WEBVTT 1 00:00:01.040 --> 00:00:03.359 M hm hm. 2 00:00:06.400 --> 00:00:09.359 Hello and welcome to another episode of Against and Donnett. 3 00:00:09.439 --> 00:00:11.800 I'm Jean Clavery, your host, and with me today is 4 00:00:11.839 --> 00:00:15.400 our special guest step On Shrines. Welcome, step On, Hey, 5 00:00:15.560 --> 00:00:19.480 thanks for having me. Hey, thanks for coming. So I 6 00:00:19.519 --> 00:00:22.679 think we're going to talk about source code generators today, 7 00:00:23.039 --> 00:00:25.079 but before we get to that, why don't you tell 8 00:00:25.160 --> 00:00:27.559 us a little bit about yourself, how you got into developed, 9 00:00:27.839 --> 00:00:30.719 into development, and what kind of thing do you work on. 10 00:00:30.719 --> 00:00:33.960 Now As of now, I'm a software developer at Team Sure. 11 00:00:34.200 --> 00:00:36.880 Basically full stack developer working on dot net in the 12 00:00:36.920 --> 00:00:40.719 back end and also the regular react type stript stack 13 00:00:40.799 --> 00:00:44.280 in the front end. Basically I got into development kind 14 00:00:44.280 --> 00:00:48.159 of the stayd ut rod so through university, and I 15 00:00:48.200 --> 00:00:50.439 also had a tech background in my family. So my 16 00:00:50.960 --> 00:00:54.119 dad is to be a system administrator at a local bank, 17 00:00:54.200 --> 00:00:57.600 and also my mom is a project manager at the 18 00:00:57.759 --> 00:01:00.960 at the same bank, so kind of already a pretty 19 00:01:01.000 --> 00:01:03.439 fine puth and so I got into it basically the 20 00:01:04.840 --> 00:01:05.519 classic crowd. 21 00:01:05.560 --> 00:01:08.400 I would say, So some people have doctors in the family, 22 00:01:08.439 --> 00:01:10.680 and you're just a family of developers. 23 00:01:10.719 --> 00:01:14.519 I guess yeah. Basically, Basically I can remember my dad 24 00:01:14.799 --> 00:01:17.400 trying to teach me C plus plus and pointers at 25 00:01:17.560 --> 00:01:20.719 thirteen didn't quite work out, but I still managed my 26 00:01:20.799 --> 00:01:21.400 way somehow. 27 00:01:21.879 --> 00:01:24.480 Nice. So how'd you get into dot net? 28 00:01:24.599 --> 00:01:26.959 Yeah, it was basically a bit of a wide ride 29 00:01:27.079 --> 00:01:30.319 until I got there. So through university. I went to 30 00:01:30.879 --> 00:01:34.120 university in a program which is common in Germany, where 31 00:01:34.120 --> 00:01:37.439 you spent three months in university and three months at 32 00:01:37.640 --> 00:01:41.280 a company, just trying to get some practical experience. And 33 00:01:41.400 --> 00:01:44.400 basically I went through the whole cycle of C plus 34 00:01:44.400 --> 00:01:48.280 plus phyphon, then your front and developer and some show. 35 00:01:48.560 --> 00:01:51.079 But in the end, when I finally picked up my 36 00:01:52.200 --> 00:01:55.799 full position, then I got into a net aspn net 37 00:01:55.799 --> 00:01:59.120 team and that's kind of where I spent my time 38 00:01:59.159 --> 00:02:01.640 so far. Really really like it all. 39 00:02:01.560 --> 00:02:05.760 Right, cool, So let's talk about source code generators. Kind 40 00:02:05.799 --> 00:02:08.199 of have to go back to the beginning of them, 41 00:02:08.400 --> 00:02:12.120 just kind of give an introduction of what they are 42 00:02:12.159 --> 00:02:14.360 and kind of what they're used for is because you know, 43 00:02:14.400 --> 00:02:16.800 I've I've listened to other people talk about it, but 44 00:02:16.879 --> 00:02:18.319 they're still very confusing to me. 45 00:02:18.599 --> 00:02:24.120 Basically, source generators in different I just it's basically a 46 00:02:24.199 --> 00:02:26.879 new API which came in with Dotnet five and c 47 00:02:27.039 --> 00:02:30.479 sharp nine. So you're probably familiar with a lot of 48 00:02:30.520 --> 00:02:33.960 approaches which we'll use ahead of this technology in order 49 00:02:34.000 --> 00:02:37.039 to generate source code. So there was a T four 50 00:02:37.280 --> 00:02:41.520 so text templating, and then maybe even even earlier we 51 00:02:41.800 --> 00:02:44.520 had to set up of having post built and rebuilt 52 00:02:44.520 --> 00:02:48.199 steps just trying to generate some some sauce which you 53 00:02:48.240 --> 00:02:51.400 can then use within your code. And basically with this 54 00:02:51.520 --> 00:02:54.520 new API, it's now possible to actually hook into the 55 00:02:54.560 --> 00:02:58.120 compilation cycle of your of your project and kind of 56 00:02:58.360 --> 00:03:01.680 also get everything set up without leaving any artifacts, so 57 00:03:01.800 --> 00:03:06.360 without fluttering your directory with new files. And actually you 58 00:03:06.400 --> 00:03:10.000 can also take advantage of a new thing with these 59 00:03:10.039 --> 00:03:13.240 sauce generators, so you can basically hook into the compilation 60 00:03:13.400 --> 00:03:17.159 and you actually get the entire past syntax three of 61 00:03:17.280 --> 00:03:21.080 your compilation, and you're able to inspect everything in there 62 00:03:21.159 --> 00:03:23.159 and generate code accordingly. 63 00:03:23.479 --> 00:03:25.879 So is it kind of like the T four templates 64 00:03:26.000 --> 00:03:29.080 it's something that you use during your development that you're 65 00:03:29.159 --> 00:03:31.439 going to spit out some final code for you to 66 00:03:31.439 --> 00:03:33.840 put into your project, or is it something that actually 67 00:03:33.919 --> 00:03:34.879 can go at runtime. 68 00:03:35.080 --> 00:03:38.400 It's somewhere in between. So T four, as you already mentioned, 69 00:03:38.439 --> 00:03:41.039 it would actually split out code, but you would usually 70 00:03:41.039 --> 00:03:45.199 get it in actual cs files depending on what you generate. 71 00:03:45.800 --> 00:03:48.919 It's also not quite at one time, it's in between, 72 00:03:49.080 --> 00:03:52.680 so it's actually happening in the compilation step. So when 73 00:03:52.719 --> 00:03:55.120 you would go ahead and start up your compilation cycle, 74 00:03:55.680 --> 00:04:00.479 you would basically start passing your entire code and then 75 00:04:00.599 --> 00:04:03.400 at this point the source generator will kick in. You 76 00:04:03.439 --> 00:04:06.719 will basically push your code at that point in time 77 00:04:06.759 --> 00:04:09.280 so that the past artifact, you would push it to 78 00:04:09.280 --> 00:04:12.520 the source generator. The source generator can then analyze it, 79 00:04:12.680 --> 00:04:15.199 generate new source code, and this new source code you 80 00:04:15.319 --> 00:04:18.959 generate will then go straight back into the compilation and 81 00:04:19.000 --> 00:04:21.800 then the compilation will resume and at the end you 82 00:04:21.839 --> 00:04:25.639 will obviously get your IL. So you're actually within the 83 00:04:25.639 --> 00:04:30.319 compilation cycle, so it's not during other events before or 84 00:04:30.360 --> 00:04:32.519 after compilation, but straight within. 85 00:04:32.839 --> 00:04:35.360 So it's not going to kick out a CS file 86 00:04:35.439 --> 00:04:38.240 for you. It's something that kicks out for the run 87 00:04:38.319 --> 00:04:40.319 time to pick up and then go from there when 88 00:04:40.319 --> 00:04:41.160 you actually run it. 89 00:04:41.160 --> 00:04:44.839 It's possible to also emit it at a CSFI, but 90 00:04:44.920 --> 00:04:47.439 basically it will just be there at run time. The 91 00:04:47.519 --> 00:04:51.560 tooling is i would say a bit somewhere in between 92 00:04:51.680 --> 00:04:55.399 right now. So if you run this generator once you 93 00:04:55.439 --> 00:04:58.600 would usually also get intellicent support within visual Studio for 94 00:04:58.720 --> 00:05:01.199 this because at that point and a visual studio can 95 00:05:01.240 --> 00:05:04.680 also pick up the generated generated code and you can 96 00:05:04.720 --> 00:05:09.000 also use it within your regular development cycle get intelligence 97 00:05:09.079 --> 00:05:12.279 and support and whatnot. But you will not have it 98 00:05:12.360 --> 00:05:17.079 on disc by default, so it has some varying pooling support. 