timjentzsch / bevy_mod_localization

A prototype to implement localization support in Bevy.

Hi there 👋

I am a software engineer working with TypeScript, Python and Rust.

Some of the bigger projects I worked on:

• bevyengine/bevy: A data-driven game engine in Rust. I am helping with issue triaging, plugin development and optimizing the CI pipelines.
• @GrafeasGroup: GrafeasGroup was a non-profit organization aiming to improve internet accessibility. I helped to develop and maintain our database, API and network of six bots.
• GameFeeder/GameFeeder: A notification bot for Game updates, available for Discord and Telegram. One of my first bigger programming projects, serving over 7000 combined users.
• stonefish_engine: An amateur chess engine implementing the UCI protocol. This was my first bigger project in Rust.

bevy_mod_localization's Issues

Update to Bevy 0.10

Update to the current Bevy version.

Revise architecture

We need a more modular architecture that allows us to support more than just text localization.
I'm still thinking about our best option here, but this might work:

• Locale resource for setting the locale
  • Enum, either fixed language or system language
• LocaleFallbackMap resource to define fallback languages in case some keys/files are not localized in all languages
• LocaleDefaultFallback resource to define the "default" language that should have all keys/files localized
• LocalizedAsset component to load select an asset file from a folder based on the locale
  • Handles asset loading when locale changes
  • Handles fallbacks based on map and default
  • Offers events and change detection when new assets have finished loading
• LocalizedText component to set and update text based on the locale
  • Uses Fluent localization files
  • Obtains correct file from LocalizedAsset
  • Updates text when locale changes and the new asset has finished loading
  • Allows setting arguments for interpolation
  • If arguments change, text is also updated
  • If possible: Fallback works per key, instead of only per file

Add ergonomic API for text with component/resource args

Project Fluent allows to insert values in the text as arguments. This is already possible, but the user will have to manually define a system to localize the Text component, ensure that it is run every time when the locale or argument changes.

Instead, it would be nice to just define which components/resources should be inserted and with which key, automatically handling the updating in the plugin.

Add localization of arbitrary assets

It is important to provide a feature to localize any type of assets. Example use cases include localizing text in a texture, in-game dialogue in a sound file, a cut scene, etc.

This should work similar to the current text localization. The user defines the path to a folder, the output format and the extension of the asset. This folder should then contain the localized files named after the language identifier and the given extension. The plugin should then select the correct asset based on the already used algorithm.

We should provide options to the user to configure how the assets are loaded. For most assets it will make sense to only load the first matching locale, not all possible fallbacks like we have right now.

The text localization should be adapted to use this new asset localization system under the hood.

Improve CI workflows

The current GitHub workflows were basically just copied over from Bevy or Leafwing.
Since that time, there have been a lot of improvements to these workflows, such as using Leafwing-Studios/cargo-cache and dtolnay/rust-toolchain.

The workflows should be updated to use these new tools.
https://github.com/Leafwing-Studios/Emergence and https://github.com/janhohenheim/foxtrot can serve as a basis.

Change name to `bevy_mod_localization`

I want this crate to be a complete solution for localization, not only for text (see #10 and #11).
The name should reflect that, so bevy_mod_localization sounds better.

It won't be a prototype anymore, but I don't want to steal bevy_localization for a potential official solution in the future. So the mod prefix seems to be the best option.

Add support for time-synchronized subtitles via WebVTT

For cut-scenes and in-game dialogue, it would be very useful to provide support for time-synchronized text. This can serve as captions or subtitles.

The most suitable format to provide that feature seems to be WebVTT. It is used in the web stack for videos, but also adopted in other areas like YouTube. It allows you to specify timed cues, as well as positioning and styling for them.

Since styling is still being developed in bevy_ui and the limited support for font families, simple support for plain-text cues is sufficient for now. Of course, the user should be able to define the subtitles in multiple languages, so this feature should use the localized assets from #10 as well.

Depend on specific bevy subpackages instead of bevy with feature flags

We currently depend on bevy directly, with a sparse set of feature flags to avoid pulling in too many crates.
However, it would probably be better to depend on the concrete sub-packages like bevy_asset and bevy_text.
I assume that it will make it easier for Cargo to de-duplicate the dependency tree.