the new iOS Target for compose multi-platform means that you can take your mobile kotlin skills to new destinations and to the next level. Because it means that you can start building your mobile applications in 100 kotlin, including the UI. The iOS Target for composed multi-platform is currently in Alpha, which makes it the perfect time for us to take a closer look at it and see what it's all about. And I thought, what better way to take a closer look at the technology than building a small app from start to finish without sweating any details and looking at everything that goes into building a proper multi-platform app. The end result of this video is going to be a small working application that works on both iOS and Android, which will show images of different types of birds that we've consumed from an API. And throughout, we will see how we can even share the UI layer with compose multi-platform, whereas previously with kotlin multi-platform, you could already share most of your business logic and pretty much everything up to the view layer in your application. Compose multi-platform.

ads that missing piece and allows you to also share the user interface for your apps. So let's get started and write some actual code. The easiest way to get started with Compose Multi-platform is to use one of the templates we provide. The easiest way to access those templates is to go to jb.gg/compose, which will bring you to the Compose Multi-platform landing page. From here, if we scroll down and click the iOS link, it'll actually bring us directly to the Compose Multi-platform mobile application template that targets both Android and iOS. It comes with detailed instructions on how to set up, but most things we'll tackle today along the way. Since this is a template, I can just use the use this template button to create a new repository. I'm going to call this one my Bird app. It's okay to have this as a public repository because, as usual, you'll be able to see the source code after we're done with everything. Let me create this repository, and after a few short seconds, I'll have my own repo where my source code can now

live. So now that I have a, we can move over to Android Studio. We're pretty much going to spend the rest of this video in Android Studio. So since our project is on GitHub, we can just click the Get from VCS button, add the URL to our repository, and let it clone. I'll give the project a couple of seconds to synchronize, and then after that is done, we can start exploring and developing. But before we do that, I want to make sure that my development environment is set up correctly. So let's check a couple of things. So to check that my development environment is set up correctly, I can run KDoctor. KDoctor tells me that my operating system, Java, Android Studio, Xcode, and CocoaPods are all set up correctly. If I were missing any of these things, KDoctor would tell me how to fix it. A small reminder that since we are developing for iOS, you do need a Mac to run the simulator and deploy on actual targets. If you want to build a multi-platform application that targets both iOS and

Android. That's just important to note in advance, okay? I also want to make sure that in Android Studio, I have the Kotlin multi-platform plug-in installed. This is the piece of software that makes it possible for me to run iOS run configurations from Android Studio as well. So, first things first, let's close this README file. And because we are developing multi-platform, the Android view doesn't actually show us everything we want to see. So, let's switch to the Project View. And here we can see the whole structure of our project. We have an iOS app, which is a real Xcode project or a real iOS app. So, this is the part that will then actually get deployed. We also have an Android app folder in which a regular Android app is defined in its regular shape, including things like the Android manifest configuration files. And then in shared, we find all the logic that is actually targeting both platforms. If we click ourselves through, we can actually see that inside the app KT, we find a bit of a scaffold for a first Compose multi-platform app.

we can start this app just to see what we're working with. The Android, uh, run configuration is already configured to run on my emulator, so that's perfectly fine. I can also take a look in my iOS app run configuration, and I can make sure that this one runs on an emulator as well, and we can run that one too. After a short build time, we can see that we have the same app running on iOS as well as on Android. It doesn't do much besides show us a target-specific string up in the button text as well as do this nice little animation of the Compose Multi-platform icon. But of course, the most notable part of this whole story is that both the Android and the iOS app render a consistent user interface that has only been described in code once. We didn't have to write two separate user interfaces here. In fact, everything that you see on the screen right here is defined in this composable. If you've done any modern Android development recently, then I'm sure these APIs will feel quite familiar to you because

they're the exact same ones that you're used to from Jetpack Compose. And if not, you're in luck because the Compose APIs follow pretty much the same principles that all modern UI frameworks follow: they're declarative APIs with some sense of state management, and functions as little blocks that can be composed together to build more complex user interfaces. In this example, the greeting text and the show image variable, which determines whether the image should be shown or not, are stored in remembered mutable states, which is just the Compose paradigm for storing state. And then, this little app uses layout primitives like columns together with modifiers and other settings to lay out the whole application. The button is responsible for actually modifying parts of our state, and the animated visibility clauses are responsible for doing that nice little animation that happens when I click the button. So, these APIs are the exact same ones that are already being actively used in modern Android development to ship apps. In fact, the only part of this code that is Compose multi-platform specific is the way that we read.

