You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The general idea here is to let folks define a js module to use as a
data source for a "sender."
This could be general purpose but the initial use case is for polling
APIs that don't expose webhooks.
Currently this diff contains only support code for config handling.
Some implementation details were worked out separatly as a
[POC in a separate repo.](https://github.com/svix-onelson/poller-input-poc/)
By using `fetch` to issue HTTP requests, we can track state in the
module itself to decide if the data is worth sending a webhook for.
Example:
<https://github.com/svix-onelson/poller-input-poc/blob/main/fetcher.js>
There are many barriers to integrating the code in the POC linked above
which need to be cleared first.
We need:
- a custom module loader to let us source modules from the
bridge config instead of js files on disk.
- Glue code to connect the `op_forward` deno extension calls to either a
transformation or svix sender output.
- existing transformation code in bridge needs to be refactored to allow
us to use "newer deno" without also introducing a memory leak.
For the latter, <svix/monorepo-private#5670>
aims to solve this.
In addition to the above, deno ops (i.e native extension code) are
supposed to be able to register state with the runtime, but I wasn't
able to get it to work for keeping track of which worker was which
(allowing us to propagate payloads to the appropriate output).
0 commit comments