feat!: fix liquid_glass layering and add opaque background option #199

ahonn · 2025-11-22T05:45:41Z

Fixes #198 and enhances the apply_liquid_glass API with improved layer management and customization options.

Changes

1. Fix WebView layering and corner radius issues

Resolves edge misalignment by reparenting the WebView into the glass view's contentView instead of placing it as a sibling. This ensures proper layer hierarchy and correct corner radius rendering.

2. Add opaque background option

The Glass Effect View's high transparency can be problematic. Adding a background directly in the WebView blocks the liquid glass effect. This introduces an optional opaque parameter that creates a separate NSBox background layer positioned below the glass view, preserving the glass effect while providing background control.

3. Refactor API with `LiquidGlassOptions` struct

Breaking Change: Consolidate parameters into a structured options object for better ergonomics.

Before:

apply_liquid_glass(&window, style, tint_color, radius)

After:

apply_liquid_glass(&window, LiquidGlassOptions {
    style: NSGlassEffectViewStyle::Clear,
    radius: Some(26.0),
    opaque: Some(false),
    ..Default::default()
})

…ssue

FabianLars · 2025-11-22T09:40:45Z

Resolves edge misalignment by reparenting the WebView into the glass view's contentView instead of placing it as a sibling. This ensures proper layer hierarchy and correct corner radius rendering.

The reason why i didn't copy your changes into this repo yet is because i don't think we should have webview specific code here. If possible i'd rather expose an option for the contentView and let users specify the webview themselves.
Most users won't even see this since we do that for them in Tauri.

This introduces an optional opaque parameter that creates a separate NSBox background layer positioned below the glass view, preserving the glass effect while providing background control.

I'll try to check it out later but what glass effect is left when you have an opaque background and what's the difference from a opaque webview background (eg either wkwebview's "native" background or just the html/css color)

Refactor API with LiquidGlassOptions struct

I don't think this fits the style of this crate tbh.

ahonn · 2025-11-22T12:36:50Z

The reason why i didn't copy your changes into this repo yet is because i don't think we should have webview specific code here. If possible i'd rather expose an option for the contentView and let users specify the webview themselves.
Most users won't even see this since we do that for them in Tauri.

I agree that having WebView-specific logic inside this crate isn’t ideal. My implementation was mainly a pragmatic fix for my own app, and I don’t think it’s necessarily the right abstraction. Exposing a contentView option and letting users attach their own WebView sounds like the cleaner long-term approach, and I’m happy to adjust the PR in that direction.

I'll try to check it out later but what glass effect is left when you have an opaque background and what's the difference from a opaque webview background (eg either wkwebview's "native" background or just the html/css color)

Based on my tests, an opaque WebView background blocks the glass effect completely. But an opaque layer under the glass still preserves the system tinting. You can verify this easily by moving the window over a strong-colored background, the glass keeps its subtle tint, while an opaque WebView background removes it.

I don't think this fits the style of this crate tbh.

That makes sense. I agree it doesn’t fully match the crate’s current style. I only grouped the parameters because there were getting to be a few, but I’m completely fine reverting to the original API approach.

FabianLars · 2025-11-22T12:46:51Z

Exposing a contentView option and letting users attach their own WebView sounds like the cleaner long-term approach, and I’m happy to adjust the PR in that direction.

awesome, that would be appreciated :)

Based on my tests, an opaque WebView background blocks the glass effect completely. But an opaque layer under the glass still preserves the system tinting. You can verify this easily by moving the window over a strong-colored background, the glass keeps its subtle tint, while an opaque WebView background removes it.

got it, sounds reasonable then.

That makes sense. I agree it doesn’t fully match the crate’s current style. I only grouped the parameters because there were getting to be a few, but I’m completely fine reverting to the original API approach.

If we keep the opaque option and add a contentView option then it's indeed getting a bit much and i'm okay with keeping the options in that case though i'd like to hear @amrbashir opinion on that if he has one.

amrbashir · 2025-11-23T02:09:03Z

I don't mind changing the API, we can even extend the same pattern to other platforms as well. My only concern is proofing against future breaking changes, so I would suggest to not make the structs fields public and provide a builder-pattern struct (or just methods on same struct).

ahonn · 2025-11-25T23:42:26Z

Thanks for the suggestions!

I've exposed a generic content_view() option instead. This keeps the crate framework-agnostic while still solving the layering issue. For the API style, I've converted options to use the builder pattern as @amrbashir suggested. The WebView detection logic has now been moved to the tauri example as a reference for users who need it.