resources because we need to provide an implementation that works on both Android and iOS, but that one is included with the Compose Multi-platform framework. And as you can see, it even reads XML-based vector graphics just fine. Well, this is all well and neat, but it's of course still far from a useful app. So, I think it's time we bring in some actual data and some useful stuff to our app. So, since my idea was to show pictures of different birds in the app, I actually prepared a small API. It's nothing fancy and hosted on GitHub pages, and really it just consists of one large JSON file. And in that JSON file is an array of objects, each object representing an image as well as an author and a category for that image that identifies the bird. And then we can take this path, and if we replace the pictures part with that path part, we'll actually see the retrieved image. So, I'm thinking we move on and we bring some of this data right into our app. But let's not go overboard, and let's

start simple and just show a single image from this API in our app. Now, just like in Jetpack Compose on Android, in Compose multi-platform, we would use an image loader library to bring in images from the network. Image loading libraries range from very simple to very sophisticated and can introduce things like caching or asynchronous loading for your application. A number of image loader libraries already exist that also support Compose multi-platform. One of them would be Compose Image Loader, for example, and the other one is Camel, which I believe stands for Kotlin Asynchronous Media Loader, at least that's what the README suggests for this project. I'm going to use Camel because it has a super simple API and quite sensible defaults, so I don't actually need to make many changes to get started, but of course, you're welcome to use whichever one you like. Just make sure it supports Compose multi-platform. So first things first, let's add a dependency to our project. We can see that Camel has a nice little dependency snippet down here that we can just copy. We'll open the Gradle file that's

part of our shared holder. And just like in any other Kotlin multi-platform project, we can go to the common main dependency section and just add it here. This library is Kotlin multi-platform compatible, which means I will be able to use all of its functionality on both platforms that we target. And most importantly, in the common UI code that we're currently writing. Under the hood, Camel uses Ktor to make the actual network requests. So, we need to add a couple of dependencies for that one as well. The README file of Camel sends me over to the engine section on Ktor. But in fact, I first want to add my core dependency for Ktor client. So, I'm going to add that one to my common block of dependencies. I'm going to hard code the current version number 2.3.1 for Ktor here. But you're of course welcome to use anything from properties to version catalogs to organize the way that you manage your version numbers. Now, for each specific target, we also need to have an engine so that Ktor knows how to make the

actual requests we have two Targets Android and Native so let's click through those thankfully these are also quite easy Snippets. we have the ktor client Android dependency which we will just add to our Android main source set. that's that and we will once again just replace the version number. and we need to add the Darwin implementation which is for darwin-based operating systems such as Mac OS iOS and tvos. which covers Us quite nicely. we don't actually have any iOS dependencies yet so we need to open that dependencies block ourself. but once we have that it's as simple as copy pasting and adding the correct version number. now we can just quickly sync our project and get everything in great. so now that this is set up we can actually already use camel and a switch to camel and loading an image from the network is actually quite quick. instead of an image we'll use a camel image and instead of a painter resource we'll use an async painter resource. which now takes a URL. I can just go back to my API and grab.

myself one of those strings and grab myself this cute picture of a couple of pigeons. And we'll just hard code that for the time being. And let's try running this on iOS. Now, if I click the hello world button, it's indeed going to show me the image of the pigeons. Isn't that cute? It's probably also a good time to take a second and add a proper content description, uh, just to make our things accessible. Okay, if we try to run the same app on Android, we're actually going to see something that might be a little bit surprising. Indeed, the app runs fine, but the moment we try to load an image, nothing happens. And that's because while on iOS we can always call into the network, Android requires explicit Network permissions. And with Compose multi-platform projects, that's no different. But luckily, we can just look up Android network permissions on a search engine of your choice, and you will very quickly find the snippet that is required to add to your manifest the permissions for internet and access network state.

once we have those in our clipboard, we can just move to our Android app. Go to the Android manifest file and add that one above the application block. Now, after a quick restart, we can once again press our button, and we will see a wonderful, cute little pigeon on our Android device as well. Neat! So, already with those minor changes, we were able to use third-party composables in our application to pull in network images. That's kind of cool, but of course, our API has a little bit more to offer than just a single pigeon image. So, let's go ahead and actually use the rest of this API, consume that, and actually parse out the JSON objects in here. Now, whenever I have a task for JSON parsing, the first thing that comes into my mind is Kotlin X serialization. It's one of the libraries built by the Kotlin team, and it's fast, it's Reflections, and most important for us, it is multi-platform. So, we can quickly set this one up, and then we'll be able to read information from our API in no time.

and give our pigeon a couple of friends. If we scroll down to the setup guide, we will see that Kotlin X Serialization comes in two parts. One of them is the Gradle plugin. That one we can just add at the top of our build file. And the second part is the runtime library, for which you have to scroll down just a little bit and find it right here. This one is a common library, so I can once again add that one to the common source set. But the ideal situation would be for Ktor, which we're going to use to make our HTTP requests, and Kotlin X Serialization to kind of click together and work together. So maybe let's also add the integration for that part. If I head on over to the documentation under the Ktor clients, there's a section called content negotiation and serialization. There are two pieces we need to add to our build file. One of them is the content negotiation itself. So that one we can add right here to the common block. And we'll once again provide the correct version number, as well as the integration for.

serialization with kotlin axialization. That's the second line, and we'll add that one as well. And just quickly hard-code the version number so we can sync this one again as well. Very good. So with that out of the way, we can actually move on and start writing the business logic that is responsible for making that network request and parsing out the individual JSON elements. First step in that case for me is always to create a model so that I have a type-safe representation of whatever my JSON information contains. So let me just add a new model directory. The cool thing about Kotlin X serialization is that you can just use very simple data class definitions, annotate them with @Serializable, and the compiler plugin and library will figure out the rest for you pretty much. But I am see, so I like installing the JSON to Kotlin class plugin, and that one actually provides me with a new option when I create a new file, which is Kotlin data class file from JSON. So what I can do now is I can go back to my API, I can grab myself one of these

object definitions. I can give it a class name like bird image, and I can just hit generate. We can add this one to git, of course, and we already get the data class. Now all that's left is annotating this one with @Serializable from the kotlinx.serialization package, and serialization works. That's all there is to it. We can actually write a short snippet of code to convince ourselves that what I just said actually works. We're going to ignore app architecture for just a moment, but don't worry, we'll get back to that in just a second. So we can just write a small function called suspend fun get images, and that one is supposed to return a list of bird images. To make the actual HTTP request, we of course need an HTTP client, so we construct one from the Ktor Constructor function. And since we want to integrate it with serialization, we need to do a little bit of configuration. Luckily, the Ktor DSL for setting up an HTTP client is very simple. We can just use install and then install, for example.

the content negotiation plugin and instructor to use Json. And just like that, we have an HTTP client that knows how to do content negotiation and deserialize Json using kotlin X serialization. So now we can actually make a get request. So we can say our images that we're getting is HTTP client dot get, and this one can just take a URL. We are going to hard code this one for the time being. Let me slightly reformat this code. And then out of this one, we want to get the body, but we want to have it in a specific type. And here we essentially tell kotlin X serialization how it should attempt to deserialize it. If we look at the format once more, we'll realize this is an array or a list of these kind of objects. So we'll just say give us a list of bird images, and that's exactly our return type as well. So we can just return this value. In fact, Android Studio rightfully tells us that we could already inline this variable and probably even make this an expression body function, but we'll keep

it simple for now because I don't know how this function will change just yet. So to convince ourselves that what I just said actually works, it would be nice to actually run this kind of function. However, we have to be a little bit careful even with this proof of concept because generally we don't call arbitrary business logic functions inside composables. And why is that? Well, because the Compose framework itself actually reserves the right to call your composable functions at pretty much any time, like for example when State changes and it has to re-render. But luckily, there is a mechanism to execute side effects, which is essentially what we're doing here. So we can use the `LaunchedEffect` builder function to specify our call to get images. `LaunchedEffect` does take one parameter called the `key`, which is the variable that has to change in order for the `LaunchedEffect` to rerun. In our case, we don't really want to rerun our `LaunchedEffect` ever, so we can just by convention use `Unit`, which is a constant that never changes. So this will only

