llmII: here's a taste of what you can do with parser/* stuff: https://github.com/sogaiu/janet-editor-elf/blob/master/janet-editor-elf/delims.janet#L16-L53
this is what i use to find missing delimiters in "partial" code
I'll probably want the parser api, if I could somehow parse multiple blocks, or I guess I could just wrap the entire file in
[] to get it all back as an array of toplevel values
parsing multiple things should be doable too
I'm guessing the full parser API would be needed for the multiple things?
unless you wrap like you described
that's why my first thought was array-wrapping a string, keeps things simple to be just a single parse call
yes :)
hehehe, I thought you were about to mention an edge case where doing so means breaking everything lol
hehe, I see, still getting used to matrix, and tab complete selection, good to know they're both you
anyway, this part of Janet is the one I like a lot. Not that I am using it that often, but there is something calming in the knowledge that it exists :-). And is not that hard to use, when needed
2 replies
pp: got to thinking about it, the context stuff, saw an example of that somewhere actually, very useful, as to do-file... could spawn it in another fiber with a deadline and between limiting context so they basically only get def/var and specials and maybe something you want to insert, and setting a deadline (no infinite while loops), can possibly prevent that
by prevent that, I mean, prevent arbitrary code exec
I'm unsure if the deadline for a spawned coroutine will work if all that is done is blocking actions
ev/with-deadline won't work for that, as I feared, might be better to just start a process to read the config, or start a thread really
(ev/with-deadline 5 (ev/thread (fn [] (while true ())))) works 😁
only issue I guess is there's no way to kill a thread like that that I know of
that said, I'd guess dofile and the like are ok where things can be trusted, but using the parser is the best bet for stuff that the user supplies and you don't want to trust... not sure what to do about the ability to have hung threads if someone inserts a "while true" loop, sort of I guess want a way to say "stop running the script if it takes too long", processes have better support in this case than threads, since such can be killed if it runs away
oh well, configuration file parsing can be done with parse, doing it with nice error messages can be figured out later, just have to think on how to use the parser to do multiple values instead of array-wrapping possibly
possibly of interest? https://github.com/janet-lang/janet/blob/master/src/boot/boot.janet#L3512-L3522
that's of interest indeed, if I could prohibit
fn defn defmacro and variants that'd pretty much make flycheck do great for checking for errors without having to worry about arbitrary execution
made it easier for me to look through recent janet commits
as there was saikyun/freja@bea027e that referred to
debug/stacktrace
I hoped to be funny
And thanks sogaiu as always!
There are some improvements to the numarray example. But they are in C, so I would love to get any feedback. 🙏
If anyone here has some upward facing triangles, I would be glad here https://lobste.rs/s/2poljt/neil_bohemian_rhapsody_for_five_channels
I would also be glad for any feedback :-)
pp: I don't have an account there but I did a quick read when I was groggy and it seemed interesting, you've put a lot of work into things that's for sure. I can't say I actually understand reactive streams but I believe when I reread it I'll get the gyst even if I wouldn't know of something to directly apply it to. I found it interesting you used jhydro and spun up an encrypted netrpl ordeal (I still haven't found a way to use netrpl, which probably hints more at my lack of imagination and my approach to things being solve some things that are necessary for my needs in the long run)
I'm having issues using a table with keys that are arrays i mean something like that:
@{@[1 2] 1 @[2 3] 4}