02:25 <omarkj> Anything similar to rust's racer in existence or being worked on for Pony?

02:33 <SeanTAllen> Don't know

02:33 <SeanTAllen> What is racer?

02:35 <omarkj> It's a CLI program that provides code completion, and can be integrated into IDEs and editors. So you can run stuff like `racer complete std::io`, and it will return potential completions.

02:35 <omarkj> I've integrated with in in the past in Emacs, and it's quite nice since it is that it can quite easily be reused cross editor integrations, etc.

03:23 <SeanTAllen> No such thing.

03:23 <SeanTAllen> Most likely way that would happen... we work on a language server for VSCode that provides that and then we backport functionality into ponyc or similar and make available across the board. Definitely something we want.

03:32 <omarkj> Right. I was looking just now at the docs generation part of ponyc (docgen.c) which has contains lots of the parts needed to build something like this. That, or a tool that uses the output from `ponyc --astpackage --pass=parse`, would seem like a way forward.

03:33 <omarkj> (I do not know what a language server for VSCode is, reading about that now)

03:41 <SeanTAllen> There is a plan that has been moving slowly to break apart the compiler into different tools with a "ponyc" front end

03:41 <SeanTAllen> So that you can do all those sorts of things. But like I said, it hasn't moved forward very much.

03:42 <SeanTAllen> I would be quite excited if someone started working on that.

03:45 <omarkj> Is that an Issue on the github project, or something discussed on one of the mailing lists?

03:55 <jemc> omarkj: I don't think an issue ticket exists, it's a bit more of a general aspiration at the moment

03:57 <jemc> a lot of the intention has been to try to build at least some parts of the broken-apart compiler in Pony, though that has led to some yak shaving toward making Pony a more pleasant language to write a compiler in

03:58 <jemc> which is of general benefit to the language, but also slows the particular effort

05:06 <SeanTAllen> omarkj: if its something you are interested in, i would say, dive in and start work on it and we can figure out how to incorporate with other parts of the ecosystem.

23:15 <carado> I thought env had to be passed down to have any kind of side-effect, but I just found it it seems you can just File("some-global-state.txt") without access to an Env.

23:16 <carado> so there’s no way I can assert that some code is determenistic (always returns the same thing) or has no side effects ?

23:19 <SeanTAllen> you cant do that

23:19 <SeanTAllen> File takes a FilePath

23:20 <SeanTAllen> FilePath requires a FilePath as a relative base or AmbientAuth

23:20 <SeanTAllen> carado ^^^

23:21 <carado> oooooh, neat, my bad then. so I guess it’s similarly restricted to open a socket or stuff like that ?

23:24 <SeanTAllen> yes

23:24 <SeanTAllen> anything where it isnt is an oversight and a bug

23:24 <SeanTAllen> if you find issues like that, please open an issue or submit a PR

23:27 <jemc> (with the caveat that any package with FFI access can break absolutely everything :)

23:27 <carado> cool, thanks !

23:27 <jemc> but ideally, FFI usage would be wrapped in packages that respect that convention (no ambient authority)

23:28 <carado> of course, so if I mike something like OpenGL bindings I’d want to require AmbiantAuthority even thougt I won’t do anything with it

23:28 <SeanTAllen> ᕕ( ᐛ )ᕗand you have the option when starting an application to limit the packages that are allowed to use FFI

23:28 <carado> what’s the rationale for making the FilePath require an AmbiantAuthority rather than the File require it ?

23:29 <SeanTAllen> Then you can grant access to an entire directory tree.

23:29 <SeanTAllen> You can sandbox a program to only having access to that location.

23:29 <carado> oh, but not parent files ? so Pony hand-checks that there’s no '..' in the path before handing it over to open() which parses the path again ?

23:29 <carado> (I mean unistd’s open())

23:30 <carado> (well, I mean posix’s C open)

23:31 <SeanTAllen> It should, yes

23:31 <SeanTAllen> That is the intent

23:31 <SeanTAllen> But if you have a symlink to somewhere else in that directory, well... ¯\_(ツ)_/¯

23:32 <carado> hmm

23:32 <carado> symlinks are evil anyway