00:36 <unique_id> i should probably just dereference when calling the function or possibly make the function take ptr to const if its a local one. These kinds of corner cases are unavoidable.

00:36 <wilsonk> does zig have a builtin isalpha function?

01:01 <kristate> woke up 10am JST

03:11 <scientes> does zig have attribute(cleanup)?

03:18 <kristate> scientes: hello, can you rephrase your question?

03:20 <scientes> he cleanup attribute runs a function when the variable goes out of scope.

03:20 <scientes> http://echorand.me/site/notes/articles/c_cleanup/cleanup_attribute_c.html

03:21 <kristate> scientes: please checkout the defer keyword: https://ziglang.org/documentation/master/#defer

03:22 <kristate> scientes: also look into the errdefer keyword: https://ziglang.org/documentation/master/#errdefer

03:30 <kristate> scientes: does that help you?

03:40 <scientes> yes

03:40 <scientes> also @firldParentPtr is awesome

03:41 <scientes> you don't see the magic of linux's container_of() get used much outside the kernel

03:46 <scientes> and is impossible in go and IIRC rust

03:58 <kristate> scientes: exactly.

03:58 <kristate> scientes: welcome to the community :-)

05:11 <scientes> the translate-c is out of date and produces struct.?.member code

05:11 <scientes> instead of orelse

05:13 <scientes> wait that seems wrong, is ? a safe dereference operator?

05:14 <scientes> so nvm

05:18 <MajorLag> unique_id: `const nothing = ([]void{})[0..];` `void` can be replaced by any type you might need.

05:19 <MajorLag> wait... nevermind that, I was reading a really old chat log somehow

05:19 <MajorLag> ...or not...

05:19 <MajorLag> whatever, I'm too tired to make sense I guess.

05:41 <scientes> https://ziglang.org/documentation/master/#struct

05:42 <scientes> is there a special case for "self" in struct methods?

05:42 <scientes> cause dot has two arguments, and then is used with one argument (and two).....???

06:10 <andrewrk> scientes, `a.b(arg1, arg2)` is syntax sugar for `@typeOf(a).b(a, arg1, arg2)`

09:53 <unique_id> MajorLag: thanks!

10:07 <unique_id> An inheritance feature could allow us to go from parent.base.field to parent.field. I have an additional use case for why this might be useful and that's to wrap types like ArrayList to handle OOM error in the wrapper and exit the program. This is wildly inappropriate for many applications, but for others not exiting just brings performance issues and noise from handling errors that you have no reason to handle. If the OS didn't

10:07 <unique_id> clean up for us things would be different of course, then propogating up would be needed.

13:13 <kristate> andrewrk: let me known when you're up, have some questions

16:29 <scientes> tree.insert(&(ns[0].node));

16:29 <scientes> tree.insert(ns[1].node);

16:29 <scientes> ^

16:29 <scientes> hello.zig:39:16: error: expression value is ignored

16:29 <scientes> tree.insert(&(ns[0].node));

16:30 <scientes> hello.zig:40:22: error: expected type '*Node', found 'Node'

16:30 <scientes> tree.insert(ns[1].node);

16:31 <scientes> pub fn insert(tree: *Tree, node: *Node) ?*Node {

16:33 <scientes> what does "expression value is ignored" mean?

16:34 <scientes> OH!

16:37 <unique_id> it means insert is returning something which you are ignoring. If you wish to ignore it explicitly, you can do "_ = tree.insert(...)"

16:37 <scientes> thats why i said "OH!

16:43 <scientes> how best to do this so zig accepts it?:

16:43 <scientes> if (parent != null) {

16:43 <scientes> parent.set_child(node, is_left);

16:45 <unique_id> if (maybe_parent) |parent| {parent.set_child(...); }

16:45 <unique_id> if (parent) |new_name|

16:48 <unique_id> or you can do:

16:48 <unique_id> if (parent != null) {

16:49 <unique_id> parent.?.set_child(node, is_left);

16:49 <unique_id> but that's strange and stupid I guess you could say, in debug mode that dereference would be checked

16:49 <unique_id> dereference/unwrap/whatever it's called

16:50 <scientes> can i do a compound if, like (maybe_parent && other > condition) |parent| { ?

16:50 <unique_id> I don't know, I guess not

16:51 <scientes> then i will if (maybe_parent != null) { var parent: ???

16:53 <scientes> how do i do that cast?

