I am converting my application from Syncronous to Asyncronous HTTP requests and have ran into a problem that looks like it will require quite a big reworking of how the application handles its data. Let me try to explain
Previously it was like this:
-Class1, Class2 and Class3 were all subclasses of UIViewController
-Helper class
-Content display class
They do broadly different things but the common trait is their interaction with the helper class. They gather details of a request in a number of different ways from a user and then eventually send a request to the helper class.
When it was done syncronously the helper class would return the data. Each class would then interpret the data (XML files) and pass them on to the Content display class via a segue
So something broadly like this:
Class1:
//Get user input
SomeData *data = [helperclass makerequest];
id vcData = [data process];
[self performSegueWithIdentifier:#"segueIdentifier"];
---
- (void)prepareForSegue:(UIStoryboardSegue *)segue
{
DestinationViewController *destination = (DestinationViewController *)segue.destinationViewController;
destination.data = vcData;
}
Content display class:
- (void)viewDidLoad
{
[super viewDidLoad];
[self.data presentdata];
}
Now it looks like this
I dealt with this problem by first making it work with Class1 with a view to deploying the fix to class2 and class3. So class1 and helper now interact like this
Class1:
//Get user input
SomeData *data = [helperclass makerequestWithSender:self];
id vcData = [data process];
[self performSegueWithIdentifier:#"segueIdentifier"];
---
- (void)prepareForSegue:(UIStoryboardSegue *)segue
{
DestinationViewController *destination = (DestinationViewController *)segue.destinationViewController;
destination.data = vcData;
}
Now the biggest problem I am facing is how to get the data from helperclass back to Class1. I managed to get it to work by doing
(void)makeRequestWithSender:(Class1*)sender
{
[NSURLConnection sendAsynchronousRequest:...
{
[sender sendData:data];
}
}
However, when I have came to roll this out to the other 2 GUI classed which will compose the request I am having difficulty with. My first thought was to set sender:(id) but that fails at the line [sender sendData:data] telling me that id does not have an method sendData: or similar.
Hopefully I wasn't too vague here and you guys can help. If required I will be able to post code snippets but for now can anyone help with a better suggestion about how to structure the code for this request?
You basically want to use the 'observer pattern' or a (maybe) slightly changed setup, so you can use delegation.
Observer pattern
You gain the mechanic via the NSNotificationCenter and NSNotifications. Your 3 different UIViewController subclasses each subscribe to a specific NSNotification and you notify them via posting a notification via the NSNotificationCenter.
The following code is an example of how you can approach the problem in your viewcontroller subclasses:
- (void)viewWillAppear:(BOOL)animated {
[super viewWillAppear:animated];
// subscribe to a specific notification
[[NSNotificationCenter defaultCenter] addObserver:self selector:#selector(doSomethingWithTheData:) name:#"MyDataChangedNotification" object:nil];
}
- (void)viewWillDisappear:(BOOL)animated {
[super viewWillDisappear:animated];
// do not forget to unsubscribe the observer, or you may experience crashes towards a deallocated observer
[[NSNotificationCenter defaultCenter] removeObserver:self];
}
...
- (void)doSomethingWithTheData:(NSNotification *)notification {
// you grab your data our of the notifications userinfo
MyDataObject *myChangedData = [[notification userInfo] objectForKey:#"myChangedDataKey"];
...
}
In your helper class, after the data changed you have to inform the observers, e.g.
-(void)myDataDidChangeHere {
MyDataObject *myChangedData = ...;
// you can add you data to the notification (to later access it in your viewcontrollers)
[[NSNotificationCenter defaultCenter] postNotificationName:#"MyDataChangedNotification" object:nil userInfo:#{#"myChangedDataKey" : myChangedData}];
}
via #protocol
Presuming all your UIViewController subclasses reside in a parent viewcontroller, you can implement a protocol in your helper class and make the parent viewcontroller the delegate. Then the parent viewcontroller may inform the child uiviewcontrollers via passing a message.
Your helper class declaration could look like this (presuming ARC):
#protocol HelperDelegate;
#interface Helper : NSObject
#property (nonatomic, weak) id<HelperDelegate> delegate;
...
#end
#protocol HelperDelegate <NSObject>
-(void)helper:(Helper *)helper dataDidChange:(MyDataObject*)data;
#end
In the helper implementation you would inform the delegate via:
...
if ([self.delegate respondsToSelector:#selector(helper:dataDidChange:)]) {
[self.delegate helper:self dataDidChange:myChangedDataObject];
}
...
