Video

Oh, common why not to use global thread local var which you can embed in error enum if you want? Code will be way more concise plus no ugly extra args in function calls

excellent video

I love the way you break things down. Unpacking the inside vs outside of a GenServer brings so much into focus. Thanks ❤

It's such an important detail that the calling process exits, not the overloaded GenServer.

I wish developers from "other programming environments" saw this, digested it - and fell in love with the BEAM. Thanks for educating us on these details!

Sir love your videos, keep making them

Love these

Note that the implemented function is NOT a fully faithful recreation of what actually happens inside a GenServer.call function. For that, read the source! github.com/erlang/otp/blob/ee9628e7ed09ef02e767994a6da5b7a225316aaa/lib/stdlib/src/gen.erl#L219-L279 In fact, the code that I've written as-is contains a fatal race condition (this is why you shouldn't reinvent GenServers)! See if you can find it.

I enjoy trying to extend Elixir in interesting ways, and I love me a good StarCraft reference (I've named throwaway mix projects things like Zergling, SCV, and Firebat), but I can't help but think you're looking for behaviours instead of protocols. You implement their callbacks directly in modules implementing the behaviour, and there are no hidden modules being created behind the scenes. You do lose the automatic dispatch provided by protocol consolidation, but your behaviour module can do whatever it wants in its functions before calling implementations of its callbacks.

maybe you can bring more about liveview for us.

Here for the starcraft references :-D

Alright! I dig it. So visually and ergonomically it's more akin to how one would implements in Java

Thanks for continuing this series again

When I first saw my subscription notification, I thought this was going to be a Starcraft review. :)

I often wonder how reliable libraries that plug so deeply into the "basics" are, though, and whether I'm not going to face some hard-to-figure side effects.

Beautiful, thanks for the lib!

The integration of Process.exit is so elegant.

Glad to see you back! Yeay!!

The machine code responding to the payload needs is great and all. But writing and reading this kind of code isn't the best experience. I'd rather have payloads embedded in the error types if possible.

I appreciate the workaround, but that’s all it is. With this approach you: * can return an error without setting an error payload * can set an error payload without returning an error * do not have error payload types associated with the error and are tied to one type for all errors But I believe there is a GitHub issue where the Zig maintainers agreed they are adding error payloads eventually.

I searched for this video

I was watching that primeagen's video on youtube, read that chat, and saw this video. The explanation you did with godbolt was great. Personally I think error payload is just unnecessary, at least in this particular example. The users of the function could read the document, understand the error better and handle it the way they want it. This feels like too much work.

ty

Thank you for the update. One more trick: you don’t have to pipe in bindings() because if you use dbg() without arguments it automatically uses the bindings 😉

I like `zoneinfo` as it uses OS supplied tables. Thus it is easier to keep them up to date if on something like Debian.

nice

Welcome back 🎉😄

So useful my friend I'm watching all of your videos 👏

At 29:15, why does Process.alive?(v) return true? By that time the heavy_lifting process has run Task.start() to send do_heavy_lifting and then it returned a pid, so it should no longer be alive, right?

Why is it called BEAM

Good heavens, what a detailed explanation.

Such a great demo!

Where is the next part?

Very deep and knowledgeable video. Please keep making these kind of videos in detail. Thank you 🎉

thank you👍❤

Remarkable FFI and great presentation - thanks Isaac!

Isaac you rock man

Wow, this is amazing!!! Thank you Isaac for your efforts

I'm using rustler but after this presentation I am really think about switch to zigler ^^

Man, you are a real genious you've been helping me a lot . This could be used as semaphores among cluster nodes🙌

Great video. Keep going 💪

Loved it

boom

oh snap!

Very great video. Thanks! I found that third option very clever! Although the second option is indeed very clean and I couldn't agree more with the argument to lower surface area)

Very good video! Just started to learn Rust just because of a need to run CPU heavy tasks with Rustler.. I would probably try Zig with Zigler as well (even if I'll probably go with Rust anyway just because of the sake of learning since Rust seems to be more and more relevant in the job market). Just out of curiosity, do you know if Rust memory consumption leak as well?

`cons` and `snoc` has backwards arguments probably because it's better used when piping from another method that would return the `q` which is a nice pattern actually, I pipe all the time

They did't want breaking changes so that's why they left in `lait`

So cool

Really useful and fun, thanks.