Two new features in the latest Odin release

Added: 2026-05-12

[00:00:00]A new release of the Odin programming language is out. This is the May 2026 release. Let's have a look at two new features in this release. These are the release notes and these are the two features that we're going to look at in this video. Let's start with the native array casting semantics.

[00:00:21]Previously, this is how this used to work. If you wanted to cast an array of for example free integers that looks like this then you would use lin al for example you could do this you could use lin alf32 and then you would get out an array uh that consists of uh floats instead. If we run this we see it prints v2 here.

[00:00:45]What you can do now instead is just cast it straight away. So you just type the type you want like this. F42 V1. Delete that. Run it and it just works. This is a very handy feature in Odin because you often use fixed arrays uh to represent positions or directions in two or three dimensional space if you're making a game or some kind of visualization software. And in those cases you might have a you know defined a separate type that is equivalent to this. So then maybe you would have something like uh vec 3 is uh three f32s and then maybe you have a uh integer version as well. So maybe you have something like this and then you know you could just as well just put these here. Actually you don't even need this bracket like that and you run it and it still works fine. Something to keep in mind here is that this operation is sort of equivalent to a loop. Uh it has complexity O N. So if you do this on a tiny array like this with just three elements then you know it's very cheap. But if you have thousands of elements, then you could you could see as a loop that goes through and makes a one new array of the type you're casting two and then it you know it it goes through that and casts each element. So just be aware of that that this is if you're doing this on a big array what's what's actually happening. Uh therefore you know you will probably end up using this a lot for these kind of small arrays such as the ones you use for vectors and positions. And it's also important here to to realize that this kind of cast only works on fixed arrays. It doesn't work on things that are for example slices that can be of any size and that are quite often referred to some other data in memory that's not on the stack or something like that. So you know fixed arrays in Odin they they had they live on the stack and they behave just like normal stack variables which is why the casting sort of makes sense. But if you you can't do this if it's a slice or a dynamic array or something like that.

[00:03:01]Okay, let's look at the second one here.

[00:03:04]Increase matrix limits from 16 to 64 elements allowing for up to matrix 88T.

[00:03:11]What this means is that matrices in Odin now have higher capacity. If you make a matrix for example like this you do a 2x2 matrix of 32-bit numbers and you put in four numbers into it like that and then you print it you know that that will work fine. You want three four like that. In the past the maximum here was if if we just declare a zero matrix like this. In the past the maximum was this 4x4 total number of 16 numbers. Now the matrix um like the maximum size of the matrix is 8 by8 64 numbers. Some people might be like but why can't you you know make 200 by 200 because you know there are legit use cases for really big matrices. Uh you quite often use them to solve linear uh equation systems. Note that these matrices they live on the stack so they need to be quite small like this is just a just like you know an integer lives on on the stack and takes stack space. This matrix also does that. And if you just put in 200 by 200 then you would first of all you would blow up the stacks. You would have to dynamically allocate it. And these are two kind of you know the optimizations you would want for a a matrices like you know those kind of huge matrices that solve these huge equation systems for example.

[00:04:36]You probably want different optimizations and stuff there anyway.

[00:04:40]So, so it's quite good that they're not the same. And it's the same thing, you know, if you look in C++ for example, the matrix libraries for doing like 3D math for robotics and games and stuff, those libraries are usually not the same library that you use for doing uh like mathematical analysis. You use two different kind of libraries there. And that's the same thing because they they they have these different problem domains. So so the the solution to the problem and the optimization you can put in for it is different. You don't necessarily want to mix them up. But anyways, so now we can do 8 by8 uh which is still, you know, it's a 64 uh uh values that this matrix can store instead of 16, which is still not too big to put on the stack, for example. Uh and so you could use this for example if you want a 6x6 matrix and you want to take an inverse of it. Now you can if you want to if you want to learn more about the Odin programming language then you can read my book which is called understanding the Odin programming language. It has a big chapter on for example manual memory management. So it's great if you're coming from some other language that does not do manual memory management and you want to learn about that stuff. Read more about the book on his website, odinbook.com.

[00:06:07]Okay, that's it.