Question

Below is a sample function where I try to run one main group, wait, then run two other background tasks on separate threads, wait, then return a value altered in the general blocks shown below. What is shown below is my guess at how to do this. It works if I run these blocks all in one block. It fails when I separate the blocks. Does anyone have an example of how they accomplish something similar? Thanks in advance for your help.

-(NSString *)sampleFunction:(NSString*)inputString
{
 __block NSString *returnString;

 dispatch_group_t mainGroup = dispatch_group_create();
 dispatch_group_t otherGroup = dispatch_group_create();

 void (^firstBlock)(void) = ^(void)
 {
  ...
 };

 void (^secondBlock)(void) = ^(void)
 {
  ...
 };


 void (^thirdBlock)(void) = ^(void)
 {
  ...
 };

 dispatch_group_async(oneGroup, dispatch_get_global_queue(0, 0), firstBlock);

 dispatch_group_wait(oneGroup, sizeof(int));

 dispatch_group_async(otherGroup, dispatch_get_global_queue(0, 0), secondBlock);
 dispatch_group_async(otherGroup, dispatch_get_global_queue(0, 0), thirdBlock);

 dispatch_group_wait(otherGroup, sizeof(int));

 dispatch_release(userGroup); 
 dispatch_release(otherGroup);

 return returnString;
}

Answer 1

There are lots of different ways to do this, but the most conceptually simple (or at least obvious) way would be to create a group for each stage of your “pipeline”, then have the completion callback for group A schedule the work on group B which also has a completion callback which schedules the work on group C.

The key here is not to actually “wait” for a group to complete – that’s unnecessary. Groups can have completion blocks that will get run automatically when the last block in that group finishes (see dispatch_group_notify()), and this is a good way of creating “fan in” points where you’ve fanned out a bunch of work in the group.