ever be executed once. So let's rerun this application and take a close look at the logcat output. And if we take a look, poking around just a little bit, we can see that indeed we have printed a number of bird images with their author, category, and path. And there's a whole bunch of them. So it seems like we have indeed properly parsed the JSON that we've received from our tiny API. Now from here, we could of course go on and try to wire all of this together with the top-level function and just having an HTTP client kind of floating around in the middle of the file. And it would probably work because this is a tiny app. But in reality, you probably structure your application in some kind of architectural pattern where you separate the logic that is responsible for your UI and your business logic. Those are patterns like MVVM, MVI, BLoC, all of those. For example, they all serve to separate and encapsulate your business logic and your UI logic from each other. There are libraries like PreCompose, which

provides multi-platform navigation and view models, kind of inspired by the Jetpack lifecycle. And there's Voyager multi-platform Library built for Compose that also comes with a ton of different features. One of the more simple and easy-to-use libraries is Moco MVVM. Because, granted, even with some architecture in place, this app is going to stay quite simple. We're not going to have any fancy navigation logic or any other bells and whistles. So I'll just use Moco MVVM for this example because it's a simple implementation of the view model pattern. But if you're looking to build a more complex app, you're more than welcome to check out all the other libraries that are out there as well. They all come with their own spins and twists and oftentimes provide a really unique and useful approach to structuring your application. So to use Moco MVVM, I need to add two dependencies to my application. I need the core dependency, which is just an API dependency, and I need the Compose multi-platform specific implementation as well, which is a second API dependency.

let me just do another quick sync and add these to our project. Great, with those in place, we can create a new class called BirdsViewModel, which is going to encapsulate the state and business logic for our application. The first thing I'll do is make sure that this class indeed inherits from ViewModel. This gives us two important parts. First of all, it gives us access to a ViewModel scope, which we can use to launch and manage coroutines, and it also provides us lifecycle awareness on Android so that we can do things like release our resources when the ViewModel is no longer needed. Great, so let's move our state management into the ViewModel instead of just having it be spread around in our app KT file. So, a typical pattern for a ViewModel is that it defines some kind of class that is responsible for representing the UI state. So, I would call mine, for example, BirdsUIState. In our case, we could start very simply and say that well, the images that we've loaded, which is a list of bird images, is

our UI State. And indeed, when this starts out, this is just an empty list. Then the view model goes ahead and exposes this UI State in a way that is observable from the view or the UI side of things. One modern approach to do that is to do it via a StateFlow. So, we can define a private field called underscore UI State, which happens to be a mutable StateFlow. And we'll import this function off of Bird's UI State. And by default, this is just a virtue I state with no function invocations. Let me close the side panel as well. Then we somehow need to expose this StateFlow in a read-only fashion to any consumers of this Bird's view model. Let me give us some extra space here. So, we'll just specify a public field called UI State, which is on our underscore UI State as a StateFlow. So, this way only members of The Bird's view model can actually update the state, but external consumers like our Compose-based user interface can still read the UI State as it gets changed. And indeed, we are going to collect this exact UI.

State variable later on from the compose site, but before we do that, let's finish moving our code into the view model. So, we have this HTTP client, which we no longer need to be a top-level declaration. Instead, that one can live inside the Birds view model. Likewise, the suspend function get images can also be a part of the bird's view model. And let's make sure that we get all of our imports in here as well, properly of course. The only part of our system that should actually have access to this specific HTTP client is the Birds view model. So, we can make this one private as well. And when we think about it, we should probably also make the get images function private. The reason for that is while this get images function, at least in its current signature, returns a list of bird images. But remember that the communication paradigm for this view model is that there might be some kind of external impulse, and in response to that impulse, the state flow gets changed. And then any external observer that collects this UI state state flow will

then be notified of any kind of changes. So rather than returning any kind of values directly, we would rather update the internal state of our view model. And indeed, that is something that we can do by providing a function called updateImages. And that one is explicitly public. And the task that this one has is to get the new images and then update the UI state so that the changes propagate throughout our system. However, getImages is a suspend function because Ktor makes its HTTP requests on top of the coroutines machinery so as not to block any threads but just to suspend them. However, if you look into the developer guides for Android developers, one of the best practices that are mentioned is that the view model should create coroutines. So something that should be avoided is exposing suspend functions directly and rather than that, calling suspending functions within the view model scope using one of the coroutine builder functions as the documentation lays out. This makes your business logic easier to test because it means your view model objects can be unit tested.

