This repository has been archived by the owner on Oct 19, 2022. It is now read-only.
We are renaming the prefix of our JavaScript modules from 'node-' to 'js-' #1
Comments
jacobheun
added a commit
that referenced
this issue
Sep 20, 2019
* feat: dream API for multistream-select
Here's some dream code! This is how the API could look in an async/await future. I would love to know what you think and if you have questions.
I **think** I have all my assumptions correct about how the current implementation works and is used 🤣!
* feat: select method
As per the dream code this function allows selecting a protocol from the passed list of protocols using the provided muxed stream.
We just iterate over the passed protocols and...
1. Write to the muxed stream sink with (length prefixed) protocol
2. Read response from muxed stream source
* We use `it-reader` and `it-length-prefixed` and read single bytes until we have the whole response
* I understand now why `pull-length-prefixed` has a `decodeFromReader` function and I'll add that to `it-length-prefixed` and use it here
* TLDR; We only need to read single bytes until we have a valid varint!
3. Check if response is expected protocol (not "na")
4. If expected then return stream and protocol, if not then iterate
The interesting bit is where we construct the `selectedStream` (the stream we return to the caller). We need to use the sink from the muxed stream to write our multistream protocol(s) and _then_ we need to allow it to be used by whoever needs it. This is just a concat of two sources, but the tricky bit is that the second is not known until after we return the selected stream to the caller. Line 20-24 pass a source to the muxed stream sink that iterates over our writer stream and then waits for a source to be provided to the `selectedStream` sink. The `selectedStream` sink function just resolves this promise and the content from this source is then yielded from the source we passed to the muxed stream's sink.
The source for the `selectedStream` is just the `reader`. You can think of it as just a wrapper around the muxed stream's source. It is used to read responses during multistream negotiation but can then continue to be used by whoever needs to read the rest of the stream. This is because it's just an iterable who's next function optionally takes a number of bytes to read. See the [example](https://github.com/alanshaw/it-reader#usage) for more info.
License: MIT
Signed-off-by: Alan Shaw <alan.shaw@protocol.ai>
* feat: handle and ls
This PR supersedes #1 and #2 and adds `ls` and `handle` on top of `select` and adds 100% test coverage:
<img width="1281" alt="Screenshot 2019-08-14 at 12 21 52" src="https://user-images.githubusercontent.com/152863/63018364-89eba280-be90-11e9-833e-db5097b092c0.png">
API has changed a little from #2 because we actually do need to keep some state around for `ls`.
If you use `ls` you need to remember you've done the multistream-select handshake so that you can call `select` afterwards.
So the API is now:
```js
const dialer = new MSS.Dialer(muxedStream)
const protocols = await dialer.ls() // optional!
const { stream, protocol } = await dialer.select('/echo/1.0.0')
// --- and ---
const listener = new MSS.Listener(muxedStream)
const { stream, protocol } = await listener.handle('/echo/1.0.0')
```
This work also uncovered a possible bug in `go-multistream` multiformats/go-multistream#41
I also addressed feedback from #2 here.
If this is approved I'll open a PR to `js-multistream-select`.
closes #1
closes #2
cc @jacobheun @vasco-santos @daviddias
License: MIT
Signed-off-by: Alan Shaw <alan.shaw@protocol.ai>
* fix: fixes and tests
License: MIT
Signed-off-by: Alan Shaw <alan.shaw@protocol.ai>
* docs: remove ramblings from README
License: MIT
Signed-off-by: Alan Shaw <alan.shaw@protocol.ai>
* docs: fix comment
Co-Authored-By: Jacob Heun <jacobheun@gmail.com>
* fix: listener and dialer selection results
Co-Authored-By: Jacob Heun <jacobheun@gmail.com>
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
ipfs/js-ipfs#32
The text was updated successfully, but these errors were encountered: