egui is a simple, fast, and highly portable immediate mode GUI library for Rust. egui runs on the web, natively, and [in your favorite game engine](#integrations) (or will soon).
If you just want to write a GUI application in Rust (for the web or for native), go to <https://github.com/emilk/eframe_template/> and follow the instructions there!
If you have questions, use [Discussions](https://github.com/emilk/egui/discussions). If you want to contribute to egui, please read the [Contributing Guidelines](https://github.com/emilk/egui/blob/master/CONTRIBUTING.md)
[Click to run egui web demo](https://emilk.github.io/egui/index.html) (works in any browser with WASM and WebGL support). Uses [`egui_web`](https://github.com/emilk/egui/tree/master/egui_web).
The native backend is [`egui_glium`](https://github.com/emilk/egui/tree/master/egui_glium) (using [`glium`](https://github.com/glium/glium)) and should work out-of-the-box on Mac and Windows, but on Linux you need to first run:
If you are not using Rust, egui is not for you. If you want a GUI that looks native, egui is not for you. If you want something that doesn't break when you upgrade it, egui isn't for you (yet).
The obvious alternative to egui is [`imgui-rs`](https://github.com/Gekkio/imgui-rs), the Rust wrapper around the C++ library [Dear ImGui](https://github.com/ocornut/imgui). Dear ImGui is a great library (and the main inspiration for egui), with a lot more features and polish. However, egui provides some benefits for Rust users:
egui is in active development. It works well for what it does, but it lacks many features and the interfaces are still in flux. New releases will have breaking changes.
* Render the triangle mesh with your favorite graphics API (see [OpenGL example](https://github.com/emilk/egui/blob/master/egui_glium/src/painter.rs)) or use `eframe`, the egui framework crate.
* [`egui_web`](https://github.com/emilk/egui/tree/master/egui_web) for making a web app. Compiles to WASM, renders with WebGL. [Click to run the egui demo](https://emilk.github.io/egui/index.html).
* [`egui_glium`](https://github.com/emilk/egui/tree/master/egui_glium) for compiling native apps with [Glium](https://github.com/glium/glium).
* [`egui_glow`](https://github.com/emilk/egui/tree/master/egui_glow) for compiling native apps with [Glow](https://github.com/grovesNL/glow).
* [`egui-winit`](https://github.com/emilk/egui/tree/master/egui-winit) for integrating with [`winit`](https://github.com/rust-windowing/winit). `egui-winit` is used by `egui_glium` and `egui_glow`.
If you making an app, consider using [`eframe`](https://github.com/emilk/egui/tree/master/eframe), a framework which allows you to write code that works on both the web (`egui_web`) and native (using `egui_glium`).
* [`egui-winit-ash-integration`](https://github.com/MatchaChoco010/egui-winit-ash-integration) for [winit](https://github.com/rust-windowing/winit) and [ash](https://github.com/MaikKlein/ash).
You need to collect [`egui::RawInput`](https://docs.rs/egui/latest/egui/struct.RawInput.html), paint [`egui::ClippedMesh`](https://docs.rs/epaint/):es and handle [`egui::Output`](https://docs.rs/egui/latest/egui/struct.Output.html). The basic structure is this:
For a reference OpenGL backend, see [the `egui_glium` painter](https://github.com/emilk/egui/blob/master/egui_glium/src/painter.rs), [the `egui_glow` painter](https://github.com/emilk/egui/blob/master/egui_glow/src/painter.rs), or [the `egui_web` `WebGL` painter](https://github.com/emilk/egui/blob/master/egui_web/src/webgl1.rs).
`egui` is an [immediate mode GUI library](https://en.wikipedia.org/wiki/Immediate_mode_GUI), as opposed to a *retained mode* GUI library. The difference between retained mode and immediate mode is best illustrated with the example of a button: In a retained GUI you create a button, add it to some UI and install some on-click handler (callback). The button is retained in the UI, and to change the text on it you need to store some sort of reference to it. By contrast, in immediate mode you show the button and interact with it immediately, and you do so every frame (e.g. 60 times per second). This means there is no need for any on-click handler, nor to store any reference to it. In `egui` this looks like this: `if ui.button("Save file").clicked() { save(file); }`.
* Your GUI code can easily live in a simple function (no need for an object just for the UI).
* You don't have to worry about app state and GUI state being out-of-sync (i.e. the GUI showing something outdated), because the GUI isn't storing any state - it is showing the latest state *immediately*.
In other words, a whole lot of code, complexity and bugs are gone, and you can focus your time on something more interesting than writing GUI code.
### Disadvantages of immediate mode
#### Layout
The main disadvantage of immediate mode is it makes layout more difficult. Say you want to show a small dialog window in the center of the screen. To position the window correctly the GUI library must first know the size of it. To know the size of the window the GUI library must first layout the contents of the window. In retained mode this is easy: the GUI library does the window layout, positions the window, then checks for interaction ("was the OK button clicked?").
In immediate mode you run into a paradox: to know the size of the window, we must do the layout, but the layout code also checks for interaction ("was the OK button clicked?") and so it needs to know the window position *before* showing the window contents. This means we must decide where to show the window *before* we know its size!
This is a fundamental shortcoming of immediate mode GUIs, and any attempt to resolve it comes with its own downsides.