we can follow this best practice quite easily because MoCo MVVM, just like pretty much all the other view model implementations out there for multi-platform, provides us with access to a view model scope, which is a coroutine scope that is lifecycle-aware. On Android, inside this launch block, we can now grab our images by calling the `getImages` function, which was private, and then updating our `_uiState`, which is the mutable variant of our flow. So, we can just say `uiState.update`, and then inside this lambda, we receive the original UI state, and we can just make a copy of the original state, but we will just set the images to be the images that we have received from our `getImages` function. Now, this view model really can't do much unless the `updateImages` function is actually called, so it might make sense to call `updateImages` once in its initializer. That means when this view model is first instantiated, we will also retrieve a new copy of all the images that our API has to offer. And of course, later on in the

life cycle of our application. If the user presses like a refresh button or something like that, that we might add to the UI, then that would just go through and also trigger the update images function in here, which in turn would then at some point update the UI State. While we're already talking about what's happening when this view model is created, we might as well also tackle the part where this view model is destroyed. For example, when it's no longer needed, for that we can override the on clear it method. Whereas good platform citizens were expected to release any kind of resources that we still have, in our case we would close the HTTP client so as to release any kind of threads or other health resources that might still be in there. Great, right? So we have a basic view model setup. Now it's time for us to go back to our actual UI and wire things back together because now our API offers us a nice list of different images. So I think it's time we show them in a nice little grid. So first things first, I

think it would be useful for us to get rid of all the code that we've been using to try things out so far because that was mainly related to just our little example. And we'll start from scratch. We will make sure that we have a separate composable function, which I'll just call Birds page, which then represents all the composables that are somehow associated with both the view model that we've just built. In fact, the Birds page should get the bird's view model as a view model parameter so that it can access these things. Now in our app composable, which is kind of the root composable of our whole application, we want to call the Birds page composable. And for that, we of course need an instance for the view model. Moko mvvm has a convenience function called get view model, so we could grab ourselves a bird's view model by calling get view model. But of course, if you're using any other kind of mvvm framework, they usually have a similar compose multi-platform integration for you to

use the first parameter is once again a key that determines when this view model should be recreated. In our simple case, there really never is any such situation so we can just pass unit again. The second parameter is a factory function that creates that view model. Luckily, there is a view model Factory Lambda included in which we can just pass a call to the actual Constructor, and that's it. Now we can grab that birth view model and pass it down to our Birds page, and already with that, we have the view model available inside of our Birds page composable. And just like I mentioned earlier, we can now actually observe it. In this case, we collect its state, so we'll say our UI state is looking at the view model's UI state. If we collect it as state, if we're using the equal sign, you can see that the UI state is indeed a state object that wraps the bird by state. There's a little bit of syntactic sugar we can use buy instead, which unwraps that state object for us and provides us with calls.

directly to Birds UI state, which is a little nicer. Great! So, from here, we can actually start working on our display logic and start building some layouts. Because it's a framework that's already out there and being used, it also comes with a number of layout primitives that you can use straight out of the box. So, I have a couple of considerations. Let's start easy. When the images are first loaded, I would like to show them with some kind of animation. And this part we've already kind of seen in the beginning in the template. We can use the animated visibility API. And because Compose multi-platform's modifiers and composables are exactly the same ones as the ones on Jetpack Compose on Android, we can actually just have a quick look at their documentation right here and take a look at how animated visibility is used. And we'll actually find out that it only takes a single parameter which determines on whether the element should be visible or not, and then just takes a block of composables that specify the content. And of course, there's a bunch of customizations that

