This project is currently only working for iOS/iPadOS, given it exclusively uses SwiftUI there should be no technical limitation caused by expanding support to macOS, and potentially even tvOS.

Closes #3

• Improved Catalyst Support
• Fixed tvOS Support
• Added macOS Support
• Added watchOS Support

TODO

• [x] Break SpanGrid down into a few entities
  • Move the three rowHeightLookup items into own extension
  • Move maths calculations into own file
• [ ] Remove requirement for horizontalSizeClass
• [ ] Improve highlight/focus state across supported platforms
  • Use arrow keys on supported platforms
  • Ensure consistent API across platforms
  • Ensure strong compatibility with accessibility tools (try navigate with games console, Switch Control, etc)
  • Add keyboard navigation to macOS (currently disabled, unsure why)
• [x] Move SpanGridWidthListener into own group
  • See if we can reduce duplication (publisher/notification)
• [ ] Add test suite covering all supported platforms
  • Use XCTSkipUnless to detect platform early on.
• [ ] Add missing app icons?

Separate Ticket

• [x] Add CI to run tests on all supported platforms, using GitHub Actions
• [ ] Improve sample app to showcase a few different options
  • A navigation tab will demonstrate a complex layout that is easily navigate-able using non-touch inputs
  • Sidebar on macOS, tab view on iOS, tvOS, and Catalyst, unsure on watchOS.
  • Could also try doing a SwiftUI widget too? Same reusable content.
• [ ] Improve Documentation
  • DocC tutorial?
  • Inline documentation
• [ ] Refine SpanGridDynamicColumnSizeStrategy further 👀
• [x] Add SwiftLog and remove all the print statements we have

Imagine a grid with 3 columns with the layout below:

[ X X O ] (2 items followed by whitespace) [ XXXX ] (single item filling entire row)

We are currently assigning the rowOffset of row 0 (first row) to the item with the full width tile. This is because we calculate the maths based on the first item in the technical grid, even if it's not rendered that way. We need to update the maths to take into account prefix cells.

Started testing this with a local project and it works fairly well. The biggest issue currently is that it really struggles around whitespace so we need to handle cell prefix in the maths calculations.

There also appears to be a fairly big performance penalty when dealing with large data sets in monkey mode. The numbers don't scare me since this is a manual user input which itself has time to register, but it would be nice to investigate some basic optimisations around that.

Finally, I'm trying to wrap my head around how keyboard navigation would work for cells with nested content. I'm thinking about the App Store layout where a cell can itself be another horizontal scrolling grid (a carousel).

How should keyboard navigation work with this? How can we add support for it? Could we make our solution more generic so it could be accessed by the public interface? We need to setup an example demonstrating this.

Arrow key navigation was not working on simulator, unsure why. Need to try on a physical device to see if there's a SwiftUI limitation there. Also, if we can't disable discoverability then we need to allow developers to pass through localised strings.

Video demonstrating behaviour: https://user-images.githubusercontent.com/15193942/135714274-6af1a55b-14e2-4ce3-b386-2f76302360d4.mov

TODO

• [x] 🚨 ~~Should use arrow keys instead of WASD~~ Not possible due to Apple limitation
• [x] Should disable discoverability, or allow user to localise
• [x] Fix issue with navigation around whitespace (cell prefix)
• [x] Fix performance with constantly looping over full cell cache
• [x] ⚠️ ~~Allow user to have cells opt out of keyboard navigation (would skip, similar to whitespace)~~ Out of scope for MVP
• [x] Hide the keyboard input buttons
• [x] Update documentation

Some cells will not immediately have data available to them, for example adverts, and will require a mechanism that allows to asynchronously load data. This is currently possible with SpanGrid.

What would be nice though is to add a demonstration of how you can dynamically insert or remove items into the grid efficiently based on network requests.

It may also be worth exploring the ability for SpanGrid to manage some level of async cell where, on failure, it would remove the tile.

More to think about here.

We should add the ability to navigate the grid using your attached keyboard, useful for both iPadOS and macOS (support pending).

It should support the four cardinal directions, respect whitespace and span size. It should be completely opt-in.

Implementation started in #1.