WEBVTT 00:00.000 --> 00:16.080 Hello, my name means, I don't know, we'll see, my name is Victor, hello, hello, so my name is Victor, I'm 00:16.080 --> 00:22.560 20 years old, I was born in Bulgaria, where I lived for most of my life, now after that I lived 00:22.560 --> 00:30.400 for theories in Dublin, Ireland, and now I'm a secondary student at TU Delft, today I wanted 00:30.400 --> 00:39.040 to tell you something about my recent last year's Google Summer of Code Project, where the 00:39.040 --> 00:46.160 goal was to migrate the Lua package manager from the dynamically typed Lua to statically 00:46.160 --> 00:55.040 typed dialect of Lua called Deal, and basically go over the benefits of that, as well 00:55.040 --> 01:04.240 as how planning happened and how the reality of everything worked, as you're going to see 01:04.240 --> 01:11.440 that it was not as straightforward as a dimension, then how I tackle different problems during that, 01:11.440 --> 01:19.440 and at the end what my work flow was, and then what happened at the end, so without 01:19.440 --> 01:24.560 certain, I wanted to start with what Google Summer of Code is, so Google Summer of Code does probably 01:24.560 --> 01:37.840 know it's a platform where big open source organizations can put projects with their contributors, 01:38.160 --> 01:47.840 so that people who are new to open source development can get a feel what is like to work 01:47.840 --> 01:54.560 on open source, and not only on small projects, but large and active projects, so my 01:57.920 --> 02:03.360 so the idea is that it's a kind of like a windman situation where the person gets experience 02:03.360 --> 02:10.480 while the organization gets a new feature, or a new implementation, or a new basically a new 02:10.480 --> 02:16.560 feature, so my thing on that is that if you haven't tried it and contributing to open source 02:16.560 --> 02:23.520 to something you are passionate about, definitely give it a go, it was really once in a lifetime 02:23.520 --> 02:33.920 opportunity for me, now without any more blubbing, let's go to the actual benefit of the migration, 02:33.920 --> 02:43.680 so the main benefit was obviously to introduce types to the lower works codebase, 02:44.800 --> 02:50.800 to hopefully make it clearer, more understandable, and more maintainable for the future, 02:51.360 --> 02:58.080 the second benefit, which was also a really nice other benefit, was a new improved to the 02:58.080 --> 03:14.160 teal language, which was a good idea for this project as it was a large project, large scale project, 03:14.240 --> 03:22.320 and allowed for encompassed edge cases, which teal, we saw that did not know how to handle, 03:22.880 --> 03:31.200 but now it can, so without let me show you an example, so here on the left we have a code, 03:31.280 --> 03:44.320 which I took from the manifest writer, manifest fire writer, yes, and it's an example 03:44.320 --> 03:48.800 I picked, because first of all I want you to think about what the type of storage is, 03:49.600 --> 03:54.480 and it's not written there, it's underlined there, and you can probably figure out from the code, 03:54.480 --> 03:59.680 like that's why I picked this example, it was not the most difficult one, or the most simplest one, 03:59.680 --> 04:05.360 but I believe that it's manageable to be figured out just from the code itself, 04:06.560 --> 04:12.240 so I hope you guessed it, it's actually a map from a string to an array of strings, 04:12.240 --> 04:17.760 where here you can see the string, it works, oh this works, okay, so here you can see that 04:19.760 --> 04:25.440 the storage takes an item name, the name of the item that you want to insert, and 04:26.400 --> 04:33.440 are an array of combinations between names and versions, so as you can hopefully see now, 04:33.440 --> 04:39.840 the second example, if someone were to try to make an improvement on that specific method now, 04:39.840 --> 04:44.240 they can just look at the types and easier go through the code, not spending their time 04:44.240 --> 04:52.160 trying to decode what exactly the code does, now I don't know how many times, so let's move on with 04:52.160 --> 04:58.480 what planning was, so this was my initial proposal back in April last year, it's small, 04:58.480 --> 05:04.080 I know the idea is not to read it, it's there if you want to read it, but that's not the go, 05:04.080 --> 05:14.160 I want to just point some places where planning, planning magic happened, anyway, so 05:14.480 --> 05:24.880 first I wanted to see that here, I forgot that I had exams, so this was not a week, I was 05:24.880 --> 05:32.000 able to work on, and so that's low the project, then my plan pre-planned vacation was actually 05:32.000 --> 05:38.560 not this week, but actually this one, which may seem not important, however, I what I figured is 05:38.560 --> 05:43.600 that it's really important for those changes to be, or for your schedule and for when you're 05:43.600 --> 05:52.320 available, it was really important to communicate that with your mentor, as for me or for our case, 05:52.880 --> 06:00.560 it