99 00:05:17.160 --> 00:05:20.199 And then it tends to feel like it's actually a 100 00:05:20.199 --> 00:05:22.439 source file within your project, but it's not. 101 00:05:22.920 --> 00:05:25.680 So what kind of be a basic use for it? 102 00:05:25.759 --> 00:05:27.879 You know, with with like T four templates and things 103 00:05:27.879 --> 00:05:30.680 like that, you're going to you build it and you 104 00:05:31.199 --> 00:05:33.199 run it, and then quite often you have to go 105 00:05:33.279 --> 00:05:35.480 into the generated code and make some changes to be 106 00:05:35.839 --> 00:05:38.160 you know exactly what you want. You know, if this 107 00:05:38.240 --> 00:05:39.800 is going to run a compile time, and I guess 108 00:05:39.839 --> 00:05:42.959 you really have to get things exactly how you want 109 00:05:42.959 --> 00:05:45.000 it before you run that compilation. 110 00:05:45.920 --> 00:05:50.480 Very varying increase of how you can use this. So 111 00:05:51.040 --> 00:05:54.439 for me personally, when I experimented with it, I kind 112 00:05:54.439 --> 00:05:57.199 of set up a code based in which I had 113 00:05:57.240 --> 00:06:01.120 for example, a specific interface which is inherited by some 114 00:06:01.199 --> 00:06:05.439 example classes, and then for example, during compilation, I could 115 00:06:05.519 --> 00:06:08.879 use the source generator and create a decorator for all 116 00:06:08.920 --> 00:06:12.879 of these classes implementing this interface, which would then maybe 117 00:06:13.000 --> 00:06:17.480 maybe log out something or to some additional things which 118 00:06:17.519 --> 00:06:20.680 might be common for all of these implementers. So that's 119 00:06:20.959 --> 00:06:24.879 maybe one use case. You can generate decorators and other 120 00:06:24.959 --> 00:06:28.240 things like that, basically boring stuff you don't want to 121 00:06:28.240 --> 00:06:32.199 write everywhere. Some people also use this to generate validation 122 00:06:32.360 --> 00:06:36.439 code on the fly based on some metadata, so for example, 123 00:06:36.480 --> 00:06:38.920 you put some attributes on your classes or on your 124 00:06:38.920 --> 00:06:43.079 methods and you can generate a validation on the fly. 125 00:06:43.399 --> 00:06:48.120 I think it's also really used a lot already within 126 00:06:48.240 --> 00:06:52.399 aspn net, and some dependency injection frameworks already picked us 127 00:06:52.480 --> 00:06:56.120 up because being able to inspect the compilation three on 128 00:06:56.720 --> 00:07:00.480 the fly, you can, for example, as a dependency injection framework, 129 00:07:00.519 --> 00:07:03.839 instead of relying on reflection at runtime, you can also 130 00:07:04.120 --> 00:07:07.879 take a look at the code during the compilation and 131 00:07:07.959 --> 00:07:11.720 basically see which losses need which dependency, and instead of 132 00:07:11.759 --> 00:07:15.360 generating your dependencies at runtime, you can then already make 133 00:07:15.439 --> 00:07:17.959 use of it and have a look at this and 134 00:07:18.120 --> 00:07:19.839 during the compilation cycle. 135 00:07:19.959 --> 00:07:23.680 Okay, so what's the first thing to know to get started? 136 00:07:23.759 --> 00:07:26.959 You know, writing your first source code generator but there. 137 00:07:26.800 --> 00:07:30.240 Are basically some constraints right now. The biggest one is 138 00:07:30.279 --> 00:07:34.000 probably that the project containing the Sauce generator needs to 139 00:07:34.000 --> 00:07:37.439 be Donut Standard two dot zero, so it's not yet 140 00:07:37.480 --> 00:07:41.519 on a Net five. That's basically a constraint of how 141 00:07:41.560 --> 00:07:46.399 the Rosslyn compiler, which the source generators originally come from, 142 00:07:46.959 --> 00:07:49.560 is allowed to run. And then basically you need a 143 00:07:49.600 --> 00:07:53.360 set up containing of one project which hosts the generator, 144 00:07:53.639 --> 00:07:55.639 and then you will probably want to consume it from 145 00:07:55.680 --> 00:08:00.879 another project. So within a project containing the generator, you 146 00:08:00.959 --> 00:08:04.879 basically need to import two packages, which is the Microsoft 147 00:08:04.959 --> 00:08:09.240 COUTE Analysis c Shop and Microsoft COUDE Analysis Analyzer. And 148 00:08:09.279 --> 00:08:12.160 then within the project you want to reference it, you 149 00:08:12.199 --> 00:08:15.839 would have the Microsoft dot net dot Compilers dot tool set, 150 00:08:16.079 --> 00:08:19.680 a package that will also bring in some intelligence magic. 