I'm not sure about how people would feel about this. I have mixed feelings. It feels like a loss of many things. What are the gains from ditching continuations?
First-class continuations remains the hardest nut to crack to implement any full specification of Scheme, especially if performance and/or compactness and/or simplicity of implementation and/or integration with other languages (C, C++, Java, etc.) is a priority. Scheme--, a subset with only downward continuations, i.e. continuations that could be implemented using only C setjmp and longjmp or equivalent, would still be an extremely useful language, but it is much harder to gather a community for such a project.
> What is needed is a small, portable compiler that generates more or less "natural" C code with minimal dependencies and runtime system that supports at least the basic constructs of the language and that puts an emphasis on producing efficient code, even if some of the more powerful features of Scheme are not available.
Since C doesn't support continuations, it is not possible to have continuations and at the same time generate "natural" C code.
Consider how you would implement first class continuations. It's not possible to do CPS transformation (given the goal of natural code generation) - since that's a whole program transformation.
Given the intent to have relatively straightforward C code and a tiny runtime, you pretty much have to ditch them or you're not going to get anywhere at all.
Plus honestly most of the uses of continuations that would make sense in the sort of relatively low level code this is aimed at can be built out of a few special purpose macros (and possibly a small amount of crying) - scheme code tends to lean on first class continuations for a lot of things that don't really need them, Because It Can.
e.g. Paul Graham's On Lisp shows how to build (a subset of) continuations out of macros, and getting something like clojure's 'recur' out of that would probably give you simpler macros than pg's code uses.
It's fine because this is a static subset of Scheme that you could use to, say, implement a runtime for a full Scheme. As a Scheme programmer, it feels nice to be able to implement your Scheme in something very close to Scheme instead of having to use C.
neonscribe|1 year ago
soegaard|1 year ago
> What is needed is a small, portable compiler that generates more or less "natural" C code with minimal dependencies and runtime system that supports at least the basic constructs of the language and that puts an emphasis on producing efficient code, even if some of the more powerful features of Scheme are not available.
Since C doesn't support continuations, it is not possible to have continuations and at the same time generate "natural" C code.
Consider how you would implement first class continuations. It's not possible to do CPS transformation (given the goal of natural code generation) - since that's a whole program transformation.
sctb|1 year ago
Since CHICKEN actually does this, I'll add the proviso of "without a GC or costly runtime".
mst|1 year ago
Plus honestly most of the uses of continuations that would make sense in the sort of relatively low level code this is aimed at can be built out of a few special purpose macros (and possibly a small amount of crying) - scheme code tends to lean on first class continuations for a lot of things that don't really need them, Because It Can.
e.g. Paul Graham's On Lisp shows how to build (a subset of) continuations out of macros, and getting something like clojure's 'recur' out of that would probably give you simpler macros than pg's code uses.
davexunit|1 year ago