search

storacha-network / w3ui

šŸ‘©ā€šŸŽ¤ Headless, type-safe, UI components for the next generation Web3.Storage APIs.
Other
63 stars 25 forks source link

feat!: fix example, buncheefication, cleanup #603

Closed travis closed 7 months ago

travis commented 7 months ago

I was hoping to keep these changes separate, but ran into some issues with using the new packages in examples that necessitated some build changes.

1) I've updated the file-upload example to use the new packages. I ran into a weird build issue trying to do this - something about Authenticator not being exported from the UMD builds? - and did some research into how other projects I like are built, which lead me to believe that bunchee, the tool SWR uses for builds (https://github.com/vercel/swr/blob/main/package.json#L94), is probably our best bet...

2) I love bunchee so far - it's much simpler and seems to do "the right things" when building TS into libraries suitable for use in a variety of modern JS environments. You can read more about bunchee here:

https://github.com/huozhi/bunchee

I've used bunchee to pull the build step into each of the packages separately, which should make it easier and more efficient to work on a single package in isolation. Because bunchee doesn't need a bunch of manual config, this was possible with a minimum of boilerplate and setup.

3) I've removed the old packages and a bunch of configs and docs that point to them. This made (2) easier, since I could just make a wholesale switch rather than trying to keep some stuff working with the old rollup builds and other stuff working with bunchee.

gobengo commented 7 months ago

We'll probably want some kind of followup task to this to unpublish the npm packages we no longer plan to maintain? Otherwise probably people will stumble upon them via search results, try to use them, and be frustrated that they are no longer supported.

gobengo commented 7 months ago

@travis should this probably be feat! in the title? It looks like it's not strictly a breaking change for core but it could be nice to start a new major version here just so we can clearly distinguish the really big change.