blob: c7a5ccb4377772fa2619a56d35b27162de819b59 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
|
[2026-06-26]
- annotations do not work properly because macroexpansion creates an entirely new body for lambdas
- we realized that we are handling macro application in two separate places: pit_expand_macros expands lambda bodies, and pit_eval expands macros encountered while evaling. we ought to unify this so that only one is used (probably pit_expand_macros, because we need to expand macros eagerly to identify free variables to capture)
- we probably can make pit_expand_macros and pit_eval much nicer
- we can probably make pit_expand_macros operate in place
- if we want to be really smart, cool, happy, rich:
let's just make stuff translate to a little VM before it evaluates, and let's store VM programs as functions instead of sexps
* little vm thing
let's have a "program" (representation of one function) be a cons-list of instructions
during execution, we maintain a stack of programs.
calling a lisp function entails pushing a new "frame" to this stack
we iterate over the frame by cdring as usual and we pop the frame when we reach the end
#+begin_src pit
(defun! add2 (x y) (+ x y 2))
#+end_src
#+begin_src pit
[
(literal x)
(get)
(literal y)
(get)
(literal 2)
(literal +)
(fget)
(apply 3)
]
#+end_src
#+begin_src pit
(defun! foo (x) (* (add2 x x) 4))
#+end_src
#+begin_src pit
[
(literal x)
(get)
(literal x)
(get)
(literal add2)
(fget)
(apply 2) ;; evaluating this pushes a new frame with the code for add2. next "tick" of evaluator continues in that code. once it returns, result is on the stack (for free)
(literal 4)
(literal *)
(fget)
(apply 2)
]
#+end_src
|