we could do, but that we're not gonna touch today. So, if we specify ourselves some animated visibility, it becomes clear that whether we can show something or not depends on whether our UI State actually has some kind of elements. So, as long as the images list in the UI state is not empty, we want to show whatever composables are in here. Now, we have a lot of images to show, so I want to show them in a visually pleasing manner, so probably a grid. And I also want to show them with performance in mind, so ideally, I would like to lazy load the images as they come onto the screen, rather than pre-loading everything and, you know, filling up my memory and drawing all that bandwidth. Now, thankfully in Compose multi-platform, we have a lazy vertical grid that pretty much perfectly fits this use case, and it also comes with a bunch of customization options that we can make use of. We can hit command P to see a couple of the different parameters that this one accepts, and we can start filling them in. Well, I know how many

columns. I want, uh, pretty much, I just want to have a fixed number, which is just two of them, uh, I kind of want to have some consistent spacing between them. So, I'll provide a horizontal Arrangement, which is an arrangement spaced by 5dp, and we gotta make sure that we import that extension property as well. I probably also need a vertical Arrangement, and to keep it consistent, we'll do the same kind of spacing with 5dp as well. So, this defines the spacing between the elements, but I also want to make sure that on the sides, we get the padding as well. So, when I specify the general modifier for this element, I first make sure that this fills the maximum available size, but I also want to make sure that it provides a horizontal padding of 5 DP as well. And then the last parameter we need to provide is the content. So, the content is everything that's supposed to be rendered inside this lazy grid. We already have a collection of items that we want to render, and we do that.

precisely using the items function, specifically the overload that uses the list. And this one, we can just feed the UI State images. You can kind of think of the items clause as somewhat of like a map function. So inside the Lambda that follows the items declaration, you specify how each individual item in the lazy vertical grid should actually be represented visually. In our case, I'm thinking we create a custom new composable called bird image cell, which just takes the definition of that bird image and then renders it out. Great, and we can actually just move on to that bird image cell next, which once again we specify as a composable function. We'll call it bird image cell, and we know that it takes an image, which is our bird image. Now, this code is kind of a repeat exercise of what we already had in the beginning. It's pretty much just a camel image, which takes an async painter, which source. But this time, instead of hard coding the entire URL, we only have to hard code the part that kind of defines

our API endpoint, and then everything after that is defined by the image's path. And of course, we will once again provide some kind of image description, which could, for example, be the category concatenated with the image's author. Because I want to make sure that this image looks good no matter what the input is, we can make a couple of smaller adjustments. For example, we can set the content scale to be `contentscale.crop` so the image gets cropped to the right aspect ratio. And then, using the modifier, we can indeed say we want the element to fill the maximum available width given an aspect ratio of one, so constraining it to a square. Great! So with that, we have reached a point where our application should compile and work just fine again. So let's run it here, and indeed on Android, it runs nice and well. And we can see that as we scroll, new images are loaded in at the bottom, and we see all the images that we have available in our API. And here's the magical part: we can run the same app on iOS as well.

works just as nicely with the images coming in, and us being able to scroll up and down, just like we defined it in our app's logic. Both the UI once and then used it twice, and all in 100 kotlin. That's already pretty good if you ask me. Alright, so I'm already enjoying scrolling through these images of these Majestic Birds, but I would love for the app to be slightly more interactive still. And if we look back into our API, we'll actually realize that we have just the information for that, which is the category. So with that, why don't we add a couple of buttons to our app that allow the user to filter which birds they would like to see. And to do that, let's first actually build out a little bit of business logic, which of course means moving over to our view model. This specifically, there's two things we need to do. We need to add the information about the categories and the selected category to our UI State class. And we also need to be able to somehow react.

when the user selects one of the available categories. Okay, so an interesting observation is that we actually don't need to manually set the categories. We can derive them from the list of bird images that we have. So we can just say that the categories are our images mapped to their categories and then exposed as a set because these are distinct. However, what we do need to store is which of the categories our users actually selected. So we'll provide a selected category field, which is just a potentially null string, which by default is null because you probably won't have anything selected. So once a user has actually selected a category, we can derive one more field, which is the list of selected images. So we can just say `val selected images` and that's our images but filtered where its category equals the selected category. So that's our update for the UI state with the information that we need to store and information that we can derive. However, the view model still needs to be able to react to the event of the user selecting a category. And for that, we can just

