* Apply spotless by default
- Upgrades to spotless 8.2.1
- Moves spotless setup from an init.gradle.kts to build-logic
- Narrows down the scope of `target` in spotless configuration to
be more precise to workaround https://github.com/diffplug/spotless/issues/2717
- Updates all references to init.gradle.kts
Ran gradle-profiler ./gradlew build --dry-run to validate performance
impact.
Before PR:
Mean 10,527.96 ms with 289.01 ms std dev
After PR:
Mean 11,251.78 ms with 530.29 ms std dev
Regression is there, but quite minor.
Test: ./gradlew spotlessCheck
* Address comments from AI overlords
* Fix usage of rootProject
* Enable spotless for build-logic via root project
The notification icons are now stored in `:core:notifications`.
This forces `:sync:work` to depend on it.
Another solution could be to provide the resource id through Hilt, but it would require more changes.
- Reduce the visibility of multiple Hilt `Module`s and implementations of public interfaces
- Correctly configure the visibility of dependencies:
- `api` when it's part of it's public `api`
- `implementation` when it's part of it's internal `implementation`
- Remove unnecessary dependencies in build.gradle.kts files
- Remove unnecessary dependencies provided by plugins
- Remove unnecessary applied plugins (i.e. roborazzi)
- Sort dependencies in `build.gradle.kts`
- Delete unused entries in `libs.versions.toml`
- Remove unnecessary nullable types
- Replace no-op method bodies with Unit
- Convert to expression body
- Replace if with when
- Remove braces from 'when' entries
- Remove braces from if statement
- Convert to single line lambda
- oneline if/returns
- Replace 'contains' call with 'in' operator
Following this refactor, it could be great to envision a more "strict" code formatter like ktlint 1.0 (we are currently stuck at 0.48.1)
The initial reason for this `StringDecoder` was to make sure the AndroidX navigation route would not use incompatible chars in the resulting String.
For that, `Uri.encode(String)` was used, but this is an Android API (`android.net.Uri`).
Unit tests would fail because Uri is not mocked (since we are not using Robolectric).
Therefore, a `StringDecoder` interface was introduced with 2 implementations: `UriDecoder` and `FakeStringDecoder`, Hilt modules were added, etc.
FWIW, the naming of the API was misleading: `StringDecoder.decodeString(encodedString: String)` does not inform how encoded the input is, so `UriDecoder` could "decode" something that was not necessarily uri "encoded".
The solution to this problem was to simply use regular Java APIs:
- `URLDecoder.decode(urlEncodedTopicId, "UTF-8")`
- `URLEncoder.encode(topicId, "UTF-8")`
This moves the responsibility for joining the UserData and the
NewsResources to UserNewsResourceRepository. This way, the work can be
done once and shared with all consumers in a SharedFlow, rather than
having each consumer perform the join itself by invoking the UseCase.
* Move Hilt setup to convention plugin
* Disable Hilt on modules where it's not needed
* Enable non-transitive R classes
* Fix `configuration` blocks called in wrong scope
* Move imperative Flavors code from build.gradle
* Disable unnecessary build features globally
* Disable configure-on-demand as it's not encouraged
* Don't use internal AGP packages in plugins
Change-Id: I1bee3e3fd0103054637b979a350f9fd2312ae8ba
Aurimas
Jaehwa Noh
Simon Marquis
Don Turner
Don Turner
Melwin Magalhaes
Simon Marquis
James Rose
Márton Braun
Manuel Vivo
Wojtek Kaliciński
Milosz Moczkowski
madroid