WEBVTT 00:00.120 --> 00:00.690 Hello again! 00:00.990 --> 00:04.350 In this video, we are going to look at miscellaneous template features. 00:06.460 --> 00:12.550 The assert() function, or macro, I think, is inherited from C. It is in the header. 00:13.570 --> 00:15.660 assert() will check its argument at run-time. 00:16.000 --> 00:22.030 If the argument is zero, it will call the library abort() function, which will terminate the program 00:22.090 --> 00:22.690 immediately. 00:23.380 --> 00:28.900 Otherwise, if the arguments is not zero, then the program will continue executing, as normal. 00:29.830 --> 00:33.000 So if 'x' is equal to 42, then nothing will happen. 00:33.010 --> 00:34.090 The program will continue. 00:34.810 --> 00:39.070 If 'x' is not equal to 42, then the program will terminate immediately. 00:40.090 --> 00:42.340 And this is useful for checking invariants. 00:42.880 --> 00:44.560 So these are things that must be true. 00:45.220 --> 00:49.900 For example, if you are going to pass 'x' to a function, and it is very important that the function argument 00:49.900 --> 00:58.240 has the value 42, then this is one way that you can enforce that. There is some runtime overhead to 00:58.240 --> 00:59.170 doing these checks. 01:00.010 --> 01:06.620 And also this is really something that you do not want, in code that is being used by customers. Just 01:06.620 --> 01:08.260 existing with no explanation. 01:08.950 --> 01:10.930 So it is possible to turn this off. 01:11.440 --> 01:17.800 If you define the symbol 'NDEBUG' in the pre-processor, then the assert macro does nothing at all. 01:20.160 --> 01:26.070 In C++11, we have a compile-time version of assert(), called static underscore assert. 01:27.030 --> 01:29.700 This takes a bool argument, and a string. 01:30.540 --> 01:32.880 These have to be available at compile time. 01:33.330 --> 01:40.500 So the bool has to be a constant expression, and the string has to be a string literal. Not a C++ library 01:40.500 --> 01:41.010 string. 01:42.180 --> 01:44.730 The compiler will evaluate this bool. 01:45.450 --> 01:50.850 If the bool evaluates as false, then the compiler will immediately stop processing the source code. 01:51.360 --> 01:55.320 And it will produce an error, which includes the string as part of the error message. 01:57.400 --> 02:03.310 If the bool evaluates to true, then the compiler will carry on, and compile the rest of the source code. 02:04.720 --> 02:09.010 So if the size of int is equal to 8, then nothing will happen. 02:09.850 --> 02:15.760 If size of int is not equal to 8, then the compiler will stop. And the error message will include 02:15.760 --> 02:16.210 the words, 02:16.510 --> 02:19.210 "This program requires a 64-bit compiler" 02:21.200 --> 02:23.730 The main use for this is in template 02:23.780 --> 02:26.390 metaprogramming, again for checking invariants. 02:27.290 --> 02:32.810 If you have a template function, and your code assumes that one of the arguments can be copied, you 02:32.810 --> 02:36.650 can check this, by using static_assert() with a type trait. 02:40.440 --> 02:44.370 So let's quickly try that out. And we get a compiler error. 02:44.730 --> 02:47.370 "This requires a 64-bit compiler" 02:48.180 --> 02:54.660 And in fact, I am compiling in 32-bit mode. And in 32-bit mode, the size of int is 4. 02:55.290 --> 02:59.940 So this boolean is false, and we get this error message when the compiler stops. 03:02.140 --> 03:07.390 We can have template classes which have a default type parameter, or parameters, in the plural. 03:08.350 --> 03:09.730 We write them like this. 03:10.270 --> 03:14.410 So by default, the template parameter is type int. 03:16.710 --> 03:22.950 And then we can just instantiate this, like a normal template class. So we can put long double as the 03:22.950 --> 03:26.460 type parameter, and then this member will be a long double. 03:27.540 --> 03:33.390 If we leave the angle brackets empty, then we gets the default. And this member will be an int. 03:36.530 --> 03:43.190 So here is that class. We have a constructor, to initialize the member. And a function, which will print 03:43.190 --> 03:46.150 out the member. In the main function, 03:46.160 --> 03:49.630 we instantiate this with long double as the type parameter. 03:50.030 --> 03:51.290 So this will be long double. 03:52.370 --> 03:59.570 And then we instantiate again, using the default type parameter of int. 04:01.200 --> 04:02.600 And then we print out the member. 04:07.400 --> 04:13.790 So with the long double, we get 1.99999, and with the default we get 1, because the int has 04:13.790 --> 04:17.090 been truncated. And you might get a compiler warning about that. 04:19.860 --> 04:25.350 In C++11, it is now possible to have a template function with default parameters. 04:26.070 --> 04:27.480 It is the same syntax again. 04:27.810 --> 04:32.850 So by default, "t1" and "t2" will be const reference to int. 04:34.410 --> 04:39.030 And then we just call the template function as normal, and the compiler will deduce the types. 04:40.650 --> 04:45.060 So here we have this function, which will just print out the sum of its arguments. 04:46.140 --> 04:50.490 Then we can have long double and int variables, which we pass to the function. 04:51.390 --> 04:54.510 And then when we call, it the component will deduce it. 04:55.440 --> 04:59.970 And that may not seem very interesting, but in the next video, I will give you an example where this 04:59.970 --> 05:01.320 is actually rather useful. 05:02.340 --> 05:02.640 Okay. 05:02.640 --> 05:03.690 So that is it for this video. 05:04.110 --> 05:04.980 I will see you next time. 05:04.980 --> 05:07.470 But until then, keep coding!