewModel location and general cross-platform approach, which is just what I need. However, I'm a bit stuck on splitting out some of the dependency injection depending on the platform.
So, we have the basic application, with a shared portable library, that initialises the service references when starting up:
public TwitterSearchApp()
{
InitaliseServices();
}
private void InitaliseServices()
{
this.RegisterServiceInstance<ITwitterSearchProvider>(new TwitterSearchProvider());
}
Fine. That defines the service implementations that will be used across all the platforms. But what about the situation where I will need different implementations on different platforms - for instance perhaps storage/caching, where the core requirement is the same, but needs to be handled differently on a phone than on a tablet.
I thought it might go in Setup somewhere:
public class Setup : MvxBaseWinRTSetup
{
public Setup(Frame rootFrame): base(rootFrame)
{
}
protected override MvxApplication CreateApp()
{
var app = new TwitterSearchApp();//set platorm specific IoC here maybe?
return app;
}
protected override void AddPluginsLoaders(Cirrious.MvvmCross.Platform.MvxLoaderPluginRegistry loaders)
{ // or perhaps here?
loaders.AddConventionalPlugin<Cirrious.MvvmCross.Plugins.Visibility.WinRT.Plugin>();
base.AddPluginsLoaders(loaders);
}
}
but I'm not sure. I've seen the references to replacing the ViewModel locator, but is there are similar way of replacing the other IoC services?
thanks, great job on the framework in general, I really like how it works (apart from this bit, which I don't understand properly yet)
Toby
There are three basic options:
1. Add the platform specific services in your UI project and then register them in an override during setup - which override you use depends on when your services are needed, but for most cases you can just use the InitializeLastChance override which gets called at the end of initialization:
protected override void InitializeLastChance()
{
this.RegisterServiceInstance<IMyService>(new SingletonMyService());
this.RegisterServiceType<IMyService2, PerCallService2>();
base.InitialiseLastChance();
}
If 'last chance' is too late for your service - if you need the service during the core app startup - then you can override any initialisation step after InitializeIoC - e.g. InitializeFirstChance. For the list and order of initialisation steps, see InitializePrimary and InitializeSecondary in MvxBaseSetup.cs
2. Add the platform specific registration in some other bit of the UI code - e.g. in the constructor for a specific View (this option isn't used much... but you could use it in some odd cases, if you wanted to...)
3. Use a plugin - all plugins are is a wrapper around IoC. Plugins have the disadvantage that they add some development overhead (you have to add the separate projects and the plugin boilerplate files), but they have the advantages that they can be reused across apps and it's easier to write test apps and test harnesses for them. For more info on plugins, see Making mono cross platform support for Task/Intent and see http://slodge.blogspot.co.uk/2012/10/build-new-plugin-for-mvvmcrosss.html
My general advice - start with the first option and migrate out to a plugin if you want to reuse the code in later projects...
I'm attempting to follow this tutorial 'Demystifying Mail.app Plugins on Leopard' to build a Mail.app plugin. Instead of using PyObjC I'm trying to use MacRuby. I've got MacRuby 0.6 loaded up and I've gotten to this step in the tutorial (PyObjC code):
MVMailBundle = objc.lookUpClass('MVMailBundle')
I've search the web a bit but can't seem to find any information about loading the private framework 'MVMailBundle' in MacRuby. Any ideas?
Thanks in advance - AYAL
I think the idea is that this plugin will be loaded into Mail.app, which will already have loaded the private framework in question. So we just want to look up a class at runtime (which is what that Python snippet is doing — not loading a framework). The way to do this in MacRuby is MVMailBundle = NSClassFromString 'MVMailBundle'.
(You will need to include framework 'Cocoa' in order to get the NSClassFromString method, but I assume you'll have already done this.)
MacRuby uses garbage collection and Mail doesn't. You can't load a GC bundle into a non-GC app. So this is a dead-end.