Compile block design questions and links

[mariari]

Currently to include c code, it looks something like

foreign c {
  int readPre(const char *key) {
      if (strcmp("change1-key", key)) {
          return 100;
      \} else if (strcmp("change2-key", key)) {
          return 90;
      \} else {
          return -1;
      \}
  \}
};

axiom readPre : String → Int;
compile readPre {
  c ↦ "readPre";
  ghc ↦ "readPre";
};

or perhaps

foreign c {
  include #<libanoma>
}
axiom readPre : String → Int;
compile readPre {
  c ↦ "readPre";
  ghc ↦ "readPre";
};

This currently works, as we just spit out a single C file (not included as I don't have wasi setup.).

However, if future plans change (say your compilation unit changes from the entire program, to each package, or maybe wish to do linking similar to OCaml to speed up compile times massively), this would not scale with it, as semantically the system doesn't know how things relate and what to bring into scope for another library who wishes to use this, so it may present problems.

Something like factor's or CL's Alien functionality may help.

https://docs.factorcode.org/content/article-alien.html

https://pmsf.eu/pub/cmucl/doc/cmu-user/aliens.html
https://cffi.common-lisp.dev/ (portability layer, has some interesting documents)

https://docs.factorcode.org/content/word-FUNCTION__colon__,alien.syntax.html

I use it in https://github.com/mariari/Misc-Stack-Scripts/blob/master/Factor/compiling-lisp/unary/unary.factor

! Missing FFI Functions
LIBRARY: libc
FUNCTION: c:int mprotect ( c:void* addr, c:size_t len, c:int prot )

! Click on the url above to see how I use this!

I will note I'm not the most familiar with using FFI at this level, but these may help with future design around the problem. From what I understand these languages handle the FFI problem fairly decently

Also a follow up question: How does GC work in regards to foreign blocks, could GC make certain features of this hard to maintain or remove invariants that is relied upon.