Your parent viewcontroller would need to be the delegate of the helper class and implement its protocol; a rough sketch, in the declaration
#interface ParentViewController : UIViewController <HelperDelegate>
and for the implementation in short version
// you alloc init your helper and assign the delegate to self, also of course implement the delegate method
-(void)helper:(Helper *)helper dataDidChange:(MyDataObject*)data {
[self.myCustomChildViewController doSomethingWithTheNewData:data];
}
Besides..
You might ask yourself which method to prefer. Both are viable, the main difference is that via the observer pattern you get more objects to be informed 'at once', whereas a protocol can only have one delegate and that one has to forward the message if needed. There are a lot of discussions around about pros and cons. I'd suggest you read up on them once you made up your mind (sorry ain't got enough reputation to post more than two links, so please search on stackoverflow). If something is unclear, please ask.
Some reasonable ideas here. To elaborate/add my opinion:
First, which object ought to tell the downloader (HelperClass) to begin downloading? My practice is to do this in the view controller that will present the data. So I generally start network requests after a segue (like in viewWillAppear: of the presented vc), not before.
Next, when one class needs to execute code provided for another, I first think about if it makes sense to do it using a block. Very often (not always) blocks make more sense and provide more readable code than, say, delegate, notification, KVO, etc. I think NSURLConnection completion, for example, is better suited to blocks than delegate. (and Apple kind of agrees, having introduced + (void)sendAsynchronousRequest:(NSURLRequest *)request queue:(NSOperationQueue *)queue completionHandler:(void (^)(NSURLResponse*, NSData*, NSError*))handler).
So my pattern for your app would be this:
// Class1.m
// when user has completed providing input
...
// don't do any request yet. just start a segue
[self performSegueWithIdentifier:#"ToContentDisplayClass" sender:self];
...
- (void)prepareForSegue:(UIStoryboardSegue *)segue sender:(id)sender {
// don't do a request yet, just marshall the data needed for the request
// and send it to the vc who actually cares about the request/result
if ([segue.identifier isEqualToString:#"ToContentDisplayClass"]) {
NSArray *userInput = // collect user input in a collection or custom object
ContentDisplayClass *vc = segue.destinationViewController;
vc.dataNeededForRequest = userInput;
}
...
Then in ContentDisplayClass.m
// this is the class that will present the result, let it make the request
- (void)viewWillAppear:(BOOL)animated {
[super viewWillAppear:animated];
HelperClass *helper = [[HelperClass alloc]
initWithDataNeededForRequest:self.dataNeededForRequest];
// helper class forms a request using the data provided from the original vc,
// then...
[helper sendRequestWithCompletion:^(NSURLResponse *response, NSData *data, NSError *error) {
if (!error) {
// interpret data, update view
self.label.text = // string we pulled out of data
} else {
// present an AlertView? dismiss this vc?
}
}];
This depends on HelperClass implementing the block form of NSURLConnection
// HelperClass.m
- (id)initWithDataNeededForRequest:(id)dataNeededForRequest {
// standard init pattern, set properties from the param
}
- (void)sendRequestWithCompletion:(void (^)(NSURLResponse *, NSData *, NSError *))completion {
NSURLRequest *request = ...
// the stuff we need to formulate the request has been setup in init
// use NSURLConnection block method
[NSURLConnection sendAsynchronousRequest:request
queue:[NSOperationQueue mainQueue]
completionHandler:completion];
}
Edit - there are several rationale's for making the VC transition before starting the network request:
1) Build the standard behavior around the success case: unless the app is about testing network connections, the success case is that the request works.
2) The cardinal principal for an app is to be responsive, to do something sensible immediately upon user actions. So when the user does something to initiate the request, an immediate vc transition is good. (what instead? a spinner?). The newly presented UI might even reduce the perceived latency of the request by giving user something new to look at while it runs.
3) What should an app do when a request fails? If the app doesn't really need the request to be useful, then doing nothing is a good option, so you'd want to be on the new vc. More typically, the request is necessary to proceed. The UI should be "responsive" to request failure, too. Typical behavior is to present an alert that offers some form of "retry" or "cancel". For either choice, the place the UI wants to be is on the new vc. Retry is more obvious, because that's where it always is when it tries to fetch the data. For cancel, the way to be "responsive" to cancel is to go back to the old vc, a vc transition back isn't ugly, it's what the user just asked for.
