Junk Drawer
For all those little papers scattered across your desk
Junk DrawerFor all those little papers scattered across your desk
This post serves primarily to host an indexable copy of some chat replies about the Racket programming language.
a: hm. what stops a #lang from reimplementing whatever algorithm a
r7rs-large compiler does?
b: in particular, nobody has yet worked out a way to maintain the Separate Compilation Guarantee™ when the compiler does phasing automatically
c: You can certainly implement that (the same way that algol60 is
implemented) but you would lose the kind of integration between r7rs
modules/macros and racket modules/macros that eg the #lang r6rs
implementation has
me: Would you have to? I’m hand-waving and clearly a non-expert, but it seems like it should be possible (procedural macros!) to implement enough of the compiler and expander as a compile-time process of the lang, and then eventually expand into an equivalent Racket module with all the relevant annotations added.
This doesn’t seem efficient (since it’s sort of like expanding the whole program twice, or worse possibly in some quadratic way), but IUIC typed racket has to do something similar in terms of duplicate expansions. Obviously the effort to implement this way would be higher (?).
c: You might be able to do this especially for a single module, but it’s not obvious if you could get the necessary information from other modules. It’s possible you could build a lot of infrastructure to do it, but I’m not certain.