provide a short little function called select category. This one is, of course, public because it will be called from the UI. And all this really does is it takes our UI State and it updates it. And the only change that it makes is it takes the selected category and sets it to whatever category was just selected by the user. And that is all there is to it. Thanks to the magic of State flows, this is going to automatically propagate properly throughout the app. So now that we have this information exposed and we also have a way of reacting to the selection of a category, we can actually go back to our UI and start adding some category buttons. So the layout I'm thinking about is buttons on top and then the grid below it. So essentially what I'm looking at is a column layout on the top because we have a column of multiple elements. And then inside, for the button specifically, we have a row that has a number of individual buttons kind of side by side. So let's introduce this kind of organization for our Birds page composable. We'll wrap the whole

part in a column, so the closing brace for this goes after our animated visibility. And we want to make sure that this column does indeed fill the maximum size of the screen because it's our whole application. Uh, we'll also set the horizontal alignment, uh, to center things horizontally. That's just going to make things hopefully a little bit nicer. And then the vertical arrangement, uh, will also provide some centering just to keep things nice and packed. Now, if we were to rerun this application, I can just do that very quickly on Android here just to demonstrate it. You will see that nothing much has changed because we've just wrapped it in a column that has the full size. So let's actually add the buttons, and those are organized in a row. Let me just import that function. For this row, we want to make sure that it does fill the max width. We want to provide the same kind of 5 DP padding, which we've used all throughout our whole application. And since we have multiple elements, it's probably interesting for us to specify a

horizontal arrangement. Specifically, I'd like to do a 5dp spacing again between the individual buttons. And inside this row, we just want one button for each category. And in Compose, that couldn't be simpler. We can just do a for loop. So we can just say, for category in UIC dot categories. And then for each one of those, we'll specify a button. That button mainly does two things. It has an onclick handler that will get executed once the user actually clicks. And in the content, this will just specify some text, which is just the category title. Unfortunately, we've already laid all the groundwork for the onclick handler. So we can just go to our view model and say, select category named the same as the button. Okay, so we could already run this, but let's make sure that our grid that we have at the bottom is actually also aware of the selected images. Because right now, it's just showing all of them. So we can just replace the animated visibility conditional, as well as the content down here with selected images.

instead of the regular images. And now it will use only those images that have been filtered by the category. So let's rerun our application, and let's do that on both platforms. And just like that, both on Android and on iOS, we have category selectors. We can click them, and they will switch out the images. If we scroll around, we can see only the images from a specific category. So if I grab myself, for example, the Eagles here, you can also see just the Eagles, or just the owls, or just the pigeons. Isn't that neat? Now, however, I'll admit that our app could use a little bit of extra polish. These button elements that we have here by default come from the Compose Material Design components Library, which on the one hand, first of all, might look a little bit foreign on iOS devices. But they also just kind of don't mesh that nicely with the rest of our minimalist and very blocky look. Now, of course, what you can always do is you could redesign these components using layout primitives like boxes or use some

kind of third-party UI components library that just have a different style, but at least for this simple app, we can take an easier route, and we can do a couple of customizations because we can do things like change the material theme, as well as change the parameters for the individual elements that we have in our app. So let's kind of look into what that would look like. The first thing I want to do is Define a custom theme for our application, and the general pattern that themes for compose follow is that they are composable functions themselves. For example, we could call this one the Bird app Theme, and essentially they are kind of higher-order components, which means they take content, which themselves is a composable function that doesn't return anything, and then they would just call this content. Usually we use the material theme as the base for our theme, so we'll just use material theme and then content, and we can switch out the invocation of material theme down here with our Bird app. Now we have a convenient and encapsulated point where we can adjust.

this material theme and where we can adjust things like colors, typography, or shapes. The adjustments that I want to make for our app here are in the color and the shapes. First of all, I want to make sure that our buttons are in a sleek black. So I can just go in and set my colors to be the original material theme colors, but will make them a copy, and we'll make sure that the primary color is indeed black. And now that they're in a sleek black, I also want to make sure that they use very sharp edges so that they're not kind of rounded buttons, but instead that they're just rectangles. So I can specify shapes as the original material theme shapes, but we'll copy those. We can see that there's three different shapes here: small, medium, and large. And for all of them, I'm going to set an absolute cut corner shape with a size of 0dp, which essentially says this is a rectangle cut off the edges but only for zero DP, so don't cut them at all, which just means they're