I'm not 100% clear on how you're handling the data now, but to change your data to asynchronous calls, I would use blocks. For instance your current synchronous code like this:
//Get user input
data = [helperclass makerequest]
sendData = [data process]
would turn into something like this:
//Get user input
data = [helperclass makerequestWithSuccess:^{
sendData = [data process]
}];
Using a success block will allow you to wait to process the data until the makerequest was finished.
Your new makerequest function would now look like this:
-(void)makerequestWithSuccess:(void (^)(void))success{
// Put your makerequest code here
// After your makerequest is completed successfully, call:
success();
}
Hope this helps!
I'm not sure that I understood your problem correctly, but if it's sort of:
Start task A asynchronously.
When task A finished successfully, get its result and start task B whose input is result A.
When task B finished successfully, get its result and start task C whose input is result B.
...
When finished successfully, be happy, otherwise print error.
A code example would look like this:
typedef (void)(^completion_block_t)(id result);
-(void) asyncTaskA:(completion_block_t)completionHandler;
-(void) asyncTaskBWithInput:(id)input completion:(completion_block_t)completionHandler;
-(void) asyncTaskCWithInput:(id)input completion:(completion_block_t)completionHandler;
-(void) asyncSomethingWithCompletion:(completion_block_t)completionHandler;
-(void) asyncSomethingWithCompletion:(completion_block_t)completionHandler
{
[self asyncTaskA:^(id resultA){
if (![resultA isKindOfClass:[NSError class]]) {
[self asyncTaskBWithInput:resultA completion:^(id resultB){
if (![resultB isKindOfClass:[NSError class]]) {
[self asyncTaskCWithInput:resultB completion:^(id resultC) {
completionHandler(resultC);
}];
}
else {
completionHandler(resultB); // error;
}
}];
}
else {
completionHandler(resultA); // error
}
}];
}
And you use it like:
[self asyncSomethingWithCompletion:^(id result){
if ([result isKindOfClass:[NSError class]]) {
NSLog(#"ERROR: %#", error);
}
else {
// success!
self.myData = result;
}
}];
The "continuation" and error handling makes this a bit confusing (and Objective-C syntax doesn't really add for more readability).
Another example with a third party library support:
The same logic can be written as this:
-(Promise*) asyncTaskA;
-(Promise*) asyncTaskBWithInput;
-(Promise*) asyncTaskCWithInput;
-(Promise*) asyncSomething;
- (Promise*) asyncSomething
{
return [self asyncTaskA]
.then(id^(id result) {
return [self asyncTaskBWithInput:result];
}, nil)
.then(id^(id result) {
return [self asyncTaskCWithInput:result];
}, nil);
}
And it is used as follows:
[self asyncSomething]
.then(^(id result) {
self.myData = result;
return nil;
},
^id(NSError* error) {
NSLog(#"ERROR: %#", error);
return nil;
});
If you like the latter more, the "Promise" framework is available on GitHub: RXPromise - I'm the author ;)
I'm not sure if what I've done in the past is relevant to your problem, but what I've done is create a download class that has a delegate protocol with a single method: -(void)downloadFinished:(id) data.
Any class that needs to get asynchronous data, creates an instance of this download class, and sets itself as the delegate. I call downloadFinished: from both connection:didFailWithError: and connectionDidFinishLoading:. Then, in the implementation of that method in the delegate, I check whether the data's class is NSData or NSError, and evaluate that data however is appropriate for that class.
I'm new to the concept of delegates and selectors when used with notifications. So my first question is,
1) Let's say you have a button that has a delegate that implements some doWork method. If you want the same functionality that's in the method, is it 'ok' to just call that method? I didn't know if that was considered good coding practices and/or if you should do that, or do something different in getting that type of functionality. Like if that is ok architecture?
2) Similarly, with NSNotificationCenter, I see some code that posts a notification. Then there's a HandleSegmentedControl:(NSNotification *)notification method. If I want to manually have that functionality, but without pressing the segment control, is it 'ok' to just take that functionality out of that method and put it in a new method so it would look like this:
Original:
- (void)HandleSegmentedControl:(NSNotification *)notification {
NSDictionary *dict = [userInfo notification];
// do stuff with the dictionary
}
New:
- (void)HandleSegmentedControl:(NSNotification *)notification {
NSDictionary *dict = [userInfo notification];
[self newMethod:dict];
}
- (void)newMethod:(NSDictionary *)dict {
// do stuff with the dictionary
}
- (void)myOtherMethodThatNeedsTheSameFunctionality {
NSDictionary *dict = // create some dictionary
[self newMethod:dict];
}
Sorry if these are basic questions. I'm not sure what the best practices are for things like this and wanted to start the right way. Thanks.
