Overcoat is a small but powerful library that makes creating REST clients simple and fun. It provides a simple API for making requests and mapping responses to model objects.
Overcoat is built on top of AFNetworking and uses Mantle to map responses into plain or Core Data model objects.
If you need to learn more about Mantle, we recommend these resources:
Overcoat 4.0 is the latest major release and introduces several API-breaking changes to support envelop and error responses, Core Data serialization, and a new method to specify how to map responses to model objects.
If you are upgraded from Overcoat 3.x, check the migration note
Check who's using Overcoat here. You're welcome to add your project/app into this wiki page.
Overcoat supports OS X 10.9+ and iOS 7.0+.
Add the following to your Podfile
and run $ pod install
.
pod 'Overcoat', '~> 4.0.0-beta.1'
If you don't have CocoaPods installed or integrated into your project, you can learn how to do so here.
Add the following to your Cartfile
and run $ carthage update
.
github "Overcoat/Overcoat" "4.0.0-beta.1"
If you don't have Carthage installed or integrated into your project, you can learn how to do so here.
Overcoat includes a simple Twitter client that shows some new features:
- Mapping model classes to resource paths.
- Specifying an error model class.
- Core Data serialization.
- Promises.
You can find the sample code here.
Note that you'll need to run pod install
to install all the dependencies.
Overcoat provides 2 different classes to subclass when creating your own clients:
Class | Usage |
---|---|
OVCHTTPSessionManager |
Using with NSURLSession and Mantle |
OVCManagedHTTPSessionManager |
Using with NSURLSession, Mantle, and CoreData. This is also a subclass of OVCHTTPSessionManager |
Both classes have identical APIs.
#import <Overcoat/Overcoat.h>
@interface TwitterClient : OVCHTTPSessionManager
...
@end
To specify how responses should be mapped to model classes you must override +modelClassesByResourcePath
and return a dictionary mapping resource paths to model classes.
+ (NSDictionary *)modelClassesByResourcePath {
return @{
@"statuses/*": [Tweet class],
@"users/*": [TwitterUser class],
@"friends/ids.json": [UserIdentifierCollection class],
@"followers/ids.json": [UserIdentifierCollection class]
};
}
You don't need to specify the full path, and you can use *
and **
to match any text or #
to match only digits.
If you use *
and #
, they are strict path matchings, so the number of path components must be equal. The **
just matches any text and has no path components number limitation.
Match String | Path | Result |
---|---|---|
statuses/* |
statuses/user_timeline |
Matched |
statuses/# |
statuses/user_timeline |
Missed (wrong type, the path component after statuses should be dights only) |
statuses/* |
statuses/retweets/12345 |
Missed (wrong number of path components, there should be only one path component after statuses ) |
statuses/*/* |
statuses/retweets/12345 |
Matched |
statuses/** |
statuses/retweets/12345 |
Matched |
statuses/** |
statuses/retweets/12345/extra |
Matched (the number of path components doesn't matter) |
statuses/retweets/* |
statuses/retweets/12345 |
Matched |
statuses/retweets/# |
statuses/retweets/12345 |
Matched |
Also you can specify different model classes by request method or response status code, like:
+ (NSDictionary *)modelClassesByResourcePath {
return @{
@"statuses/*": [Tweet class],
@"users/*": @{
@"PUT": [UpdatedTwitterUser class], // For PUT request method,
@"201": [NewCreatedTwitterUser class], // For 201 response status code
@"*": [TwitterUser class], // For all other cases, as fallback
},
@"friends/ids.json": [UserIdentifierCollection class],
@"followers/ids.json": [UserIdentifierCollection class]
};
}
Check the documentation of OVCURLMatcherNode
for further explaination.
Different REST APIs have different ways of dealing with status and other metadata.
Pure REST services like Twitter use HTTP status codes and a specific JSON response to communicate errors;
and HTTP headers for other metadata like rate limits. For these kind of services, you may want to override
+errorModelClassesByResourcePath
to map error responses into your own model.
+ (Class)errorModelClassesByResourcePath {
return @{@"**": [TwitterErrorResponse class]};
}
Other services like App.net use an envelop response, which is a top level JSON response containing
the data requested and additional metadata. For these kind of services, you must create your own OVCResponse
subclass and specify the data key path.
@interface AppDotNetResponse : OVCResponse
...
@end
@implementation AppDotNetResponse
+ (NSString *)resultKeyPathForJSONDictionary:(NSDictionary *)JSONDictionary {
return @"data";
}
@end
You can then specify which response class to use in your client by overriding +responseClassesByResourcePath
.
+ (Class)responseClassesByResourcePath {
return @{@"**": [AppDotNetResponse class]};
}
To support CoreData serialization, you have to use Overcoat+CoreData
extension if you're using CocoaPods
pod 'Overcoat+CoreData', '~> 4.0' # Use this,
Or if you are using Carthage, you also have to add OvercoatCoreData.framework
to your project.
And the main classes would be changed to OVCManaged
prefixed one. (For instance,
OVCHTTPSessionManager
-> OVCManagedHTTPSessionManager
)
If you initialize your client with a valid NSManagedObjectContext
,
it will automatically persist any model object(s) parsed from a response,
if the model supports Core Data serialization (that is, implements MTLManagedObjectSerializing
).
Note that, if you provide a context with an NSMainQueueConcurrencyType
,
a private context will be created to perform insertions in the background.
You can see Core Data Serialization in action in the provided example.
// Lookup Twitter users
NSDictionary *parameters = @{
@"screen_name": @"gonzalezreal",
@"user_id": @"42,3141592"
};
[twitterClient GET:@"users/lookup.json" parameters:parameters completion:^(OVCResponse *response, NSError *error) {
NSArray *users = response.result; // This is an array of TwitterUser objects!
}];
Note that Overcoat automatically parses the JSON into model objects, that is, in this case response.result
contains an array of TwitterUser
objects.
From 2.0, Overcoat adds support for ReactiveCocoa.
To add ReactiveCocoa support, you have to use Overcoat+ReactiveCocoa
podspec if you're using CocoaPods.
Or if you're using Carthage, add OvercoatReactiveCocoa.framework
Now you can make HTTP requests and get cold signals to handle responses:
#import <OvercoatReactiveCocoa/OvercoatReactiveCocoa.h>
...
[[twitterClient rac_GET:@"users/lookup.json" parameters:parameters] subscribeNext:^(OVCResponse *response) {
...
} error:^(NSError *e) {
...
}];
If you're looking for a better way to handle asynchronous calls but you're not ready to embrace ReactiveCocoa, you may try PromiseKit.
To add PromiseKit support, you have to use Overcoat+PromiseKit
podspec if you're using CocoaPods.
Or if you're using Carthage, add OvercoatPromiseKit.framework
Now you can get PMKPromise
objects when making HTTP requests:
#import <OvercoatPromiseKit/OvercoatPromiseKit.h>
...
[twitterClient pmk_GET:@"users/lookup.json" parameters:parameters].then(^(OVCResponse *response) {
return response.result;
});
In order to build the library and run unit tests, you will first need to install 2 tools: cocoapods
and xctool
- cocoapods: https://cocoapods.org
- xctool: https://github.com/facebook/xctool
After you setup these tools (or you already have these tools), you could run tests via
make test
Check the Makefile
to run other test target.
Overcoat is available under the MIT license. See LICENSE.md.