…der pattern

FabianLars · 2025-11-26T10:48:25Z

src/macos/liquid_glass.rs

+}
+
+unsafe fn apply_corner_radius_layer(view: &NSView, radius: f64) {
+    let _: () = msg_send![view, setWantsLayer: true];


just one example but most msg_send uses can be replaced with an explicit objc2 function, for example here it's

view.setWantsLayer(true);

ref https://docs.rs/objc2-app-kit/latest/objc2_app_kit/struct.NSView.html#method.setWantsLayer

that should make the code a bit nicer to look at x)

Thanks! I’ve updated the code to use the explicit objc2 APIs instead of msg_send! Here and in the other spots as well.

src/macos/liquid_glass.rs

FabianLars

didn't look through everything yet cause my battery is almost empty x)

src/macos/liquid_glass.rs

FabianLars · 2025-11-27T10:05:43Z

src/macos/liquid_glass.rs

 }
+
+unsafe fn apply_corner_radius_layer(view: &NSView, radius: f64) {
+    view.setWantsLayer(true);


we have to monitor this very closely. https://developer.apple.com/documentation/appkit/nsview/wantslayer sounds like it could cause conflicts with idk what we're doing elsewhere in tauri land.

I updated it only to call setWantsLayer(true) when the view doesn’t already have a layer, and then modify the existing layer. Not entirely sure if this is the safest approach, so let me know if we should handle it differently.

yeah idk either, this wasn't really a call for action for you. just in case something acts weird (eg with tauri's unstable feature flag) we know to have this in mind 🤷

FabianLars · 2025-11-27T10:06:48Z

src/macos/liquid_glass.rs

+        self
+    }
+
+    pub fn content_view(mut self, view: NonNull<c_void>) -> Self {


Let's just take an NSView directly here

I've replaced all NonNull<c_void> with &NSView.
Maybe we need to update apply_vibrancy and clear_vibrancy, or their parameter types will be inconsistent.

FabianLars · 2025-11-27T10:09:45Z

src/macos/liquid_glass.rs

+}
+
 pub unsafe fn apply_liquid_glass(
    ns_view: NonNull<c_void>,


i know this is my bad but i think we can take an NSView here as well - the NonNull thing is from before we had the objc2 crates

Yes, I've considered this as well. I will change this to NSView.

FabianLars · 2025-11-27T10:10:07Z

src/macos/liquid_glass.rs

    Ok(())
 }

 pub unsafe fn clear_liquid_glass(ns_view: NonNull<c_void>) -> Result<bool, Error> {


socket-security · 2025-11-27T10:48:39Z

Review the following changes in direct dependencies. Learn more about Socket for GitHub.

Diff	Package	Supply Chain Security	Vulnerability	Quality	Maintenance	License
	cargo/serde@1.0.228
	cargo/serde_json@1.0.145
	cargo/tauri@2.9.3
	cargo/tauri-build@2.5.2

View full report

ahonn added 3 commits November 16, 2025 11:54

fix: move webview into glass view contentView to fix layered radius i…

3d96494

…ssue

feat!: add opaque parameter and refactor liquid_glass api

635f6ea

docs: update README with liquid_glass documentation

c621aa0

ahonn force-pushed the dev branch from 0557c4e to acf067f Compare November 25, 2025 23:48

feat!: add content_view option and convert LiquidGlassOptions to buil…

e997383

…der pattern

ahonn force-pushed the dev branch from acf067f to e997383 Compare November 26, 2025 10:48

FabianLars reviewed Nov 26, 2025

View reviewed changes

src/macos/liquid_glass.rs Outdated Show resolved Hide resolved

ahonn added 2 commits November 26, 2025 22:43

refactor: use explicit objc2 methods and objc2::ffi functions

029510f

refactor(examples): replace msg_send with objc2 type-safe interfaces

73b75ed

FabianLars reviewed Nov 26, 2025

View reviewed changes

src/macos/liquid_glass.rs Outdated Show resolved Hide resolved

ahonn added 2 commits November 27, 2025 02:04

refactor: use objc2-quartz-core and type-safe layer APIs

f1e8cc9

refactor: improve type safety and reduce unsafe code

fe0db07

ahonn force-pushed the dev branch from 960cc38 to fe0db07 Compare November 27, 2025 10:09

FabianLars reviewed Nov 27, 2025

View reviewed changes

ahonn added 2 commits November 27, 2025 21:46

refactor!: replace NonNull<c_void> with &NSView references

4168566

refactor: add defensive checks before modifying view layer properties

d211aa0

ahonn requested a review from FabianLars November 28, 2025 11:20

feat!: fix liquid_glass layering and add opaque background option #199

Are you sure you want to change the base?

feat!: fix liquid_glass layering and add opaque background option #199

Uh oh!

Conversation

ahonn commented Nov 22, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

1. Fix WebView layering and corner radius issues

2. Add opaque background option

3. Refactor API with LiquidGlassOptions struct

Uh oh!

FabianLars commented Nov 22, 2025

Uh oh!

ahonn commented Nov 22, 2025

Uh oh!

FabianLars commented Nov 22, 2025

Uh oh!

amrbashir commented Nov 23, 2025

Uh oh!

ahonn commented Nov 25, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

FabianLars left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

socket-security bot commented Nov 27, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

ahonn commented Nov 22, 2025 •

edited

Loading

3. Refactor API with `LiquidGlassOptions` struct