Sockets on basic-cli · beginners

In this case, I also managed the lifetime of the file to make sure the handle got closed. Only exposing a function that took a lambda for what to do with the file.

Brendan Hansknecht (Feb 28 2023 at 02:02):

Oh, though I used U64, which is wrong on 32bit platforms, so I wouldn't copy that.

Agus Zubiaga (Feb 28 2023 at 02:03):

I see

Agus Zubiaga (Feb 28 2023 at 02:04):

So Nat is like usize?

Brendan Hansknecht (Feb 28 2023 at 02:04):

Yep

Agus Zubiaga (Feb 28 2023 at 02:05):

Alright, on it! Thank you :)

Agus Zubiaga (Feb 28 2023 at 03:58):

It's getting more interesting chat.gif

Ayaz Hafiz (Feb 28 2023 at 03:59):

This is pretty fire, nice Agus!

Brendan Hansknecht (Feb 28 2023 at 04:29):

I love the gif. Really cool to see.

Brian Carroll (Feb 28 2023 at 06:54):

Oh wow, so cool! :tada:

Richard Feldman (Feb 28 2023 at 12:55):

wow, that's awesome!!! :heart_eyes:

Richard Feldman (Feb 28 2023 at 12:55):

@Agus Zubiaga was this all on your first day getting into Platform development?

Agus Zubiaga (Feb 28 2023 at 12:56):

I did cross midnight but yeah :grinning:

Agus Zubiaga (Feb 28 2023 at 12:58):

It was my first day actually using Roc at all. I had only seen some talks before.

Agus Zubiaga (Feb 28 2023 at 12:59):

Props to you folks for making it so straightforward!

Richard Feldman (Feb 28 2023 at 12:59):

wow, amazing! :heart_eyes_cat:

Agus Zubiaga (Feb 28 2023 at 13:09):

It's been fun! I'll play with the API when I get some more free time. I have to make some things fail at the Task level instead of panicking in the platform, and figure out how to read arbitrary bytes.

Agus Zubiaga (Feb 28 2023 at 13:09):

If everything goes well I can probably make a PR later in the week!

Agus Zubiaga (Mar 01 2023 at 21:33):

Got the API to a pretty good state I think: https://github.com/agu-z/roc-basic-cli/blob/sockets-support/src/Tcp.roc
Example: https://github.com/agu-z/roc-basic-cli/blob/sockets-support/examples/tcp-client.roc

Agus Zubiaga (Mar 01 2023 at 21:35):

It can now read all data in the stream (was reading into a fixed buffer before). Also, you can read/write raw bytes if you want to.

Agus Zubiaga (Mar 01 2023 at 21:38):

Replaced Tcp.connect with Tcp.withConnect which manages the lifetime of the connection so you can't forget to close, similar to: https://github.com/roc-lang/roc/blob/main/examples/cli/false-interpreter/platform/File.roc

Agus Zubiaga (Mar 01 2023 at 21:40):

I think the only missing part for a PR is error handling. I wrote the types but glue doesn't work so I can't return them from Rust.

Agus Zubiaga (Mar 01 2023 at 21:41):

I'll look into how much code per type we need, maybe I can just write them by hand

Agus Zubiaga (Mar 01 2023 at 22:14):

Yeah, maybe not lol

Agus Zubiaga (Mar 01 2023 at 22:16):

I'd rather spend the time investigating why glue doesn't work in the first place. Gotta run now but I might do that later or tomorrow.

Richard Feldman (Mar 01 2023 at 22:17):

I'd love to have someone else get involved with glue, so if you want to look into it that would be awesome! I'd be happy to help with answering any questions :smiley:

Agus Zubiaga (Mar 01 2023 at 22:22):

Yeah, it’ll probably take me a while to get familiar with the compiler internals but happy to!

Agus Zubiaga (Mar 04 2023 at 18:54):

I managed to generate the glue code for my errors with this trick: https://roc.zulipchat.com/#narrow/stream/316715-contributing/topic/Broken.20glue/near/339383878

However, even if a I return a RocResult::ok from the rust, when I pattern match on the Roc side it always matches the Err case. Even if the platform is never creating RocResult::err at all.

Folkert de Vries (Mar 04 2023 at 18:55):

and neither of the a and x in Result a x are the [] type?

Agus Zubiaga (Mar 04 2023 at 18:56):

Not in the Effect module or my Tcp module

Agus Zubiaga (Mar 04 2023 at 18:56):

Only the main function because it needs to be Task _ []

Folkert de Vries (Mar 04 2023 at 18:56):

yeah so not in relevant code paths.

Folkert de Vries (Mar 04 2023 at 18:56):

ok, well this sounds like just a mismatch between where glue is storing and roc is reading the bytes

Agus Zubiaga (Mar 04 2023 at 18:57):

Yeah, that's what I thought. Lemme try something like a RocStr from roc_std an error.

Agus Zubiaga (Mar 04 2023 at 18:58):

Well, I guess that wouldn't matter because I'm never generating an error yet

Nick Hallstrom (Mar 04 2023 at 19:00):

Is that the same issue as this? Folkert’s response from the Roc Clock thread (can’t seem to copy the link from mobile):

well. this is interesting and frustrating: rust will send a RocResult<u8, ()> over to roc as a u16. we "receive" it as { [0 x i8], [1 x i8], i8, [0 x i8] }, which is reasonable (if you've seen LLVM IR before), but it is _not_ valid according to the C ABI