> I would have thought that the usage of parseInt, was fairly obvious from the type-signature:
Blog author here. In hindsight yes it's obvious.
I think my problem was a lack of understanding of comptime types. They're a little different from C++ / Rust since they're passed as args. Looking back I think I found the Zig docs less "sticky" than other languages. The concepts are familiar but just new enough I don't understand and don't fully remember them.
There's a point while working on AoC that I wasn't sure if Zig had first-class runtime types or not. It has comptime types and it also has a TypeInfo built-in with some degree of reflection? Is that runtime reflection maybe? I think partially? I honestly don't know.
As far as I know, typeinfo is used at compile time. In general, doing runtime type shenanigans involves implementing things like vtables, and some other weirdness. This breaks "no hidden control flow".
If you do that at comprime instead, the thing that runs at runtime is just memory offsets. You can for example generate code that parses json into 30 types you know at compile time, or you can use things like hashmaps and arraylists and whatnot to do it at runtime.
You get a lot of mileage by just doing the dynamic code away from runtime, and that's one of the benefits of comptime. And it's a slow thing to realize, also not necessarily something everyone cares about.
> It has comptime types and it also has a TypeInfo built-in with some degree of reflection? Is that runtime reflection maybe?
I do wish this was explained more prominently! The way this seems to work is that Zig has a lazy runtime for Zig code at compile time that does some interning (such that identical types obtained separately are equal, identical strings obtained separately are equal, etc.), and some types such as type can only exist in this runtime, not in the actual uh "run time" runtime. TypeInfo is for iterating over struct, union, or enum definitions at compile time to generate code. You'd use it if you wanted to write a generic json parser and serializer or if you wanted to write a type that's like an arraylist but internally uses a struct-of-arrays layout. Both of these are in the standard library and make informative reading.
TypeInfo itself is a union that can exist at runtime, but I don't think you can get into a situation where you have something at runtime and you don't know what type it is already. So actually using the TypeInfo at runtime may not be very useful.
> Zig appears to not report compiler errors for functions that get optimized out.
This is apparently caused by the laziness. The docs suggest using std.testing.refAllDecls(@This()) to have the compiler check unreachable code. I mean the actual usable docs at https://ziglang.org/documentation/master/#Nested-Container-T... rather than the automatically generated stdlib docs.
Blog author here. In hindsight yes it's obvious.
I think my problem was a lack of understanding of comptime types. They're a little different from C++ / Rust since they're passed as args. Looking back I think I found the Zig docs less "sticky" than other languages. The concepts are familiar but just new enough I don't understand and don't fully remember them.
There's a point while working on AoC that I wasn't sure if Zig had first-class runtime types or not. It has comptime types and it also has a TypeInfo built-in with some degree of reflection? Is that runtime reflection maybe? I think partially? I honestly don't know.
Either way I was confused. :)