One workaround is to store the size and use it the next frame. This produces a frame-delay for the correct layout, producing occasional flickering the first frame something shows up. `egui` does this for some things such as windows and grid layouts.
You can also call the layout code twice (once to get the size, once to do the interaction), but that is not only more expensive, it's also complex to implement, and in some cases twice is not enough. `egui` never does this.
For "atomic" widgets (e.g. a button) `egui` knows the size before showing it, so centering buttons, labels etc is possible in `egui` without any special workarounds.
#### CPU usage
Since an immediate mode GUI does a full layout each frame, the layout code needs to be quick. If you have a very complex GUI this can tax the CPU. In particular, having a very large UI in a scroll area (with very long scrollback) can be slow, as the content needs to be layed out each frame.
If you design the GUI with this in mind and refrain from huge scroll areas (or only lay out the part that is in view) then the performance hit is generally pretty small. For most cases you can expect `egui` to take up 1-2 ms per frame, but `egui` still has a lot of room for optimization (it's not something I've focused on yet). You can also set up `egui` to only repaint when there is interaction (e.g. mouse movement).
If your GUI is highly interactive, then immediate mode may actually be more performant compared to retained mode. Go to any web page and resize the browser window, and you'll notice that the browser is very slow to do the layout and eats a lot of CPU doing it. Resize a window in `egui` by contrast, and you'll get smooth 60 FPS at no extra CPU cost.
There are some GUI state that you want the GUI library to retain, even in an immediate mode library such as `egui`. This includes position and sizes of windows and how far the user has scrolled in some UI. In these cases you need to provide `egui` with a seed of a unique identifier (unique within the parent UI). For instance: by default `egui` uses the window titles as unique IDs to store window positions. If you want two windows with the same name (or one window with a dynamic name) you must provide some other ID source to `egui` (some unique integer or string).
`egui` also needs to track which widget is being interacted with (e.g. which slider is being dragged). `egui` uses unique id:s for this awell, but in this case the IDs are automatically generated, so there is no need for the user to worry about it. In particular, having two buttons with the same name is no problem (this is in contrast with [`Dear ImGui`](https://github.com/ocornut/imgui)).
`egui` is a 2D user interface library for laying out and interacting with buttons, sliders, etc.
`egui` has no idea if it is running on the web or natively, and does not know how to collect input or show things on screen.
That is the job of *the integration* or *backend*.
It is common to use `egui` from a game engine (using e.g. [`bevy_egui`](https://docs.rs/bevy_egui)),
but you can also use `egui` stand-alone using `eframe`. `eframe` has integration for web and native, and handles input and rendering.
The _frame_ in `eframe` stands both for the frame in which your egui app resides and also for "framework" (`frame` is a framework, `egui` is a library).
### Why is `egui_web` using so much CPU in Firefox?
On Linux and Mac, Firefox will copy the WebGL render target from GPU, to CPU and then back again: https://bugzilla.mozilla.org/show_bug.cgi?id=1010527#c0
To alleviate the above mentioned performance issues the default max-width of an egui web app is 1024 points. You can change this by overriding the `fn max_size_points` of [`epi::App`](https://docs.rs/epi/latest/epi/trait.App.html).
egui can't do 3D graphics itself, but if you use a 3D library (e.g. [`glium`](https://github.com/glium/glium) using [`egui_glium`](https://github.com/emilk/egui/tree/master/egui_glium), or [`miniquad`](https://github.com/not-fl3/miniquad) using [`egui-miniquad`](https://github.com/not-fl3/egui-miniquad)) you can render your 3D content to a texture, then display it using [`ui.image(…)`](https://docs.rs/egui/latest/egui/struct.Ui.html#method.image). You first need to convert the native texture to an [`egui::TextureId`](https://docs.rs/egui/latest/egui/enum.TextureId.html), and how to do this depends on the integration you use (e.g. [`register_glium_texture`](https://docs.rs/egui_glium/latest/egui_glium/struct.Painter.html#method.register_glium_texture)).
There is an example for showing a native glium texture in an egui window at <https://github.com/emilk/egui/blob/master/egui_glium/examples/native_texture.rs>.
egui uses the builder pattern for construction widgets. For instance: `ui.add(Label::new("Hello").text_color(RED));` I am not a big fan of the builder pattern (it is quite verbose both in implementation and in use) but until Rust has named, default arguments it is the best we can do. To alleviate some of the verbosity there are common-case helper functions, like `ui.label("Hello");`.
Instead of using matching `begin/end` style function calls (which can be error prone) egui prefers to use `FnOnce` closures passed to a wrapping function. Lambdas are a bit ugly though, so I'd like to find a nicer solution to this.
The one and only [Dear ImGui](https://github.com/ocornut/imgui) is a great Immediate Mode GUI for C++ which works with many backends. That library revolutionized how I think about GUI code and turned GUI programming from something I hated to do to something I now enjoy.
*`NotoEmoji-Regular.ttf`: [google.com/get/noto](https://google.com/get/noto), [SIL Open Font License](https://scripts.sil.org/cms/scripts/page.php?site_id=nrsi&id=OFL)
*`Ubuntu-Light.ttf` by [Dalton Maag](http://www.daltonmaag.com/): [Ubuntu font licence](https://ubuntu.com/legal/font-licence)
