WEBVTT 00:00.150 --> 00:00.660 Hello again! 00:00.990 --> 00:03.090 In this video, we are going to look at unions. 00:04.350 --> 00:07.500 Unions are inherited from the C programming language. 00:08.370 --> 00:10.740 A union is a compound data structure. 00:11.100 --> 00:14.910 It is similar to a struct in some ways, and different in some ways as well. 00:15.870 --> 00:18.630 Each member of a union must have a distinct type. 00:18.990 --> 00:21.390 We cannot have two members which have the same type. 00:22.500 --> 00:28.740 All the members of a union are stored at the same address in memory, and only one member of a union 00:28.950 --> 00:31.080 can be in use at any one time. 00:32.110 --> 00:33.220 So what does this mean? 00:34.900 --> 00:36.670 Let's try and visualize this. 00:37.150 --> 00:42.760 If we imagine we have a struct whose members are character, int and double, the character will be stored 00:42.760 --> 00:44.800 at the beginning of the struct memory. 00:45.550 --> 00:49.960 The int member will follow the character, and the double will come after the int. 00:50.080 --> 00:52.000 So they all come after each other in memory. 00:53.110 --> 00:56.020 The compiler may also put in some so-called "padding" bytes. 00:56.410 --> 00:57.820 These mean that the program 00:57.820 --> 01:02.050 will access the character value as a word and not as a byte. 01:02.320 --> 01:03.550 And that is much more efficient. 01:04.210 --> 01:07.840 There may also be some padding bytes here as well, depending on the architecture. 01:09.670 --> 01:15.400 If we have a union with the same members, then the character will again begin at the start of the union's 01:15.430 --> 01:15.850 memory. 01:16.660 --> 01:22.060 The int will also begin at the start of the union's memory, and its storage will overlap the character 01:22.060 --> 01:23.850 storage. And the double 01:23.870 --> 01:29.770 will also begin at the start of the union's memory and its storage will overlap the character and 01:29.770 --> 01:30.280 the nt. 01:32.420 --> 01:34.760 So you may think this is a trick for saving memory. 01:35.060 --> 01:40.340 In fact, the main application of unions is if you are processing data, which could be one of 01:40.340 --> 01:41.030 several types. 01:41.330 --> 01:46.100 And that comes in useful if you are parsing soure code, for example, or processing data that you have 01:46.100 --> 01:47.240 received over the network. 01:50.040 --> 01:52.260 All the members of a union are public by default. 01:52.380 --> 01:55.530 The same as a struct. In older versions of C++, 01:55.530 --> 02:01.530 the members of a union must be simple data types: no constructor, copy constructor, assignment 02:01.530 --> 02:02.880 operator or destructor. 02:03.930 --> 02:08.040 So that means, for example, that you cannot have a string as a union member. Which is pretty limiting, 02:08.400 --> 02:13.190 because strings come up all the time when you are processing data. In C++11, 02:13.230 --> 02:18.720 this was relaxed a bit, but it is a bit messy, because the compiler will not generate calls to things 02:18.720 --> 02:19.500 like destructors. 02:19.890 --> 02:23.070 So you have to call them yourself. And I am not going to go into that here. 02:24.450 --> 02:30.330 Unions can have member functions, but they cannot have virtual member functions. And you cannot use 02:30.330 --> 02:33.060 a union as a base or a derived class. 02:36.340 --> 02:37.840 To use a union, 02:37.840 --> 02:40.150 we start by assigning to one of the members. 02:40.540 --> 02:45.430 So we could assign to the character member for example, and that means that the character member is 02:45.430 --> 02:46.090 now in use. 02:46.720 --> 02:51.810 So the character member will have a well-defined value, and the int and the double will have undefined values. 02:53.110 --> 02:58.480 The only safe thing we could do is to read from the character value, or we could assign to one of the other 02:58.690 --> 02:59.170 members. 02:59.500 --> 03:04.060 And then, that member will come into play and the character member will have an undefined value. 03:06.120 --> 03:07.050 Let's have an example. 03:07.530 --> 03:09.010 So we have a union. 03:09.030 --> 03:11.170 The members are character, int and double. 03:12.420 --> 03:14.160 We create an object of this union. 03:14.580 --> 03:16.230 And then we assign to the character member. 03:16.560 --> 03:22.740 So the character member is now in use. And it has a well-defined value, which is the letter, capital 03:22.740 --> 03:23.040 'Z'. 03:24.270 --> 03:28.680 If we try to access the double member, this will have an undefined value. 03:29.580 --> 03:35.190 What will happen is that the program tries to interpret the member used by the union as a double. 03:35.970 --> 03:41.730 The first byte of this memory will contain the ASCII code for capital 'Z' and the rest of the bytes will 03:41.730 --> 03:43.230 contain - well, who knows? 03:44.100 --> 03:46.410 So we will probably get some kind of garbage number. 03:48.030 --> 03:49.080 Like that, for example. 03:51.750 --> 03:54.210 So this is fairly typical of the C legacy. 03:54.630 --> 03:55.380 It is low level. 03:55.890 --> 04:00.000 It is easy to make bad mistakes, and it is difficult to use safely. 04:00.660 --> 04:05.190 The main problem here is that the programmer is expected to know which type is in use. 04:05.940 --> 04:09.360 If you are working on a small problem on your own, that is not really a problem. 04:09.990 --> 04:14.910 If you have a large code base with several programmers working on it, then sooner or later there will be 04:14.910 --> 04:15.450 a mistake. 04:16.830 --> 04:21.030 One of the things you can do is to add a so-called "tag" member to the union. 04:21.450 --> 04:25.650 So this is an extra member, which will just keep track of which member is in use. 04:26.160 --> 04:27.930 And that is known as a tagged union. 04:30.070 --> 04:31.180 So let's try that out. 04:31.660 --> 04:37.480 We start off by creating an enumeration, with values for each possible allowed type. 04:37.480 --> 04:39.280 So we have character, int and double. 04:40.330 --> 04:43.270 Then we have an extra member, which is this token type. 04:44.500 --> 04:50.010 If we are going to bring the character member into use, then we need to set the token type to character. 04:50.920 --> 04:55.210 And then later on, if we want to use the double member, we can just check this. 04:55.570 --> 04:58.270 And if the token type is double, then we know it is safe to use this. 04:58.810 --> 05:01.870 And if it is not, then we have avoided something which is dangerous. 05:05.490 --> 05:08.130 And there are, we get a much more sensible result this time. 05:10.300 --> 05:15.080 But there is still a problem: we have to rely on the programmer who assigned to the character member. 05:15.340 --> 05:18.670 We have to assume that they also remembered to set the token type. 05:19.570 --> 05:22.350 We will come back to this in the next video. But 05:22.780 --> 05:24.220 meanwhile, that is all for this one. 05:24.610 --> 05:25.420 I will see you next time. 05:25.420 --> 05:27.700 But until then, keep coding!