If the delegate protocol implements the doWork method as a required method, then yes. However, if it is an optional method, or if you want to be extra safe, you should use
if ([delegate respondsToSelector:#selector(doWork)]) {
[delegate doWork];
}
Sure, that seems like a reasonable thing to do. It is a common way to make your code more robust. The notification's userInfo is there so you can send data as you like. However, I think in your code you meant [notification userInfo] not [userInfo notification].
I will try to make myself as clear as possible. Let start from the beginning. I have an application with a tableview that contains a list of places with distances from myLocation. Now everytime I get an update in the gps location I run the following code
- (void)locationUpdate:(CLLocation *)location {
myLocation = location;
for (Trek * trek in list) {
CLLocation *loc = [[CLLocation alloc] initWithLatitude:[trek latitude_start] longitude:[trek longitude_start]];
double dis = [locationManager getDistance: loc];
[trek setDistance:dis];
[trek setDistanceUnit];
[loc release];
}
[self.tableView reloadData];
}
Now this piece of code [trek setDistanceUnit]; calls
-(void) setDistanceUnit {
if (self.distance < 1000.0)
self.distanceString = [NSString stringWithFormat:#"%.0lf m", self.distance];
}
Now if I use only distanceString the application crash. Now I think it may have something to do with the fact that those updates may run concurrently (in parallel) to the access required by the view to draw the cells. Anyone has any idea? I can post more code if helpful, I just didn't want to post too much to make this post too long.
I tried to search everywhere but I could not found anything so far.
Thanks in advance,
Umberto
PS Now the application is working but I would like to understand what is going on.
If your distanceString is a retain property, assigning it without self sets it up for a crash because you bypass the setter, and assign the string without retaining it. So when the string gets deallocated on being sent to the autorelease pool, your app crashes.
By synthesizing the accessors using #synthesize and using the dot notation (or setDistanceString:), the object will retain the string for you so that it always has a pointer to it for itself (until it's released).
Whenever the user scrolls map or zooms in/out, this method gets called instantaneously. I want to delay the call to this method by say 2 secs. Is it possible to do that?
You could implement that method like this:
-(void)mapView:(MKMapView *)mapView regionDidChangeAnimated:(BOOL)animated
{
NSNumber *animatedNumber = [NSNumber numberWithBool:animated];
NSArray *args = [[NSArray alloc] initWithObjects:mapView,
animatedNumber,nil];
[self performSelector:#selector(delayedMapViewRegionDidChangeAnimated:)
withObject:args
afterDelay:2.0f];
[args release];
}
Then, somewhere in the same class:
-(void)delayedMapViewRegionDidChangeAnimated:(NSArray *)args
{
MKMapView *mapView = [args objectAtIndex:0];
BOOL animated = [[args objectAtIndex:1] boolValue];
// do what you would have done in mapView:regionDidChangeAnimated: here
}
Of course, if you don't need one of those arguments (either mapView or animated), you could make this considerably simpler by only passing the one you did need.
If you can't just edit the code for your MKMapViewDelegate, perhaps you could do something similar with method swizzling, although then you're getting really hacky.
You can send a delayed message with performSelector:withObject:afterDelay: or one of its related methods.
I'm currently asynchronously using NSURLConnection with several UIViews (every view handles NSURLConnection as delegate). The problem I have is when user switches views too fast and the delegate becomes NSZombie the app crashes - that is NSURLConnection doesn't have living delegate anymore. So, the first question is if there's a way to circumvent this?
The second question is simple - how do I handle NSZombie? Simple if(myObject != nil).. doesn't work at all.
You need to cancel the NSURLConnection before you dispose it's delegate. Simply keep a reference to the NSURLConnection in your UIView that acts as a delegate and call [urlConnection cancel].
After you release a message you need to set your pointer to it to nil if you continue using that pointer. As an example:
id myObject = [[SomeObject alloc] init];
/* Some code */
[myObject release];
myObject = nil;
/* Some more code */
if (myObject != nil) {
[myObject doSomething];
}
Notice however that it is valid to send a message to nil so you don't need to safe guard the message sending. It simply won't have any effect if myObject == nil.