Mir libraries on "BetterC" dialect of D
This is a sequel, and correction, to a previous post:
There, I failed to get a simple string datatype test to work in the variant of D without a runtime. This could be very useful as a way to extend languages that have their own runtime, e.g. Lisp and Prolog dialects. However, when I tried this again, I got it to work this time.
There are a few things I did differently. And more importantly, my aim has changed in the meantime. My interest now lies in coding "hot loops" of programs written in higher-level languages; I don't find rewriting systems in a procedural/OO language very interesting.
- Run "dub upgrade" to upgrade the dependencies (just mir-core & mir-algorithm) to the latest released versions.
- Need to add `"buildOptions": ["betterC"]` to dub.json.
- I figured out that [] converts the string datatype to an array, then .ptr converts that to pointer which can be passed to C stdlib routines. So if s is a SmallString object, you can use it like `puts(s[].ptr)`. No reallocation or similar complexity is required.
After this, "dub build" works. There are a few other things I've learned:
- It's best to make main() @safe and a small percentage of leaf functions @trusted. This is similar to the other languages that have a safe/unsafe divide.
- I use dmd on Windows, ldc on macOS and gdc on OpenBSD. For OpenBSD, dub isn't in the package repos any more so I have to download & build the mir libraries manually, but this isn't terribly difficult.
Like I said earlier, being able to compile without a runtime is almost essential to extend higher-level languages. And this is the favourite way I've found so far (C++ with the Core Guidelines is another option, but I think it's complicated).
Source