14:14 <zevlg> here is my very first attempt to create issue in RFC repo - https://github.com/ponylang/rfcs/issues/77 please tolerate possible stupidity

14:31 <SeanTAllen> zevlg: that's not how you create a RFC

14:31 <SeanTAllen> check out the instructions here in the README.md: https://github.com/ponylang/rfcs/blob/master/README.md

14:32 <SeanTAllen> we open issues for requests to create a RFC

14:32 <SeanTAllen> perhaps that was your intent

14:32 <SeanTAllen> in which case, you are basically saying "someone please think about this and create a RFC for it"

14:32 <SeanTAllen> vs creating the RFC yourself and having to work through any issues

14:33 <SeanTAllen> if the former is what you intended then you are good

14:33 <SeanTAllen> just want to make sure that was your intent

14:36 <zevlg> SeanTAllen: I'm not that confident in the topic in order to create RFC, I think jemc could do much better. I can intensively test the implementation or help in any kind

14:36 <zevlg> that was my intention in creating that issue

14:38 <SeanTAllen> sounds good then

14:38 <SeanTAllen> ^5

16:59 <zevlg> jemc: hey, I wrote https://github.com/ponylang/rfcs/blob/master/README.md I'm not so confident about creating RFC for this issue, could you please consider making RFC for this thing? I could test the implementation and help in as I can

17:01 <SeanTAllen> zevlg: i dont think that is the link you intended

17:02 <zevlg> oops, this is correct one - https://github.com/ponylang/rfcs/issues/77

17:14 <dougmacdoug> just curious.. what is the problem with just putting the data in the struct yourself? what use case do you have where you need multiple ways to access a data from union ? that seems like it would only be the concern of the internal C api.. you could even write a pony union type with a simple let match to populate the correct U8 tuple ..

17:26 <zevlg> dougmacdoug: adhoc solution is of course possible right now, you just calculate biggest field in union and add it as field, and then just cast to type you need. It does not feels natural however, having builtin union type doing all that work for you - that is feels natural :)

17:29 <zevlg> also, functionality for https://github.com/ponylang/rfcs/issues/75 might require union type as well

18:22 <jemc> dougmacdoug: many

18:22 <jemc> sorry, didn't finish typing there

18:23 <jemc> dougmacdoug: many C libraries have unions as part of their public interface, so being able to deal with them from Pony would be a prerequisite to using thoe from FFI

18:24 <jemc> also, bitcasting isn't possible in Pony without using a lot of C hackery

18:24 <jemc> zevlg: I'll bring it up for discussion on today's sync call, to get some basic feedback on the idea and what pitfalls we might run into with it

18:28 <dougmacdoug> well, i've only done a few things with ffi so far, but it just seemed to me you could write a wrapper class that would convert a pony union type to the correct collection of U8 tuples to pass to the C call and reverse

18:43 <dougmacdoug> ah, nvm.. i see the issue now..

20:22 <jemc> zevlg: which C library are you trying to bind to, by the way?

20:26 <aav> if I have Array[U8] ref - how is it possible to create Array[U8] val out of it? (obviosly it must be a copy)

20:26 <aav> Array.clone() type is Array ref

20:50 <aav> this is the only solution i got to work

20:50 <aav> https://gist.github.com/aav/4d88bfb077d1b6bbd8fe844d49fddc87

21:04 <aav> but it will not work, if source array would be a class member (cannot be consumed)

21:22 <SeanTAllen> aav: you need to copy from the one to the other

21:54 <jemc> aav: we have a solution for making `Array.clone` work better with capabilities, but it requires this RFC (which has already been accepted) to be implemented: https://github.com/ponylang/rfcs/pull/62

21:54 <jemc> for now the workaround is to do a loop and copy one element at a time with `push`