withastro-utils
Package is incompatible with Astro's Cloudflare adapter
Hi,
This is not necessarily a "bug", but that's the only issue template available.
I tried combining this package with https://docs.astro.build/en/guides/integrations-guide/cloudflare/ but with no success. It, and its dependencies don't follow the requirement to
have all node internals' specifiers prefixed with node:.
Here's a list of those that might be prefixed and may get this closer to working in workers environment:
-
assert -
path -
crypto -
util -
buffer -
stream
But I also see those:
-
fs -
fs/promises -
os
and those don't have support in Cloudflare (see https://developers.cloudflare.com/workers/runtime-apis/nodejs#nodejs-api-polyfills)
There's also a dependency on https://www.npmjs.com/package/snappy and it's a hard no-go for Workers runtime, because it's written in Rust and is published as NAPI node extension.
Question: is making this package compatible with workers (or extracting its subset which might at some point become) is considered "in scope"?
Hi @jrencz, sorry for the delay (did not get GitHub notification, need to figure this out). For your question, I am afraid not, this package does a lot of the hard lifting and depends on the context.
The fs/fs-promises is required for file uploads and snappy is required for the context for ASP.NET like experience.
It is possible to create a subset of this package without file upload and replace snappy with something else, but for now it is the best I found (I term of speed and compression)
Could you explain more what is your use case?
Hi, thanks for the answer.
I'm primarily interested in JWT session and CSRF protection - those things seem like they could live without a link to filesystem.
I try to deploy it to Cloudflare Pages
Maybe a way to go would be to extract a smaller package with just workers runtime compatible stuff?
https://developers.cloudflare.com/workers/runtime-apis/
I'll investigate that. As you mentioned, only "bigFileUpload" requires the fs and os modules.
However, incorporating a high-speed, highly-efficient compression library (such as snappy) might be more challenging