fix(window.ipfs): stop injecting window.Buffer #732
Conversation
| @@ -46,7 +45,4 @@ function createWindowIpfs () { | |||
| return freeze(proxyClient) | |||
| } | |||
|
|
|||
| // TODO: we should remove Buffer and add support for Uint8Array/ArrayBuffer natively | |||
| // See: https://github.com/ipfs/interface-ipfs-core/issues/404 | |||
| window.Buffer = window.Buffer || _Buffer | |||
There was a problem hiding this comment.
I think we should add support for TypedArray and ArrayBuffer before we ship this.
I'd be interested to know if this covers all the bases.
There was a problem hiding this comment.
Yeah, I've been thinking about it and (C) feels like a right way to do it.
Something I realized is that the first thing people do when they play with API is to add some text (String) to IPFS. What if we remove friction of converting to bytes and accept String input and convert it internally to a typed array?
afaik something like https://encoding.spec.whatwg.org would do:
const enc = new TextEncoder() // implicit utf-8
const bytes = enc.encode('This string will be converted to a Uint8Array'))There was a problem hiding this comment.
Yes, I've wanted this for a while and allowed it in https://github.com/ipfs-shipyard/ipfsx/blob/master/API.md#example-1
There was a problem hiding this comment.
The tricky bit with it is destinguishing between content and a path in the file system, which I think is supported somewhere
There was a problem hiding this comment.
yeah thats the problem addFromFs
There was a problem hiding this comment.
Is addFromFs the only place?
I think it is quite intuitive special case where String is interpreted as file path. It could run the generic normalizeInput like every other command and then convert bytes back to file path internally.
There was a problem hiding this comment.
Previously we had everything in the add, addReadable, addPullstream when we introduced addFromFs i think we kinda separated the /file/path input to this method only need further investigation
|
Superseded by #777 |
This PR removes code responsible for injection of
window.Bufferwhenwindow.ipfsis enabled.We did that as a convenience tweak to make it easier to write demos/pocs, but polluting global scope introduced some issues on websites (Closes #637).
Question: should we
window.ipfsaltogether in September:https://github.com/ipfs-shipyard/ipfs-companion/blob/4ca9a0ce1842a9cf5979623bbf416493a18c2b2f/add-on/src/contentScripts/ipfs-proxy/page.js#L37