16:53 <unique_id> Maybe this works for you: if (maybe_parent != null and other > condition) { var parent = maybe_parent.?; }

16:53 <unique_id> I guess this would work too: if (maybe_parent and other > condition) { var parent = maybe_parent.?; }

16:54 <unique_id> btw there's unusually few people around here right now

16:54 <unique_id> please note how "and" is used instead of "&&"

16:54 <scientes> oh, i'm a C programmer :)

17:12 <scientes> maybe types do not cast to bool

17:12 <scientes> maybe they should

17:13 <scientes> pub fn insert(tree: *Tree, node: *Node) ?*Node {

17:14 <scientes> node = grandpa;

17:14 <scientes> rror: cannot assign to constant

17:15 <scientes> how do i make these variables?

17:15 <scientes> it is because sometimes zig passes by value?

17:16 <unique_id> no, as far as I know you can not alter parameters. So you'd have to make a copy of the node pointer first

17:17 <unique_id> it makes a lot of sense to me to have that restriction by default, the question then is whether it's worth it to add a feature to mark the parameter as mutable

17:19 <andrewrk> unique_id, marking a parameter as mutable doesn't make sense because the way that would be implemented would be equivalent to `var mutable_param = param;` and then using mutable_param

17:20 <andrewrk> so you may as well just do that if you need a mutable parameter

17:20 <andrewrk> kristate, you still up?

17:30 <wink_> andrewrk: I like a thread to wait on a queue of messages until one arrives, but I don't see a mutex & condition variable or the like to sleep/wakeup a thread,did I miss it or do I need to make my own?

17:39 <andrewrk> wink_, mutexs and condition variables are not yet provided by the std lib

17:40 <wink_> k, I'll do my own for now and if you'd like create a PR.

17:41 <andrewrk> sure, just note that I'll leave it open until I can complete the macos and windows implementations too

17:43 <wink_> yep, np. I may make it universal, but for now just on linux.

17:43 <wink_> one step at a time :)

17:44 <scientes> while (node.get_parent()) |parentconst| {

17:44 <scientes> why does parentconst have to be immutable?

17:44 <scientes> i'm trying to loop over it

17:44 <unique_id> |*parent*

17:44 <unique_id> |*parent|

17:45 <unique_id> that's one solution

17:46 <unique_id> i'm sorry that's for for loops, no clue what that's doing in this case

17:49 <andrewrk> it also works for if and while

17:49 <scientes> can you give me an example

17:50 <andrewrk> var opt_x: ?i32 = 1234; if (opt_x) |*x| x.* += 1;

18:23 <wink_> I'm looking at std/atomic/queue.zig and there are private global functions, startPuts and startGets, can I assume when building an executable that these functions are optimized away?

18:25 <andrewrk> wink_, the optimizer doesn't even see them, because they're not even analyzed because they're only referenced in the test code

18:26 <andrewrk> zig has lazy top level declarations. if you don't reference something, it doesn't get compiled

18:26 <wink_> got it, so global variables to test shared state, will they be optimized away too?

19:36 <scientes> why doesn't assignement return the variable?

19:37 <scientes> to prevent crazy c constructs??

19:37 <scientes> :)

21:48 <Dirkson> Weird question: Are coroutines concurrent yet? The description in the documentation only covers behavior that would be the same with concurrent and non-concurrent coroutines.

22:17 <wilsonk> Test

22:33 <unique_id> andrewrk: My stack trace is telling me that the originator of a "index out of bounds" issue is "return self.items[0..self.len]" from a .toSlice() call. I know I've written buggy code and I can get around it / fix it, but I'm not sure how creating a slice could result in an index out of bounds crash? Reading from the slice sure, but creating it?

22:39 <unique_id> Ok the pointer which the slice is being created from is null, I can understand crashing on that. And indeed it is a bug, I don't know how that happened!

22:40 <unique_id> Though in other cases I've been creating slices that point to nothing.

22:44 <unique_id> or maybe the pointer isn't null. Time to debug properly.

23:24 <unique_id> ok my root issue was |child| should have been |*child|, I managed to return a pointer to something on the stack. My bad.

23:31 <kristate> woke up 8:30am JST

23:31 <scientes> kristate, https://github.com/ziglang/zig/pull/1335

23:32 <kristate> scientes: cool -- do you have test cases?

23:32 <scientes> only smoke testing

23:32 <scientes> i'll make something up with rand()

23:33 <kristate> scientes: docs don't have much on testing I see

23:33 <kristate> search for `test "` on the docs page

23:45 <scientes> can I use shared state between tests, or should i just make it one test?

23:50 <kristate> scientes: that is a good question -- state should be sharable between tests

23:50 <kristate> I think the rule of thumb is that if shared state could botch the test, don't use it

23:57 <scientes> how do i do first class functions?

23:58 <kristate> scientes: for callbacks, etc?

23:58 <kristate> https://github.com/ziglang/zig/issues/1144