WEBVTT 00:00.180 --> 00:05.550 Hello again! In this video, we are going to look at how to get a random number in older versions of 00:05.550 --> 00:12.240 C++. C has a function called rand(), which C++ has inherited. 00:12.870 --> 00:19.410 This is a pseudo-random number generator, so it will generate a sequence of numbers which look random. 00:19.980 --> 00:23.220 And each time we call rand(), we get the next number from the sequence. 00:24.090 --> 00:27.870 If we get to the end of the sequence, then it just goes back to the beginning, and starts again. 00:28.710 --> 00:35.760 The return value is an integer between 0 and RAND_MAX. And RAND_MAX is chosen by the implementation, 00:36.120 --> 00:37.560 so there is no standard. 00:38.940 --> 00:39.850 So this is what we do, 00:40.020 --> 00:49.620 to get a random number. And your random number today is 41. And that is OK if you want some random number between 0 00:49.620 --> 00:54.900 and RAND_MAX, whatever that is. But you may want a different range of values. 00:55.650 --> 00:59.430 And in that case, you have to do some arithmetic and rescale the result. 01:00.180 --> 01:06.180 For example, if we want a floating point number between 0 and 1, we divide the return value 01:06.180 --> 01:07.230 by RAND_MAX. 01:08.040 --> 01:13.650 But we have to be careful, because these are both integers. So that will be an integer division. So we only 01:13.650 --> 01:14.760 get 0 or 1. 01:15.900 --> 01:21.180 So we need to make sure we do a floating-point division, and to do that, we convert the result of 01:21.180 --> 01:23.280 rand() to a floating point number. 01:25.140 --> 01:30.240 If we want the range to start from a different value, then we have to add that, after re-scaling. 01:30.870 --> 01:34.860 So if we want integers between 1 and 100, we have to re-scale, 01:35.280 --> 01:38.910 to get integers between 0 and 99, and then add 1. 01:41.280 --> 01:42.630 So let's try that out. 01:44.370 --> 01:48.090 So that is a floating-point number between 0 and 1. 01:48.540 --> 01:51.810 And then we have 10 integers between 1 and 100. 01:53.280 --> 01:56.670 And if I run it again, we get the same numbers. 02:01.750 --> 02:05.260 If we want different random numbers, we have to "seed" the generator. 02:05.830 --> 02:07.630 We do that by calling srand(). 02:08.440 --> 02:10.180 We pass the seed number as 02:10.180 --> 02:16.000 the argument. And the traditional practice here is to pass the current time, because what could be more 02:16.000 --> 02:16.930 random than the time? 02:17.410 --> 02:18.610 There is no way of guessing that, is there (!?) 02:18.610 --> 02:21.700 There is another C function, called time(). 02:22.270 --> 02:27.040 If we call this with argument 0, it will return the current time, as a very large integer. 02:27.430 --> 02:33.040 The number of seconds since 1970. And then that can be used to seed the random number generator. 02:34.540 --> 02:39.250 In addition to the obvious problem, the value returned by time only changes once per second. 02:40.690 --> 02:44.800 So if you have another program which starts up at the same time, it will have the same seat and it is 02:44.800 --> 02:46.690 going to generate the same sequence of random numbers. 02:47.050 --> 02:50.650 And this does make it a bit easier for people who are trying to guess the sequence. 02:52.390 --> 02:57.550 So here is the same code again, with our call to srand(), using the current time as the seed. 02:59.420 --> 03:03.410 So if I wait one second, I should get different numbers. 03:05.810 --> 03:09.200 I do not think I can type this fast enough to do it twice in one second! 03:09.350 --> 03:12.800 Maybe you could do that from the command line, actually. Something for you to try out! 03:15.300 --> 03:20.850 OK, so that all seems pretty straightforward, so why does it need to change? If you are just writing 03:20.880 --> 03:23.850 "toy" programs, then this approach is good enough. 03:24.630 --> 03:27.780 If you are serious about random numbers, then there are quite a few issues. 03:29.040 --> 03:33.310 The first one is this the implementer can use any algorithm they want. 03:33.330 --> 03:37.380 There is no standard. And apparently there have been some pretty shocking ones in the past. 03:38.550 --> 03:41.910 The present implementations are probably okay, but not great. 03:43.650 --> 03:48.900 There's a common tendency to generate numbers which are not particularly random. Especially if you look 03:48.900 --> 03:51.720 at the last few digits. And quite a few people do that. 03:52.230 --> 03:57.360 So she gets a random number between 0 and 1000, say, they just chop off everything except the last 03:57.360 --> 03:58.050 few digits. 03:58.410 --> 03:59.880 That is not really a very good approach. 04:00.450 --> 04:06.690 If we do not want numbers between 0 and RAND_MAX, we have to re-scale the range, and most implementations 04:06.690 --> 04:08.070 are not very robust to this. 04:08.670 --> 04:14.430 If you re-scale the range, it introduces what is called "bias", and that means that some numbers are more 04:14.430 --> 04:15.240 likely than others. 04:15.870 --> 04:18.870 So the results are not quite as random as they should be. 04:22.610 --> 04:27.680 And finally, the sequence is too short, and it repeats too often, and guessing the next number is not 04:27.680 --> 04:28.640 as hard as it should be. 04:29.030 --> 04:31.490 So this is really not good enough for cryptography. 04:32.450 --> 04:34.130 Okay, so that is it for this video. 04:34.520 --> 04:35.390 I will see you next time. 04:35.630 --> 04:37.490 Until then, keep coding!