Create a dedicated section for interfacing with other specs #1073
Conversation
This incorporates much of https://fetch.spec.whatwg.org/commit-snapshots/ad9a84ad91fb6bd3a2dcdf0406321785f51c4c11/#streams, as well as https://wicg.github.io/compression/#generic-transform-stream. It also looks at how https://wicg.github.io/native-file-system/ uses writable streams, and creates wrapper algorithms to make that easier. Closes #372.
domenic
added
the
do not merge yet
|
This should also have a section on duplex streams like https://reillyeon.github.io/serial/#serialport-interface. I don't think they need any new algorithms, but stating things like "use the names |
|
Note that w3c/media-source#14 may eventually also be resolved such that Media Source Extensions becomes another interface point for the Streams API. |
|
I've had an initial read-through and it looks good. |
|
We will also need to integrate with https://wicg.github.io/web-transport/ but that should be straightforward. |
Mostly adding a lot of "-ing" linking text variants
Follows whatwg/streams#1073. Also includes minor editorial touchups to the algorithms as I was editing them.
|
OK, PRs sent to all of the above; I updated the OP with them. I got all ready to send one for WebTransport, but they don't appear to be currently at a level of rigor where this makes sense; they just talk about "Set X to a ReadableStream of Uint8Arrays representing (some abstract concept), currently empty". I guess the assumption is that all the enqueuing/closing/erroring is encapsulated in that "represents", hmm. |
No problem, we have to increase the level of rigor anyway, so we can adopt the new interfaces at the same time. Sorry for wasting your time. |
| 1. If |chunk| is not a {{Uint8Array}} object, [=reject=] |promise| with a {{TypeError}} and | ||
| abort these steps. | ||
| 1. Append the bytes represented by |chunk| to |bytes|. | ||
| 1. [=Read-loop=] given |reader|, |bytes|, and |promise|. |
There was a problem hiding this comment.
This can call ReadableStreamDefaultReaderRead re-entrantly, which is a bit scary. As far as I can tell it's safe in the standard language, but implementations can fail due to stack overflow. Easiest fix would probably be to post a microtask.
There was a problem hiding this comment.
Hmm. Given that there's already a note explaining that this might be implemented rather differently, I think it'd be best to make a note of the danger, and encourage implementations to mitigate it in a way that doesn't require "read all bytes" to delay for (number of chunks) microtasks.
There was a problem hiding this comment.
Well, I guess such a requirement wouldn't be observable either, so scratch that part of my comment. But still.
domenic
removed
the
do not merge yet
Follows whatwg/streams#1073. Co-authored-by: Andreu Botella <abb@randomunok.com> Co-authored-by: Anne van Kesteren <annevk@annevk.nl>
* Use new Streams algorithms Follows whatwg/streams#1073. * Namespacing tweaks * No subsubsteps Co-authored-by: Jake Archibald <jaffathecake@gmail.com> * Missing :: Co-authored-by: Jake Archibald <jaffathecake@gmail.com>
Follows whatwg/streams#1073. Also includes minor editorial touchups to the algorithms as I was editing them. Also improves the Makefile.
This incorporates much of https://fetch.spec.whatwg.org/#streams, as well as https://wicg.github.io/compression/#generic-transform-stream. It also looks at how https://wicg.github.io/native-file-system/ uses writable streams, and creates wrapper algorithms to make that easier.
Closes #372.
This is probably ready for review, but we shouldn't merge it until I have created update pull requests for at least:
and have validated that this suffices for their purposes.
Also, I'll be opening another issue about GenericTransformStream.
Preview | Diff