151 00:08:20.040 --> 00:08:24.000 And then basically within the reference to this source generator, 152 00:08:24.439 --> 00:08:27.720 you would also have to put some specific text like 153 00:08:27.920 --> 00:08:32.039 output item type analyzer, and it's basically a bit of 154 00:08:32.080 --> 00:08:36.720 a setup, but that's really much standard. And then once 155 00:08:36.759 --> 00:08:39.840 you do that, you can basically just create a glass, 156 00:08:40.039 --> 00:08:43.120 put a generator attribute on top of it and make 157 00:08:43.159 --> 00:08:46.639 it inherit Thesaurce Generator interface and that's basically it it 158 00:08:46.639 --> 00:08:48.919 sounds a bit more complicated than it actually is in 159 00:08:48.960 --> 00:08:52.120 the end, so it's actually really easy just taking a 160 00:08:52.159 --> 00:08:55.200 look at the dogs and then getting everything set up 161 00:08:55.639 --> 00:08:58.639 and then actually it works quite well out of the box. 162 00:08:58.799 --> 00:09:00.279 So does it take writing a lot a lot of 163 00:09:00.320 --> 00:09:04.159 code to get something that's going to generate code for you? 164 00:09:04.240 --> 00:09:06.679 And you know sometimes you know you might be spending 165 00:09:06.720 --> 00:09:09.720 more time writing the generator than what you spent writing 166 00:09:09.759 --> 00:09:10.480 the code itself. 167 00:09:10.720 --> 00:09:13.840 It depends on the usage, I would say. So basically 168 00:09:13.840 --> 00:09:17.240 a source generator it always works in two steps. So 169 00:09:17.360 --> 00:09:20.879 there's an initialized method, and within this method you can, 170 00:09:20.919 --> 00:09:27.200 for example, register your code to respond to syntax node notifications. 171 00:09:27.240 --> 00:09:30.279 So whenever the compiler picks up something new, you would 172 00:09:30.639 --> 00:09:33.639 have this event pushed to your sauce generator and you 173 00:09:33.679 --> 00:09:37.399 can maybe already analyze all the notes whether they fulfill 174 00:09:37.519 --> 00:09:40.080 some constraint. But if you don't need this, and you 175 00:09:40.240 --> 00:09:43.720 only decide okay, I just want to generate code, that 176 00:09:43.919 --> 00:09:47.039 usually happens in the execute step of the generator, and 177 00:09:47.159 --> 00:09:50.240 within there, basically all you have to do is, just 178 00:09:50.480 --> 00:09:54.679 as with regular code generation, create a string builder and 179 00:09:55.200 --> 00:09:57.879 there you can write your C sharp code and that's 180 00:09:57.919 --> 00:10:00.039 it already. So if you don't need any super a 181 00:10:00.320 --> 00:10:03.480 special setup, you can just have your string builder create 182 00:10:03.519 --> 00:10:06.399 a c shop class and you're good to go already. 183 00:10:06.639 --> 00:10:08.960 Okay, so is it kind of like using a string builder. 184 00:10:09.519 --> 00:10:11.919 You pass in some tokens and things like that you 185 00:10:11.960 --> 00:10:15.120 want to be replaced in that string as it compiles, 186 00:10:15.159 --> 00:10:17.159 and then that'll kick out the code for. 187 00:10:17.159 --> 00:10:23.320 You, essentially usually a very clorified string interpulation, and that's 188 00:10:23.360 --> 00:10:26.279 basically all there. So that's what you would mostly also 189 00:10:26.399 --> 00:10:26.879 end up with. 190 00:10:27.240 --> 00:10:29.200 Okay, Yeah, that makes a lot more sense to me 191 00:10:29.279 --> 00:10:30.759 now that I think about it that way. It's just 192 00:10:30.799 --> 00:10:34.159 something that's going to basically take a snippet or templative 193 00:10:34.200 --> 00:10:38.120 code that you have, and then I guess somewhere in 194 00:10:38.159 --> 00:10:40.720 your main code you put you know, like you said, 195 00:10:40.799 --> 00:10:44.679 you're use an attribute or something like that to set it, 196 00:10:44.759 --> 00:10:47.440 and then in that attribute you pass in the things 197 00:10:47.440 --> 00:10:49.759 that are going to replace get replaced in that string builder. 