refactor: async iterators

[alanshaw]

@jacobheun this is https://github.com/alanshaw/it-mplex + the example from this repo updated to use the new API.

Note this includes an adapter class so it can be used immediately in JS IPFS when merged and released. I've run the JS IPFS tests and interop tests against it and they pass!

TODO:

• Fix up the one failing test from refactor: API changes and switch to async await interface-stream-muxer#55 (either here or there 😂)

Depends on:

• refactor: API changes and switch to async await interface-stream-muxer#55

BREAKING CHANGE: This module now has a completely new API based on async iterators. See the README for documentation.

[jacobheun]

Just some small comments. I am running the adapter through the benchmarks at https://github.com/libp2p/pull-mplex/tree/master/benchmarks to see what the difference is. I'm hitting some issues with the adapter benchmarks, I'm looking into it. Right now the bench test is really slow and the large file test hangs.

[jacobheun]

@alanshaw the restrictSize fix took care of the issue with benchmarks. When I was running that the restrict size error ends up getting swallowed and never logged, only abort errors show in the logs. Can we have the source errors propagate up to the logs, they seem to be getting swallowed?

[alanshaw]

I'll look into it - this still needs work clearly. The slow pingpong benchmark is also worrying.

[jacobheun]

I used npm alias support to install the interface-stream-muxer@0.6 code along side the async test suite to test the adapter, as I was curious about the close tests. It's failing the close tests there as well (both close tests fail for the adapter), so I think there is something wrong with error/close propagation.

[alanshaw]

@jacobheun I'm working on a fork of the libp2p-tcp async/await lib that's used in that test. I built https://www.npmjs.com/package/stream-to-it to ensure all cases were covered when a writable stream closes or errors. When libp2p-tcp is ready the test should pass and we should be gtg. My plan is then to circle back to the interface-connection PR.

[daviddias]

Can you elaborate a bit more? The "probably" is not really comforting :)

[jacobheun]

Updated the comment for this. We should only hit the break in 2 instances.

1. We haven't gotten enough data to parse a header (most common).
2. The data we have can't be parsed, likely due to an invalid varint from bad data.

[jacobheun]

js-libp2p-mplex migration is complete

[jacobheun]

I've added some comments/jsdocs to the code, removed the adapter (we dont need it), and fixed an issue with message logging. This should be all set.

[vasco-santos]

I was trying out the examples and I am having issues with the dialer, not sure why

[jacobheun]

Fixed, the BufferList chunks need to be converted before going to node streams .write

[vasco-santos]

LGTM!