fm.div("vert <><<><weight=80% text><>><><weight=24<><button><>><>");
Wow, that layout string looks awful. There’s a visual description of what it means here, but still, oof.
Wow that is garbage.
The one question I have about any cross-platform GUI library is whether it’s using the native platform APIs or doing its own rendering, usually by means of OpenGL (aka “immediate GUI”). I didn’t find an answer on the website. So far, I only know of exactly one real cross-platform C++ GUI library that uses the native APIs, which is wxWidgets. Most libraries I have encountered promise cross-platform and then use the “immediate GUI” approach, which will never look really native and is a battery killer in my experience.
Looking at the source for the graphics class and the button class, looks like they’re just drawing their own widgets via Windows APIs that I’m not familiar with or Xlib (the X11 drawing library). Xlib is an interesting choice; most GUI libraries I’ve seen go for OpenGL, which gives you a lot more control, and isn’t slated to be replaced by Wayland. To your point, it might be friendlier on battery than an OpenGL implementation, since it’s not doing any drawing in immediate mode (and instead looks to be repainting into X11 buffers when needed).
Personally, I’ve given up trying to make the GUIs I work on feel native, especially since the GUIs I’m writing are usually for internal tools or for my own consumption. I default to using Dear ImGUI, which makes GUIs that are very utilitarian and look nothing like native, and is super super nice to use as a programmer.
My solution is to just write a native UI for each platform you care about. For me, this basically means I’ll just use Windows, and just keep a rudimentary GTK# frontend working, primarily so logic can be split out from views when reasonable.
I’m a big stickler in things feeling like that they belong to the platform. Otherwise, why not write a web app instead?
I think it depends on the target audience; usually, I’m writing a GUI for a piece of C++ code that controls a robot or a geometry system or something, and the intended users are either me or the employees of a company I’m consulting for. Writing a web frontend for some chunk of native code is irritating – I either have to jam a web server into a C++ binary, or create wrappers for some scripting language – and using a native toolkit isn’t that important to me, especially because the applications I’m writing usually have some weird widgets in them anyway (2D sliders, or line graphs, or something).
In general, I think that writing a very simple UI can really aide in debugging complex control systems, and I just pick up the tool that makes it as easy as possible to do so. In my experience, that’s an immediate GUI, without callbacks or a dedicated event loop or anything.