198 00:10:50.000 --> 00:10:53.600 That's actually the beauty of it. You can basically put 199 00:10:54.039 --> 00:10:57.879 anything anywhere, and since you get the complete syntax tree 200 00:10:58.000 --> 00:11:00.799 of your program, you can analyze every thing, so it 201 00:11:00.840 --> 00:11:03.399 doesn't even have to be an attribute. You can also 202 00:11:03.559 --> 00:11:08.960 inspect interfaces, check whether interfaces fulfill a specific name or 203 00:11:08.960 --> 00:11:12.879 whether they also implement other interfaces on top. Usually an 204 00:11:13.000 --> 00:11:15.919 attribute is a good idea, I would say, because that's 205 00:11:16.320 --> 00:11:19.720 especially designed to represent some metadata. So that's a good 206 00:11:19.759 --> 00:11:23.600 approach in my opinion. But you're definitely not constrained to attributes. 207 00:11:23.679 --> 00:11:26.840 You can mix and match as you want. So basically 208 00:11:26.879 --> 00:11:27.080 you have. 209 00:11:27.200 --> 00:11:30.559 It build a class. That class doesn't exist at the 210 00:11:30.559 --> 00:11:32.879 time that you're actually doing your development. So how does 211 00:11:32.919 --> 00:11:35.759 the other part of your code know that that's going 212 00:11:35.840 --> 00:11:37.440 to be there when you compile. 213 00:11:37.799 --> 00:11:41.320 That's actually a bit difficult now, So I think I 214 00:11:41.360 --> 00:11:44.519 already mentioned that the tooling support can feel a bit 215 00:11:44.600 --> 00:11:48.919 off here and there. Usually it might take one cycle 216 00:11:49.039 --> 00:11:52.919 of building the source generator, for example visual studio to 217 00:11:52.960 --> 00:11:55.519 pick up the generated files at least, and then at 218 00:11:55.559 --> 00:11:58.039 least visual Studio can give you a bit of a 219 00:11:58.120 --> 00:12:01.360 hand with the generated code and also give you some 220 00:12:01.440 --> 00:12:05.039 intelligence on top. But sometimes, for example, if you create 221 00:12:05.080 --> 00:12:07.399 a new generator and you never had it run before, 222 00:12:07.600 --> 00:12:11.399 then it can feel like you basically want to use 223 00:12:11.399 --> 00:12:14.840 that glass, but intelligence just acts as if this would 224 00:12:14.879 --> 00:12:17.879 not exist, which it actually doesn't at disappoint So yes, 225 00:12:18.200 --> 00:12:20.799 tooling is a bit off, but I hope the visional 226 00:12:20.840 --> 00:12:25.080 studio experience, and I personally don't know how it behaves 227 00:12:25.080 --> 00:12:27.480 in writer, but I guess they would also have picked 228 00:12:27.480 --> 00:12:30.600 it up by now. So that part is usually entirely 229 00:12:30.600 --> 00:12:33.399 on the ide and depending on the support, it can 230 00:12:33.480 --> 00:12:37.159 feel good or mediocre, but usually you can get by 231 00:12:37.320 --> 00:12:38.320 really pretty fine. 232 00:12:38.320 --> 00:12:40.759 I would say so in that in that string builder, 233 00:12:40.799 --> 00:12:43.279 you know, as as you're building that telling it what 234 00:12:43.600 --> 00:12:46.799 the code you want to output, can it see into 235 00:12:46.879 --> 00:12:50.000 that code that you're building so you can get intelligence 236 00:12:50.039 --> 00:12:51.840 and code validation and things like that. 237 00:12:52.120 --> 00:12:55.039 Not entirely sure, to be honest, I think there's also 238 00:12:55.080 --> 00:12:56.879 some tooling in that direction. 239 00:12:57.080 --> 00:12:59.399 I mean, can you do something like like a razor file, 240 00:12:59.600 --> 00:13:02.720 but be be the template of the code that you 241 00:13:02.759 --> 00:13:03.360 want to use? 242 00:13:04.519 --> 00:13:06.720 I think I read up a bit. There are some 243 00:13:06.879 --> 00:13:10.399 extensions which also look into solving that problem, but I 244 00:13:10.440 --> 00:13:12.840 think out of the boxes is sadly not possible. Not 245 00:13:13.120 --> 00:13:15.919 so you would kind of have to know where your 246 00:13:16.039 --> 00:13:19.559 classes are because in the end, when you create a 247 00:13:19.559 --> 00:13:22.840 string which should then contain like a complete class, you 248 00:13:22.840 --> 00:13:26.360 would also have to take care of using namespaces and 249 00:13:26.559 --> 00:13:31.320 everything around that. So that's definitely something which feels a 250 00:13:31.320 --> 00:13:33.960 bit difficult sometimes because you'll have to remember all of 251 00:13:34.000 --> 00:13:36.519 these things, all of these boiler plate things you would 252 00:13:36.519 --> 00:13:37.720 also have to add on top. 253 00:13:37.919 --> 00:13:40.200 Yeah, because if it did something like in a razor 254 00:13:40.240 --> 00:13:42.200 file or something like that, where you could just do 255 00:13:42.320 --> 00:13:45.440 all your loops right within the little template to build 256 00:13:45.440 --> 00:13:48.000 it all up as the final string that you're trying 257 00:13:48.039 --> 00:13:50.240 to replace into your code, that would make it pretty 258 00:13:50.240 --> 00:13:51.639 easy to work with. I think. 