almost caused the project to not be finished, so don't undermine that, now another thing is, 06:00.560 --> 06:06.640 I want to mention is the last week, I put a buffer week, which in turn is the thing that saved 06:06.640 --> 06:12.320 the project, so the mistake was the thing that almost failed it, and then the buffer week was the 06:12.320 --> 06:22.240 thing that saved it, now as to my delusions at the start, you can probably see how I had the image 06:22.240 --> 06:28.320 in my head that by week five I was going to be done with 20% of the project, and by week six I was 06:28.320 --> 06:33.520 going to be done with 80% of the project, and by week seven I was going to be done, which was not what 06:34.240 --> 06:41.040 happened, now let's go with actually how I approach the problem, so my approach to the problem 06:41.040 --> 06:47.920 as there were many files interacting with each other was to create a dependency graph, so a dependency 06:47.920 --> 06:57.440 graph where the root files were the core files of the project, which did not interact with anything 06:57.440 --> 07:03.200 outside of it, so the interactives within each other, but once they were all figured out, 07:05.200 --> 07:10.160 these two dos of foundation have already translated code, then I moved out layer by layer until 07:10.160 --> 07:19.120 the leaves, which were the mostly the files handling the commands of the CLI, so pretty up the 07:21.600 --> 07:26.800 list, then however, of course it was not a straightforward as with most things, and we had a 07:26.800 --> 07:34.240 few exceptions, first of which were the external dependencies, the external dependencies are in 07:34.240 --> 07:39.120 Lua, and so you have to provide some sort of definitions to TIO, so it can handle them, 07:41.120 --> 07:47.760 this however, it's a really nice feature of TIO, which provides definition files, which I used, 07:47.760 --> 07:53.200 however, as we wanted to improve TIO, I was on the development branch of TIO, rather than the 07:53.200 --> 07:59.040 release branch of TIO, which created a few changes which also had to be taken care of. 07:59.040 --> 08:05.520 Now the next thing were also files that after some decisions and considerations, we decided that 08:05.520 --> 08:09.520 were not in the scope of the project, so that they would not be converted, meaning that they would 08:09.520 --> 08:14.800 also need their own definition files. And the third thing are the cyclic dependencies, 08:15.760 --> 08:23.040 cyclic dependencies are handled by Lua, however, they are not a good practice, and we wanted to get 08:23.120 --> 08:30.560 rid of them, so that also happened by refactoring methods from one file to another to hopefully minimize 08:32.560 --> 08:39.520 the cyclic dependencies. Here I also wanted to add a quick note on challenging parts, which I encountered 08:39.520 --> 08:46.000 during my actual conversion, which was when I was stuck and a method that I could not figure out, 08:46.640 --> 08:55.760 instead of stopping, a solution was to continue working to down the file or to a different 08:55.760 --> 09:01.120 file so that I do not get blocked and mark that part where, which I did not understand, is a question. 09:01.120 --> 09:07.600 That led to two huge improvements, one of which was that I could be more efficient with my work, 09:07.600 --> 09:14.880 and the second one was that the meetings with my mentor had more of a agenda to them, 09:14.960 --> 09:22.000 which made them more structured, and we could go over more blockages in them. Now, I want to talk 09:22.000 --> 09:30.160 about the dependency, the development workflow about it, so how I did testing, so after I completed 09:30.160 --> 09:36.480 the file, how I was testing them. In this case, I had to start when I converted the file, 09:36.480 --> 09:42.400 I could run the already existing workflow on GitHub, on GitHub actions, which took about seven minutes, 09:42.400 --> 09:46.960 which was fine because I could continue to the next file, and only go back if it failed. 09:46.960 --> 09:54.000 However, at some point in the project, I managed to magically break up the project, 09:54.000 --> 10:00.160 so seven minutes per iteration was not a good idea. So then, after some filling around, I managed 10:00.160 --> 10:08.160 to set up the built environment locally, which brought down the runtime from minutes to seconds, 10:08.160 --> 10:15.360 which was a huge improvement, and as a note, more of the story is in best more time 10:15.360 --> 10:22.320 in testing, as it is really important for your workflow. Another thing is bunching, branching, 10:22.320 --> 10:32.160 we had two options for it, one of which was one was incremental releases, so every 10:32.160 --> 10:37.280 comment I had would go on to a branch, which then would get merged, which was the straightforward 10:37.280 --> 10:43.680 solution, however, another solution we had was a one big release at the end, which kind of means 10:44.640 --> 10:51.760 after all the code was converted, moving it to a new clean branch, where there were two benefits 10:51.760 --> 10:58.160 to it. One