Alamofire completionHandler of responseJSON is not called

Question

I have the following code to fetch the replies to a list of comments. (1 comment has many replies)

static func fetchCommentsAndTheirReplies(articleId: String, failure: (()->Void)?, success: (comments: [[String: AnyObject]], replies: [[[String: AnyObject]]], userIds: Set<String>)->Void) {
    var retComments = [[String: AnyObject]]()
    var retReplies = [[[String: AnyObject]]]()
    var retUserIds = Set<String>()

    Alamofire.request(.GET, API.listComment, parameters: [API.articleId: articleId]).responseJSON {
        response in
        guard let comments = response.result.value as? [[String: AnyObject]] else {
            failure?()
            return
        }
        print(comments)
        retComments = comments

        let group = dispatch_group_create()

        for (commentIndex, comment) in comments.enumerate() {
            guard let id = comment["_id"] as? String else {continue}

            let relevantUserIds = parseRelaventUserIdsFromEntity(comment)
            for userId in relevantUserIds {
                retUserIds.insert(userId)
            }

            retReplies.append([[String: AnyObject]]())

            dispatch_group_enter(group)
            Alamofire.request(.GET, API.listReply, parameters: [API.commentId: id]).responseJSON {
                response in
                if let replies = response.result.value as? [[String: AnyObject]] {
                    for (_, reply) in replies.enumerate() {

                        let relevantUserIds = parseRelaventUserIdsFromEntity(reply)
                        for userId in relevantUserIds {
                            retUserIds.insert(userId)
                        }
                    }
                    //TODO: need to capture commentIndex?
                    retReplies[commentIndex] = replies
                }
                dispatch_group_leave(group)
            }


        }

        dispatch_group_wait(group, DISPATCH_TIME_FOREVER)
        success(comments: retComments, replies: retReplies, userIds: retUserIds)
    }
}

The complete handler of API.listReply request is never called. dispatch_group_enter(group) is called once, and dispatch_group_leave(group) is never called. The code gets stuck at dispatch_group_wait. What's strange is, even the UI is stuck, which is strange because the entire function is async.

Answer 1

I have encounter similar problem:

on Main UI thread, call:

dispatch_semaphore_wait(loginDoneSemaphore, DISPATCH_TIME_FOREVER)

and call http using Alamofire, internal also use your httpRequest.responseJSON

-> finally found code inside responseJSON completion handler never called

-> cause find code after DISPATCH_TIME_FOREVER never called

-> finally find root cause is: Alamofire's responseJSON, default, if you not pass in the thread/queue, will run on Main UI thread

-> before call Alamofire, have inside Main UI thread, to use DISPATCH_TIME_FOREVER lock UI thread

-> so following Alamofire's responseJSON, which run also on Main UI thread, will never called

-> my solution is: designate Alamofire do http response on another thread:

let BackgroundThread:dispatch_queue_t = dispatch_get_global_queue(QOS_CLASS_BACKGROUND, 0)

func dispatchBackground_async(thingsTodo:()->()) {
    dispatch_async(BackgroundThread, thingsTodo)
}

dispatchBackground_async({
    httpRequest.responseJSON(queue: BackgroundThread, completionHandler: { response in
        gLog.debug("request=\(response.request), response=\(response.response), statusCode=\(response.response?.statusCode), result=\(response.result)")
        // [Debug] [com.apple.root.background-qos] [CrifanLibHttp.swift:21]
})

this maybe useful for you to refer.

-> maybe you can use my method: set another thread for Alamofire's responseJSON, to solve your problem.

Answer 2

I'm not familiar with Alamofire internals, but it seems likely that it fulfills all your response completionBlocks on the same serial queue. Your dispatch_group_wait is blocking other responses from finishing and calling their dispatch_group_leaves.

You can solve this problem by using dispatch_group_notify instead of dispatch_group_wait, so rather than blocking the thread it will simply submit the block to a queue when necessary.