WEBVTT 00:00.120 --> 00:04.440 Hello again! In this video, we are going to look at the subclasses of standard exception. 00:06.010 --> 00:11.800 So the immediate subclasses are bad_alloc, which is for memory allocation errors and bad_cast 00:11.800 --> 00:14.320 for when a dynamic cast fails. 00:15.280 --> 00:19.510 And then there are two sub-hierarchies, for logic errors and run-time errors. 00:22.710 --> 00:28.500 In the logic error hierarchy, we have out_of_range. So we have seen that with the at() member function of vector. 00:28.920 --> 00:35.700 This is thrown when we try to access an element outside the defined range. invalid_argument. 00:35.760 --> 00:40.770 So for example, stoi() takes a string and returns an int. 00:41.250 --> 00:47.160 But if the argument cannot be converted to a number, then the invalid_argument is thrown. 00:47.190 --> 00:50.010 So that is any case where the argument is not acceptable. 00:51.400 --> 00:52.150 domain_error. 00:52.480 --> 00:57.910 If we have something which takes a date, and we give some invalid date, like the 31st of February, 00:58.510 --> 01:01.390 then that is outside the domain of applicable values. 01:02.870 --> 01:09.170 And length_error. If we try to append too many elements to a string, then that will exceed the length limit of 01:09.170 --> 01:09.920 the object. 01:12.540 --> 01:19.200 For runtime_error, these are more numerical. overflow_error means that the result is too large for the 01:19.200 --> 01:19.650 variable. 01:20.010 --> 01:22.170 So there are too many digits to be stored. 01:23.240 --> 01:29.180 underflow_error means that the result of a floating-point computation is too small. There 01:29.180 --> 01:37.340 are too many zeros after the decimal points, before you get to the actual, non-zero digits. And then 01:37.340 --> 01:41.300 a range_error we have some intermediate result which cannot be represented. 01:44.250 --> 01:47.850 So you might think there are lots of places where the standard library could use exceptions. 01:49.560 --> 01:52.020 So here is an example with the square root function. 01:52.380 --> 01:54.840 This is from the header. 01:55.950 --> 02:00.810 It has a precondition, which is that the argument must be a positive number. 02:01.530 --> 02:04.710 So if we call it with -1, we violate the precondition 02:04.710 --> 02:11.370 and we get a logic error. And -1 is outside the range of acceptable values, so we should get a 02:11.420 --> 02:15.240 domain_error exception, and that should get caught by this catch block. 02:16.050 --> 02:17.130 So let's see what happens. 02:19.200 --> 02:28.410 And we get this weird-looking result. minus nan brackets ind. So obviously, we did not get an exception. 02:28.770 --> 02:29.970 Otherwise, it would have been caught. 02:30.600 --> 02:37.350 So the function has actually completed and returned. And the value that it returned is "not a number". 02:38.070 --> 02:41.310 So this is something from the floating-point implementation. 02:41.820 --> 02:47.220 If you get something which cannot be represented as a floating point number, like dividing by 0 02:47.220 --> 02:53.880 or getting the square root of -1, then the result is NaN. "Not a number". 02:57.730 --> 03:01.360 By default, the C++ streams do not throw exceptions. 03:01.960 --> 03:07.810 I think the argument is, that it is very easy to check the stream's state. And if you want an exception, 03:07.810 --> 03:08.890 you can throw one yourself. 03:09.550 --> 03:12.850 But in C++11, you can actually enable exceptions. 03:14.970 --> 03:20.580 To do that, you need to call the exceptions() member function for the object, for the stream object, and 03:20.580 --> 03:24.960 then you need to say which error conditions you want to cause an exception. 03:25.500 --> 03:30.900 So in this case, if fail_bit or bad_bit is set, "ifile" is going to throw an exception. 03:35.800 --> 03:41.680 So let's try that out. We need to #include to get these input/output exceptions. 03:42.760 --> 03:46.510 We create a stream object, an input file stream. 03:47.260 --> 03:54.010 Then we enable exceptions, like we did on the slide. And then we have a try block, where we have an operation 03:54.100 --> 04:00.610 that might throw an exception. If this "txt" file is not available. 04:00.610 --> 04:06.010 If it does not exist or we are not allowed to access it, then this will throw an exception which should 04:06.010 --> 04:06.940 get caught here. 04:07.570 --> 04:08.560 So let's see what happens. 04:10.320 --> 04:11.950 Right, so we do get the exception. 04:12.460 --> 04:16.150 So this has thrown an exception because the file does not actually exist. 04:17.500 --> 04:21.610 The message is maybe not very user friendly, but it is there. 04:24.710 --> 04:28.760 So why does the C++ library avoid using exceptions so much? 04:29.420 --> 04:31.250 Well, the answer really is efficiency. 04:31.940 --> 04:38.120 The library has to work with every kind of application on every kind of hardware that supports C++. 04:38.960 --> 04:43.640 And if you are going to be checking for exceptions and handling them, then that will require extra 04:43.640 --> 04:46.910 code, and sometimes that may not be acceptable. 04:47.960 --> 04:53.360 For example, the square root function is already quite slow compared to doing basic arithmetic operations. 04:54.200 --> 04:57.410 If you are doing numeric programming, or games, or multimedia. 04:57.920 --> 05:02.900 Then having a square root which checks and throws exceptions could add too much overhead. 05:04.190 --> 05:09.560 However, there are third party libraries which use exceptions, and of course, we are allowed to use 05:09.560 --> 05:10.730 them in our own code. 05:11.690 --> 05:13.220 Okay, so that is it for this video. 05:13.640 --> 05:16.880 I will see you next time, but until then, keep coding!