Try a different approach to importing "cloudflare:workers". #82
Conversation
Fixes #23 Per #23, many bundlers do not like our dynamic import of "cloudflare:workers". This PR takes a different approach, using the "exports" feature of "package.json" to point workerd at a different version of the code. That version imports "cloudflare:workers" and sticks it in the global scope before importing the rest of the library, thus allowing it to conditionally probe.
|
Missed this, will see over the weekend |
|
This pull request fixes all the issues I was having. Any idea when it will be merged and a new package created? |
|
looking at it today |
|
@scotthellewell Sorry for the delay on this and everything else Cap'n Web! In the month since Cap'n Web was released a whole lot of stuff has conspired to take of my time (and Sunil's too), and this week I'm on vacation. Next week I plan to circle back and address all the open Cap'n Web issues. |
|
@kentonv I hope you have a great vacation. I've grabbed your branch and compiled my own package from it for now which works. I just like to directly use from the npm repository when possible. But great library so far. |
threepointone
left a comment
threepointone
left a comment
There was a problem hiding this comment.
approving this because it works and I want to unblock, but I think we can simplify it. I'll work on that myself later.
|
Awesome to see this is approved now. Do you have an estimate on when this will get merged and published to npm? I'd love to just reference your package, but I need this fix. If it will be a while still, I can publish a separate scoped npm package to use for the time being. But if it will be done in the next few days I would prefer to just wait. I would love an ETA. Thanks. |
|
Yes I'm planning to do a pass through all the open issues tomorrow and then probably push a release. Meant to do it yesterday but there was some other fire that came up. :/ |
* Add changeset for past PR #82. * Throw an error when attempting to access RpcTarget instance properties. This helps people learn why instance properties are not accessible over RPC, whereas returning `undefined` leaves them confused. Fixes #55 * Implement toString() for RpcStub and RpcPromise. They just return `[object RpcStub]` and `[object RpcPromise]`, but that's better than the previous `[object Function]` which was confusing. As suggested in #55. * Fix type signature of newHttpBatchRpcSession(). An earlier version of the function was designed to have the same signature as `fetch()`, but when I added `RpcSessionOptions` it made more sense to make that the second param... but I only updated the implementation and forgot to update the type. Fixes #67. * Support serializing Infinity, -Infinity, and NaN. Fixes #80 * Polyfill Promise.withResolvers(). This hopefully improves compatibility with old Safari versions and Hermes (React Native). Unfortunately it didn't seem easy to extend the tests to cover React Native. There is a vitest-react-native package, but it looks little-used and unmaintained, so I didn't want to install it. Might help with #91, though I won't declare that one "fixed" until we actually have tests proving it. * Document missing expression types in protocol.md. Fixes #48
3 participants
Fixes #23
Per #23, many bundlers do not like our dynamic import of "cloudflare:workers". This PR takes a different approach, using the "exports" feature of "package.json" to point workerd at a different version of the code. That version imports "cloudflare:workers" and sticks it in the global scope before importing the rest of the library, thus allowing it to conditionally probe.
@threepointone is this a sane approach? I have very little understanding of JS tooling but this seem like it works.