rectangles. We're going to have this for each one of these variants. I think I can probably add the names to the call arguments here to also just make that clear in code. Now when we you rerun this, we can see that our buttons are nice and sharp, properly rectangular, but the buttons aren't really taking up all the space yet that we have available, which makes them look kind of wonky. So we can fix that by taking a look at the individual buttons and specifying a modifier. In this case, we just set them to be an aspect ratio of 1.0, which means they are square. We'll make sure that they fill the max size that is available to them and will make sure that all of them use the same amount of weight so that they kind of have an equal distribution between them. And just like that, we got some nice big squares that sit at the top and mesh quite nicely with all the paddings to provide us with a quite uniform design if you're squinting really hard and looking quite.

detailed you'll see that there's still the tiniest of shadows under the buttons which maybe doesn't look exactly the same as the rest of our rectangles. We can indeed even address that part if we would like to. Let me just reformat that for a second because there is indeed an elevation property on our buttons, and we could, for example, set those to be the button defaults elevation. But we will set the default elevation to zero DP and the focused elevation to also be zero DP. And with those properties reset, no matter if we run it on iOS or on Android. I'm just going to switch back and forth here. You will see that even when we take a close look at the buttons, there are no shadows, and they look quite nice and cubic. And there we have it, our little bird app is done, and it's looking nice, sharp, and rectangular. And most importantly, it works great on both Android and iOS, even though all the code we've written was purely kotlin multi-platform and shared between the platforms. Now there's obviously a bunch

of different things that we could still do to improve this app. Instead of just doing a bunch of filtering, we could use a proper navigation library like Voyager. Pre-compose or decompose. That would also allow us to have separate views that show us, for example, specific detailed information for each one of those bird pictures. We could use a repository pattern so instead of having the view model make the actual network requests, we could outsource that to a repository. And that would also give us a layer where we could do things like play with persistence, for example, or introduce some kind of caching layer, for example. Of course, we could also further decompose our composables here and instead, for example, provide a custom composable just for our button or maybe for the selector for the different individual categories. Just to kind of break things up into reusable smaller pieces. And I am sure you already have about a dozen ideas of how this code could be changed, improved, or modified to suit your needs even better. But I think we have done enough for one session. And looking back at it, it's been

quite a lot. We've seen how to just set up a compose multi-platform project from scratch, how to briefly validate that our development environment is indeed ready for compose multi-platform, how we can use third-party libraries and composables like the Camel image loader, like Kotlin X serialization together with Ktor, and we've essentially built a whole application that can interact with the outside world and still write pretty minimal and nice Kotlin code. We've also seen how we could use something like an MVVM Library together with State flows to separate out the business logic from the UI logic that we have here, and we've used a couple of the built-in components that are available not just in Jetpack Compose and Android but through the power of Compose Multi-platform on iOS and other platforms as well, like for example lazy grids, and we designed our user interface to be neat and cross-platform with even a smidge of customization built in as well. I do hope you enjoyed this journey through Compose Multi-platform with me, where we take a small project from start to

finish and maybe you've learned something new. As mentioned previously in the beginning, Compose multi-platform support for iOS is currently in Alpha, so we're really excited to hear your feedback and see you experiment with it and see what's possible. If you want to learn more or play around, make sure you go to jb.gg/compose or just look for Compose multi-platform where you can find all the information on our wonderful declarative UI framework that works on iOS, Android, and even desktop in the web. So thank you so much for tuning in, and I do hope that your Compose multi-platform story has only just begun. Alright, see you in a future video. Everybody, take care, thank you. [Music]

Build an iOS & Android app in 100% Kotlin with Compose Multiplatform

Expand your Kotlin skills with the official YouTube channel of the Kotlin programming language. 
Kotlin Tutorials, Kotlin Multiplatform, Kotlin Coroutines, Kotlin Tips, Talking Kotlin Podcast, ATOM Podcast.

Kotlin is a concise multiplatform language developed by JetBrains.

Subscribe for shows, the latest news, and live events!


Build an iOS & Android app in 100% Kotlin with Compose Multiplatform

Build an iOS & Android app in 100% Kotlin with Compose Multiplatform

스크립트