was that each file only had one entry, even if it was edited multiple times throughout the 10:59.120 --> 11:05.040 course, and the second was that Git Blame continued to work, so on the newly generated 11:05.040 --> 11:15.200 Lua code from the Tio code. Okay, so Git Blame, the Git Blame history was saved because of the 11:15.200 --> 11:24.720 new commits that only added the new files one at a time. Hopefully that makes sense. Now, I also wanted 11:24.720 --> 11:29.520 to talk a little bit about how everything was brought together, so after the new clean branch, 11:32.320 --> 11:38.560 we had, so after the new clean branch, we wanted to wait a bit for community feedback, 11:39.200 --> 11:44.080 which also gave us time to release the new version of Tio, the development version of Tio, 11:44.080 --> 11:49.680 which was, which the code was dependent on, so at the end it worked out nicely as we could 11:49.680 --> 11:55.760 address feedback from the community as well as have the new version of Tio, and then we merged it, 11:55.760 --> 12:10.960 that's it. So to recap, I hope this presentation, although quick, 12:12.160 --> 12:17.040 helped to gain some valuable insights, which I wanted to share as they really helped me for this 12:17.040 --> 12:22.400 project, and who definitely, I would definitely use in the future and taking to consideration. 12:22.400 --> 12:29.600 So hopefully now, can you do? This slide is for related links, which are on the 12:30.400 --> 12:37.520 first them the presentation page, on below in the resources, as well on the presentation slides, 12:37.520 --> 12:42.640 which are also there, and if you have any further questions, as I don't think we have much time, 12:43.280 --> 12:49.680 maybe, you'll have much time, so you can either talk to me outside after the presentation, 12:49.680 --> 12:56.160 or on one of those two links here. Thank you, Victor, that's awesome. 12:56.160 --> 13:06.160 I think, it's also a really nice, we get students to speak, you know, from the Geeks Summer of Code. 13:06.960 --> 13:14.640 We've also, as a theme of the day, we're going to replace all words with Geeks, so we'll have Geeks Emax. 13:19.840 --> 13:22.080 So let's have questions. 13:36.660 --> 13:42.480 Sorry, repeat, Aydberg. 13:52.020 --> 14:04.800 It was made short, a minute before the last one fight. 14:04.800 --> 14:13.800 So the question was, if I had to change anything about the signature or just add the 14:13.800 --> 14:19.800 types, in most cases, I could get away with just the types, however, with the cyclic 14:19.800 --> 14:26.800 dependencies, for example, I had to refactor code around to deal with those, which men 14:26.800 --> 14:33.800 in this signature sometimes. 14:33.800 --> 14:36.800 In fact, it reminded me of what? 14:36.800 --> 14:40.800 Yes, that was the go. 14:40.800 --> 14:41.800 Any more questions? 14:41.800 --> 14:42.800 I don't think if we have time. 14:42.800 --> 14:44.800 For more questions, but. 14:44.800 --> 14:45.800 Okay, sure. 14:45.800 --> 14:46.800 One more question. 14:46.800 --> 14:47.800 Anyway. 14:47.800 --> 14:48.800 Okay. 14:48.800 --> 14:53.800 What goal is off the cyclic when moving something right from the stack of each 14:53.800 --> 14:54.800 type language? 14:54.800 --> 14:56.800 Is the binding fixed issues? 14:56.800 --> 14:59.800 Did you find any interesting issues in the world? 14:59.800 --> 15:03.800 We actually did, but they're quite lengthy, so we cannot go into much detail. 15:03.800 --> 15:12.800 But one of which was the hash, the hash table was interestingly implemented in 15:13.800 --> 15:16.800 which we then had to change how deal dealt with it. 15:16.800 --> 15:24.800 And another good one was some code that was redundant, which we managed to remove in the process. 15:24.800 --> 15:27.800 So yes, that was something we'd do nice. 15:27.800 --> 15:30.800 Now I'm going to leave it to the next person, so we don't. 15:30.800 --> 15:31.800 Oh, we have time. 15:31.800 --> 15:32.800 Okay. 15:32.800 --> 15:33.800 Sorry. 15:33.800 --> 15:34.800 I cannot. 15:34.800 --> 15:35.800 Yeah. 15:35.800 --> 15:38.800 Did they leave myself in the work? 15:39.800 --> 15:42.800 No, I did not. 15:42.800 --> 15:44.800 I was not smart enough to do that. 15:44.800 --> 15:45.800 No. 15:45.800 --> 15:47.800 Maybe a different project, you know. 15:47.800 --> 15:49.800 Any more questions? 15:49.800 --> 15:52.800 I'm letting you stop me because I don't know. 15:52.800 --> 15:53.800 Okay. 15:53.800 --> 15:54.800 Okay. 15:54.800 --> 15:55.800 Yeah. 15:55.800 --> 15:56.800 Yeah. 15:56.800 --> 15:58.800 Five minutes to complete the talk and then you're five minutes for questions. 15:58.800 --> 15:59.800 Okay. 15:59.800 --> 16:00.800 Good. 16:00.800 --> 16:01.800 So any questions? 16:01.800 --> 16:02.800 Yes. 16:02.800 --> 16:20.800 So the idea, the way to language works is by writing the types and then compiling it back to 16:20.800 --> 16:21.800 Lua. 16:21.800 --> 16:24.800 So everything is still running from Lua. 16:24.800 --> 16:29.800 However, most of it was written from TL. 16:29.800 --> 16:31.800 So compiled from TL.