WEBVTT 00:00.150 --> 00:06.780 Hello again! In this video, we are going to look at the for_each() algorithm. for underscore each takes 00:06.780 --> 00:13.860 an iterator range and a callable object. It will iterate over the range and call this function with 00:13.860 --> 00:15.390 each element as the argument. 00:16.650 --> 00:19.290 So here we have the begin() and end() for the string. 00:19.620 --> 00:24.210 We have a lambda expression which points out the elements, followed by a semicolon. 00:25.530 --> 00:28.200 And we could write this using a range-for loop. 00:28.620 --> 00:31.620 So we just iterate over all the characters and print each one out. 00:34.170 --> 00:36.420 So here is that code, we have our string here. 00:37.950 --> 00:42.900 We call for_each() with begin() and end() as the arguments and the lambda expression, which will print 00:42.900 --> 00:43.710 out each character. 00:44.550 --> 00:46.020 And then we have the range for loop. 00:48.690 --> 00:49.410 So there we are. 00:49.560 --> 00:53.940 We get the characters in the string displayed on screen, with a comma between each one. 00:56.180 --> 01:02.000 The function can also take the element by reference, which means we can change the element. So we 01:02.000 --> 01:07.910 have the begin() and end() databases, and we have a lambda expression which takes the element by reference 01:08.630 --> 01:10.910 and replace it with the uppercase equivalent. 01:11.780 --> 01:17.870 And again, we could write this as a range for loop, with reference to auto as the type of the element. 01:22.260 --> 01:27.840 And then we have our for_each() call, the lambda expression and the range-for. 01:30.540 --> 01:33.240 And both those will convert the characters to uppercase. 01:35.130 --> 01:40.240 So you may be wondering what the point is of having this for_each() algorithm if you can use a range-for 01:40.240 --> 01:40.890 a loop instead. 01:41.490 --> 01:46.350 And the main reason is historic. Range-for loops only came in in C++11. 01:46.740 --> 01:49.890 But the algorithms came in in C++ 98. 01:50.490 --> 01:56.250 So for those 13 years, if you wanted to call a function with each element in the container as argument, 01:56.580 --> 01:58.380 then for_each() was the best way to do it. 01:59.610 --> 02:02.520 Now we have range-for loops, and usually these are the better choice. 02:02.820 --> 02:08.790 The code is shorter and clearer, and also you can put in break or return statements if you want to 02:08.790 --> 02:09.870 terminate the loop early. 02:11.730 --> 02:17.430 On the other hand, there are some situations where for_each() is still useful. If you only want a sub-range 02:17.610 --> 02:23.200 so part of a container and not the whole container. There is a single path through the code. 02:23.220 --> 02:25.620 There are no branches, and no early returns. 02:26.010 --> 02:27.810 So that makes it easier to optimize. 02:29.610 --> 02:35.100 And if there are some loop-related things which your implementation does not yet support, it is a useful 02:35.100 --> 02:35.610 fallback. 02:35.910 --> 02:39.870 Visual C++ took a long time to add support for range-for loops. 02:40.500 --> 02:44.610 So until 2015, Microsoft's advice was to use for_each(). 02:46.380 --> 02:52.170 In C++17, many algorithms are being converted to work with parallel processing, but obviously it 02:52.170 --> 02:56.280 takes time for compiler implementers to write all that code. 02:56.700 --> 02:59.070 However, there is a parallel version of for_each. 02:59.640 --> 03:05.070 So if you want an algorithm which is not yet available in parallel form, but you can convert it to 03:05.070 --> 03:10.680 use for_each(), then that is a way of getting a parallel version of that algorithm, for the time being. 03:12.130 --> 03:14.100 OK, so that is it for this video. 03:14.550 --> 03:15.390 I will see you next time. 03:15.630 --> 03:17.700 Until then, keep coding!