259 00:13:51.919 --> 00:13:55.559 In on that end, you can basically also just push 260 00:13:55.960 --> 00:13:58.759 a part of your class to the string builder, and 261 00:13:58.799 --> 00:14:01.200 then you can fall back to the regularly shop code 262 00:14:01.440 --> 00:14:04.000 and just for example, inside a loop which would then 263 00:14:04.639 --> 00:14:08.320 continuously append lines to the to the string builder, you're 264 00:14:08.320 --> 00:14:10.840 also not constrained to the string builder at all. You 265 00:14:10.879 --> 00:14:14.320 can also just use a very big interpolatest string with 266 00:14:14.320 --> 00:14:16.559 with the ad at the beginning to make it work 267 00:14:16.559 --> 00:14:21.200 across lines, and you're really free to model that however 268 00:14:21.279 --> 00:14:21.960 you want. 269 00:14:22.200 --> 00:14:25.440 Do you find yourself creating these source code generators a 270 00:14:25.480 --> 00:14:27.799 lot for different projects? You know, I'm trying to I'm 271 00:14:27.799 --> 00:14:29.519 trying to take up a case where you know, I 272 00:14:29.600 --> 00:14:34.840 really needed that that much other than like generating entities 273 00:14:34.879 --> 00:14:37.519 and things like that, where you know, Visus code Visus 274 00:14:37.519 --> 00:14:40.159 Studio is already going to help you out building those, 275 00:14:40.679 --> 00:14:43.159 just like a T four templeate does. So when would 276 00:14:43.200 --> 00:14:45.639 you use a source code generator and when would you 277 00:14:45.720 --> 00:14:48.159 use something right like a T four template. 278 00:14:48.399 --> 00:14:52.200 I think there's a very big distinction. With T four templates, 279 00:14:52.240 --> 00:14:55.639 you're usually constrained to Visual Studio because I think they're 280 00:14:56.240 --> 00:14:59.559 much tied to that I d e. And with the 281 00:14:59.600 --> 00:15:03.960 software generators, you're actually running within Rosslyn, so within a compiler, 282 00:15:04.039 --> 00:15:06.919 so you can also have it running on Linux or 283 00:15:06.960 --> 00:15:09.840 wherever you want, So that's a big difference. I would 284 00:15:09.840 --> 00:15:12.879 say you're very much free to use it wherever you want. 285 00:15:13.039 --> 00:15:15.360 When it comes to actual examples where I used this, 286 00:15:15.519 --> 00:15:18.159 So actually I had a use case for this a 287 00:15:18.159 --> 00:15:21.480 while ago. So at work, we have an older project. 288 00:15:21.480 --> 00:15:24.960 It's C plus plus cli. Maybe that rings a bell. 289 00:15:25.279 --> 00:15:28.440 It's it's a bit engined, and basically we have some 290 00:15:28.679 --> 00:15:32.080 definitions in there which we wanted to translate to completely 291 00:15:32.120 --> 00:15:36.519 manage the code models. And basically I also got started 292 00:15:36.559 --> 00:15:39.679 with this by using it as as a post built step, 293 00:15:39.759 --> 00:15:43.759 you know, the old wave generating code. But I mean 294 00:15:44.279 --> 00:15:48.279 in my case, it was pretty simple colleagues started complaining 295 00:15:48.320 --> 00:15:50.720 at some point that I was just generating way too 296 00:15:50.720 --> 00:15:54.120 many files with that, because I created a file for 297 00:15:54.399 --> 00:15:56.399 definitions and there are quite a lot of them in 298 00:15:56.440 --> 00:15:59.399 that project. And then I thought, okay, this might be 299 00:15:59.399 --> 00:16:02.279 a good use is for a source generator, and then 300 00:16:02.519 --> 00:16:06.399 I did that, and it even worked flawlessly, even if 301 00:16:06.639 --> 00:16:09.360 in a dot Net framework code