| 2 Mar 2020 |
| 18:38:51 |  kiti_nomad | Someone told me that the jvm language takes AOT when compiled into bytecode, and JIT when run in a virtual machine. Languages like cpp directly generate executables, so it takes AOT
is it right? |
| 20:56:28 |  yorickpeterse | JIT stands for Just In Time. AOT is Ahead Of Time |
| 20:56:47 | yorickpeterse | A JIT compiler is typically used in an interpreted language to compile VM bytecode into machine code |
| 20:56:54 | yorickpeterse | An AOT compiler is something like clang, gcc, etc |
| 20:57:18 | yorickpeterse | Inko has neither, it only has a bytecode compiler that takes Inko source code and compiles it into a custom bytecode format |
| 20:57:23 | yorickpeterse | The VM then just executes that as-is |
| 20:57:36 | yorickpeterse | A JIT probably will be added one day, but I suspect it will take a few years |
| 5 Mar 2020 |
| 01:12:52 | |  imransherrafique joined the room. |
| 16 Mar 2020 |
| 08:13:27 |  fsx | How is it going? |
| 18:16:04 | yorickpeterse | Slow but steady, working on implementing all the type checking rules in the self-hosting compiler |
| 18:16:25 | yorickpeterse | But this requires quite a bit of re-thinking to make sure those rules are actually what I want; instead of just porting over the Ruby rules as-is |
| 18:17:27 | yorickpeterse | We're now at about 7000 LOC of compiler tests, though a substantial portion of that is for the parser |
| 18:19:04 | yorickpeterse | It's taking a bit longer than I'd like, mostly because I'm trying not to introduce decisions I will come to regret a few months down the line |
| 22 Mar 2020 |
| 15:31:52 | |  ashunix joined the room. |
| 23:10:09 | yorickpeterse | I'd say I'm now about 70% through the type checking rules for all the types. This mostly consists out of defining methods to check if A is compatible with B, when, etc; but it's great progress so far |
| 23:11:25 | yorickpeterse | Even found some edge cases not handled by the current Ruby compiler, such as not being able to pass a type ?A to something that expects a trait B, even when both A and Nil implement the B trait |
| 23:12:23 | yorickpeterse | e.g. the Ruby compiler won't let you pass ?String to something that expects ToString |
| 23:12:41 | yorickpeterse | Even though both String and Nil implement ToString, so it's totally safe |
| 23:14:47 | yorickpeterse | Once all these rules are implemented I probably need to groom the code a bit though, pretty sure it could be a bit nicer in a few places |
| 23 Mar 2020 |
| 11:20:56 | fsx | Cool |
| 11:21:24 | fsx | Are the rules based on some existing theory, does it have a name? |
| 12:01:24 | yorickpeterse | Not really, I just wrote them based on what I think makes sense |
| 12:01:43 | yorickpeterse | The rules are pretty simple though, like object A is compatible with trait B if A implements B |
| 12:02:16 | yorickpeterse | It's certainly nowhere near the complexity of all the formulas you may see in type checking/inference papers |
| 13:14:36 | fsx | I.e. yorick-rules. |
| 13:14:49 | fsx | :p |
| 15:46:34 | yorickpeterse | Yeah basically it comes down to them just being arbitrary, though I think most other languages follow similar rules |
| 25 Mar 2020 |
| 21:26:56 | yorickpeterse | I think I just finished implementing the type checking rules, now I need to groom this code a bit more, but pfew; finally |
| 2 Apr 2020 |
| 13:12:37 | yorickpeterse | Started working on the type checking compiler pass again this week (now that most rules are in place). Still have to add tests for quite a bit of helper code, but will do that along the way |
| 13:12:52 | yorickpeterse | Took a while to get here, but hopefully progress should go up now |
