Question

-(float)tableView:(UITableView *)tableView heightForRowAtIndexPath:(NSIndexPath *)indexPath{
    UITableViewCell * cell = [self.tableView cellForRowAtIndexPath:indexPath];
    return cell.bounds.size.height;
}

What would be the disadvantage?

I changed that from

-(float)tableView:(UITableView *)tableView heightForRowAtIndexPath:(NSIndexPath *)indexPath{
    UITableViewCell * cell = [self tableView:tableView cellForRowAtIndexPath:indexPath];
    return cell.bounds.size.height;
}

Answer 1

As rmaddy points out, you can’t use the first version because -[UITableView cellForRowAtIndexPath:] can cause the table view to send you tableView:heightForRowAtIndexPath: again, resulting in infinite recursion.

If you are using a static set of cells, with one cell preallocated for each row of the table, then the second version is fine.

If you are dynamically creating cells for rows, the second version will end up draining your table view’s reuse queue and then creating another cell for every row, because tableView:cellForRowAtIndexPath: returns an autoreleased object. None of these cells will be deallocated until the end of the run loop, so in addition to the time cost of creating and destroying all of these cells, you’re also using memory proportional to the number of rows in your table. If you want to do it that way, and you have a lot of rows, you might want to use an explicit autorelease pool:

-(CGFloat)tableView:(UITableView *)tableView heightForRowAtIndexPath:(NSIndexPath *)indexPath{
    @autoreleasepool {
        UITableViewCell *cell = [self tableView:tableView cellForRowAtIndexPath:indexPath];
        return cell.bounds.size.height;
    }
}