base, so not even 302 00:16:09.399 --> 00:16:12.519 if dont five. That's probably more of a convenience thing 303 00:16:12.840 --> 00:16:16.559 because there's like no technical reason to have a source 304 00:16:16.600 --> 00:16:19.480 generator there. But I think I think I already talked 305 00:16:19.480 --> 00:16:23.200 about it a bit. There are some use cases, especially 306 00:16:23.240 --> 00:16:27.200 within dot net frameworks, which are very much focused on performance, 307 00:16:27.200 --> 00:16:30.120 so ASPNAT is a prime use case for this. They 308 00:16:30.159 --> 00:16:32.960 already tried to move a lot of the code which 309 00:16:32.960 --> 00:16:36.320 they would usually use reflection for at run time and 310 00:16:36.720 --> 00:16:39.440 try to convert it to a source generator, especially if 311 00:16:39.440 --> 00:16:44.399 the information they would try to process through reflection is 312 00:16:44.399 --> 00:16:47.679 already available at compilation time, just to shape of that 313 00:16:47.919 --> 00:16:51.120 extra bit of performance during one time. I think that's 314 00:16:51.159 --> 00:16:53.960 a pretty big use case where this is already used 315 00:16:54.000 --> 00:16:54.279 a lot. 316 00:16:54.639 --> 00:16:57.200 So I think people like me, being kind of being 317 00:16:57.320 --> 00:16:59.799 old school, the first thing I would go to is 318 00:16:59.840 --> 00:17:02.720 just make make a copy of something that I want 319 00:17:02.759 --> 00:17:05.079 and then I just copy and paste to make all 320 00:17:05.119 --> 00:17:07.279 the little changes and things like that, rather than going 321 00:17:07.319 --> 00:17:11.079 to lassole writing a source code generator. Do you really 322 00:17:11.160 --> 00:17:13.599 have to get something that really has to be used 323 00:17:13.640 --> 00:17:16.160 over and over again? And can you can you write 324 00:17:16.200 --> 00:17:19.480 this generator and then share it between different projects. 325 00:17:19.640 --> 00:17:23.079 That's actually possible. So it's a bit difficult because I 326 00:17:23.079 --> 00:17:26.000 think I already mentioned that source code generators are kind 327 00:17:26.039 --> 00:17:28.920 of constrained to dot Net standard to zero right now. 328 00:17:29.359 --> 00:17:32.599 So that was definitely a bit of a surprise when 329 00:17:32.640 --> 00:17:36.720 I started getting into that technology, because, as with any 330 00:17:37.039 --> 00:17:41.079 good citizen, I already upgraded my projects to dot Net 331 00:17:41.079 --> 00:17:44.240 five at disappoint and I basically wanted to have the 332 00:17:44.680 --> 00:17:48.599 generator use some some specific names of some types, and 333 00:17:48.640 --> 00:17:51.000 then I noticed you can't reference it from a dot 334 00:17:51.039 --> 00:17:54.000 net standard project. You can't add a reference to dot 335 00:17:54.000 --> 00:17:56.079 net five. So that was a bit of a shocker. 336 00:17:56.279 --> 00:17:58.559 So that's a constraint do you have to keep in 337 00:17:58.599 --> 00:18:01.880 mind when you try to share these source generators, But 338 00:18:02.000 --> 00:18:05.920 otherwise it's it's definitely possible, and if you already use 339 00:18:06.000 --> 00:18:09.680 the new cspproach format. So the one coming in with 340 00:18:09.799 --> 00:18:13.559 dot net standard, you would also get this source generator 341 00:18:13.559 --> 00:18:17.519 output as a transitive dependency. So, for example, if you 342 00:18:17.599 --> 00:18:22.680 have a source generator project which is referenced by another project, 343 00:18:22.720 --> 00:18:25.160 and then you would transitively add a reference to that 344 00:18:25.240 --> 00:18:30.119 project from somewhere else, you would get these generated definitions everywhere. 345 00:18:30.279 --> 00:18:32.319 And that's already a nice way to share it. But 346 00:18:32.440 --> 00:18:35.880 besides that, you're also very much free to reference it 347 00:18:35.920 --> 00:18:38.720 from somewhere else. That's just one thing to keep in 348 00:18:38.759 --> 00:18:42.319 mind because if you reference it from multiple positions in 349 00:18:42.359 --> 00:18:45.519 your code, and you reference it in a way which 350 00:18:45.680 --> 00:18:48.960 triggers the source generator to actually generate them some source, 351 00:18:49.359 --> 00:18:52.240 then it's also pretty easy to end up with double 352 00:18:52.319 --> 00:18:57.039 generations and you basically then have two conflicting types because 353 00:18:57.079 --> 00:19:01.160 the generator run twice, and that's what you have a 354 00:19:01.200 --> 00:19:02.200 new project in the end. 355 00:19:02.440 --> 00:19:05.240 Oh wow, yeah, that's kind of a good gotcha there? 