Advent of Codon and Capturing Loop Variables

[AndreiMoraru123]

Hi again folks, I'm doing Advent of Code this year to try out Codon.

Today, I discovered that Codon does not capture loop variables.
For instance, in Python, for this piece of code:

fns = []
for i in range(5):
    fns.append(lambda: print(i))

[f() for f in fns]

you'd canonically get a reference for the loop variable:

❯ python capture.py
4
4
4
4
4

using codon you do not:

❯ codon run -release capture.py
0
1
2
3
4

I appreciate that this choice is intentional, to enable loop parallelism via the convenient @par

Now for the AoC part of the story, which I should skip for brevity, in Python, you can naturally control flow via global states like this:

    state = True
    ans = 0

    for _, kind, value in instructions:
        if kind == "do":
            state = True
        elif kind == "dont":
            state = False
        elif kind == "mul" and state:
            ans += value

    return ans

which for the test sample

xmul(2,4)&mul[3,7]!^don't()_mul(5,5)+mul(32,64](mul(11,8)undo()?mul(8,5))

returns:

❯ python solution.py
48

And now for the problem: Codon cannot capture the state inside the loop branching:

❯ codon run -release solution.py
NameError: variable 'state' not yet defined

Raised from: __internal__.undef:0
/home/andrei/.codon/lib/codon/stdlib/internal/internal.codon:484:13
[1]    42058 IOT instruction (core dumped)  codon run -release solution.py

So what I had to do instead was force state on the heap (at least I think so?) to give it a stable memory location:

    state = [True]
    ans = 0

    for _, kind, value in instructions:
        if kind == "do":
            state[0] = True
        elif kind == "dont":
            state[0] = False
        elif kind == "mul" and state[0]:
            ans += value

    return ans

which only then works:

❯ codon run -release solution.py
48

Do you think there is/should be a better way to handle this?

Here is the whole code and input for reference:
https://github.com/AndreiMoraru123/aoc/blob/master/2024/d03/solution.py

[inumanag]

This should be fixed in 0.19 which introduces better captures; please reopen if that's not the case.