dispatch_group_notify被调用多次,因为在嵌套块中调用dispatch_group_leave

Question

我在下面有这个逻辑; 它们基本上是3个嵌套的调度组块.

1. 第一组(group)将执行3个短异步任务(仅从Web服务下载数据)和1个更长的异步任务:将未同步的记录上载到Web服务,在本地删除已同步的记录,最后从Web服务下载记录(首先一个包含ID和基本信息的数组,然后是这些记录中的每一个).
2. 第二组(saveGroup)是较长任务的一部分.它将等待所有未完成的记录对Web服务的请求完成.
3. 第三个(downloadGroup)将等到所有这些单个记录下载请求到服务完成.

一切顺利,直到第三个派遣小组.如您所见,我获取服务器上记录的ID和基本信息,遍历数组并使用downloadGroup调用dispatch_group_enter,然后触发HTTP请求.在dispatch_group_leave完成请求时被调用.我可以看到dispatch_group_leave每个请求被调用,最后dispatch_group_notify被调用多次.

-(void) doTheSync {
    dispatch_group_t group = dispatch_group_create();

    [self syncFirstDataWithGroup: group];
    [self syncSecondDataWithGroup: group];
    [self syncThirdDataWithGroup: group];

    [self syncRecordsWithExternalGroup: group];

    dispatch_group_notify(group, dispatch_get_main_queue(), ^{
        NSLog(@"Finished all the sync");
    });
}

-(void) syncRecordsWithExternalGroup: (dispatch_group_t) externalGroup {
    dispatch_group_enter(externalGroup);

    NSError* error = nil;
    ConfigureDataModelHandler* configDataHandler = [ConfigureDataModelHandler sharedCoredata];
    WebserviceLib* RESTClient = [WebserviceLib sharedInstance];

    //get all unsynced records (f
    NSArray* recordsUnsynced = [configDataHandler getAllRecordsWithSynced: NO ignoreDelete: YES withError: &error];

    if (!error) {
        //upload them to the BE (and mark as synced if succeed

        dispatch_group_t saveGroup = dispatch_group_create();

        //get full record dictionary, save and mark as synced
        for (CMrecord* record in recordsUnsynced) {
            NSDictionary* recordDict = [configDataHandler exportToDictionary: record.recordID.integerValue];

            dispatch_group_enter(saveGroup);

            [RESTClient saverecord: recordDict onComplete:^(AFHTTPRequestOperation *operation, id responseObject) {
                NSLog(@"Saved unsynced record (%@) to BE", record.recordID);

                //mark as synced
                [configDataHandler markrecordAsSynced: record.recordID.integerValue];
                dispatch_group_leave(saveGroup);
            } onError:^(AFHTTPRequestOperation *operation, NSError *error) {
                NSLog(@"Error saving unsynced record to BE %@", error);
                dispatch_group_leave(saveGroup);
            }];
        }

        //** NOTIFY FINISH SAVING **
        dispatch_group_notify(saveGroup, dispatch_get_main_queue(), ^{
            NSLog(@"Finished saving all unsynced records to BE");

            //delete all synced records
            //TODO: Check if this makes sense. Probably better to not delete anything until we got the full record from the BE...
            [configDataHandler deleteRecordsSynced];

            //download small records from BE
            NSString* agentNationalID = [self.coreData getLoginStatus].nationalID;
            [RESTClient getRecordsForAgent: agentNationalID onComplete:^(NSInteger completeCode, NSArray *responseArray) {
                NSLog(@"Success getting the records %@", responseArray);

                if (completeCode == 200) {
                    dispatch_group_t downloadGroup = dispatch_group_create();

                    for (NSDictionary* shortDict in responseArray) {
                            dispatch_group_enter(downloadGroup);

                            //download full records from BE
                            [RESTClient getRecordByCodeAndTimestamp: shortDict onComplete:^(NSInteger completeCode, NSDictionary *responseDictionary) {
                                NSLog(@"Success Downloading record");
                                dispatch_group_leave(downloadGroup);
                            } onError:^(AFHTTPRequestOperation *operation, NSError *error) {
                                NSLog(@"Error downloading record %@", shortDict);
                                dispatch_group_leave(downloadGroup);
                            }];

                            //** NOTIFY FINISH DOWNLOADING **
                            dispatch_group_notify(downloadGroup, dispatch_get_main_queue(), ^{
                                NSLog(@"Finished downloading all the records");

                                //This is CRASHING because this block is being called as many times as the dispatch_group_leave(downloadGroup) is called
                                dispatch_group_leave(externalGroup);
                            });
                        }
                } else {
                    NSLog(@"Error getting the records for Agent %@, with Code %li", @"AGENT ID", (long)completeCode);
                }
            } onError:^(AFHTTPRequestOperation *operation, NSError *error) {
                NSLog(@"Error getting the records for Agent %@,  %@", @"AGENT ID", operation.response);
            }];
        });
    }
}

我在其他部分使用具有正常行为的调度组(创建,输入,输入,离开,离开,通知),所以我不明白这里发生了什么.它是否与嵌套块相关？关于如何dispatch_group_notify在完成时只调用一次的任何建议,或者更好的是,如何以更干净的方式实现这种嵌套的异步任务完成依赖(意味着如何等待多个请求完成,然后再激发一些并再次等待那么)？

Answer 1

您downloadGroup每次输入时都会通知该组

for (NSDictionary* shortDict in responseArray) {
    dispatch_group_enter(downloadGroup);

    //download full records from BE
    [RESTClient getRecordByCodeAndTimestamp: shortDict onComplete:^(NSInteger completeCode, NSDictionary *responseDictionary) {
        NSLog(@"Success Downloading record");
        dispatch_group_leave(downloadGroup);
    } onError:^(AFHTTPRequestOperation *operation, NSError *error) {
        NSLog(@"Error downloading record %@", shortDict);
        dispatch_group_leave(downloadGroup);
    }];

    // BUG IS HERE
    dispatch_group_notify(downloadGroup, dispatch_get_main_queue(), ^{
        NSLog(@"Finished downloading all the records");

        dispatch_group_leave(externalGroup);
    });
}

// dispatch_group_notify should be moved HERE

你只应该通知小组一次,dispatch_group_notify离开循环.