356 00:19:05.319 --> 00:19:08.000 So all right, so don't there are two point zero 357 00:19:08.039 --> 00:19:10.200 So it's not in net six either. 358 00:19:10.200 --> 00:19:13.119 Not yet, sadly. Actually, I also reached out to the 359 00:19:13.400 --> 00:19:16.720 to the Rossland team on the c sharp discord channel, 360 00:19:16.960 --> 00:19:19.200 and as much as I would, they would like to 361 00:19:19.240 --> 00:19:21.400 do it. As of now, it seems like they are 362 00:19:21.440 --> 00:19:26.960 constrained by the support. So dot net framework projects also 363 00:19:27.200 --> 00:19:30.319 support these source generators. So basically, as of now, the 364 00:19:30.440 --> 00:19:32.960 best way they can do this is with dot net standard, 365 00:19:33.039 --> 00:19:35.839 because you can consume dot net standard from dot net 366 00:19:35.880 --> 00:19:38.960 framework projects and dot Net five or dot Net six 367 00:19:39.039 --> 00:19:42.839 projects alike. I am sadly not aware of whether there 368 00:19:42.839 --> 00:19:45.920 are any plans to also make it possible to reference 369 00:19:46.279 --> 00:19:50.000 dot net six projects because that's currently a bit of 370 00:19:50.000 --> 00:19:53.880 a paint point. Definitely, if you want to just even 371 00:19:53.880 --> 00:19:55.559 if you just want to get a name of the 372 00:19:55.680 --> 00:19:58.799 type you would want to check against this compilation three, 373 00:19:59.400 --> 00:20:00.680 then it's a of a hassle. 374 00:20:00.880 --> 00:20:04.480 Yeah, so I'm being done in stented two point zero only. 375 00:20:04.839 --> 00:20:06.240 You know, I think that could be kind of a 376 00:20:06.279 --> 00:20:09.559 discouragement for getting people to start using them. It's a 377 00:20:09.599 --> 00:20:13.000 new feature, but if they're not bringing it forward that much, 378 00:20:13.079 --> 00:20:17.279 then people go, well, it's probably going to get abandoned 379 00:20:17.359 --> 00:20:20.759 and deprecated at some point in time, so why should 380 00:20:20.759 --> 00:20:21.400 I use it. 381 00:20:21.400 --> 00:20:25.200 It's definitely a good point, but basically it's also not 382 00:20:25.359 --> 00:20:29.160 as bad as it might sound. So this whole generation step, 383 00:20:29.200 --> 00:20:32.079 I think you already talked about how the generation would 384 00:20:32.119 --> 00:20:34.640 create a syntax three for you, and you can then 385 00:20:34.720 --> 00:20:37.720 analyze the syntax three. It's a bit difference. It's a 386 00:20:37.720 --> 00:20:41.359 bit of a difference than with reflection because, for example, 387 00:20:41.400 --> 00:20:44.240 with reflection, you would also just go ahead and try 388 00:20:44.240 --> 00:20:48.400 to compare types or assemblies, and the syntax tree is 389 00:20:48.440 --> 00:20:52.359 basically one layer above, so you're not dealing with types 390 00:20:52.400 --> 00:20:55.119 at all, and you would usually just do it with 391 00:20:55.759 --> 00:20:59.839 either strings or you actually try to model your sauce 392 00:21:00.000 --> 00:21:02.880 generator in a way which can easily work with all 393 00:21:02.920 --> 00:21:06.200 of these syntax notes. So it's not as bad as 394 00:21:06.240 --> 00:21:09.759 it sounds, to be honest, and usually the benefits kind 395 00:21:09.799 --> 00:21:13.519 of outweigh the drawbacks. So I'm also hoping and I'm 396 00:21:13.599 --> 00:21:17.480 rooting for some improvements in that regard, especially when it 397 00:21:17.480 --> 00:21:21.119 comes to support for neo dot net versions. But as 398 00:21:21.119 --> 00:21:23.920 of now, it is also not really big deal to 399 00:21:24.000 --> 00:21:24.599 me at least. 400 00:21:24.799 --> 00:21:29.000 Okay, okay, so what are the limitations of what you 401 00:21:29.039 --> 00:21:31.720