David Neil comments

Results 25 comments of


                                            David Neil

add out-of-order variants of AsyncIterator.prototype methods

Aren't Streams the out-of-order Iterators? I wonder if all of the out-of-order concerns can just be handled by making a convenient `Iterator`->`ReadableStream` helper. And maybe `ReadableStream` would be well served...

add out-of-order variants of AsyncIterator.prototype methods

> ReadableStreams are ordered I guess you are correct, once the items are enqueued. I suppose I always just reached for `ReadableStream` when I wanted any-order iteration, because it allowed...

--experimental_output_paths=strip is not effective when actions are scheduled in parallel

We have some developers using local execution with remote cache. Personally, I am using remote execution with the dynamic spawn strategy, (and `--noexperimental_dynamic_exclude_tools`) And our CI system uses only remote...

Focusing only on signals interoperability

My primary concern is performance. It would be interesting to run these interop implementations through some benchmarks on to see if they are slowed down overly much.

Signals should be able to efficiently return to an old value

So my org's legacy implementation of Signals does that kind of thing, where it stores the lastValue on the edges of the dependency graph. It is a design decision we...

Signals should be able to efficiently return to an old value

I like using the internal cache because then it is behavior that you can specifically opt into in cases where it is actually helpful, and has almost no cost on...

Signals should be able to efficiently return to an old value

We have some Puppeteer tests that load our application, and then puts the application in a state where we expect there to be 100s of instances of class `C`, then...

Signals should be able to efficiently return to an old value

> and then I can opt-out when I know I have to There is a fixed overhead of another ~8 bytes per edge that I can't think of a reasonable...

Consumers should be held weakly to allow all nodes to be live

In my opinion `Signal.subtle.unwatched` and `Signal.subtle.watched` would be invoked when the number of consumers moves to/away from 0, regardless of the type of that consumer (`Watcher` or a read of...

Consumers should be held weakly to allow all nodes to be live

> Actually, WeakRef looks terrible in benchmarks. IIRC, Angular used it for both producers and consumers. I remember benchmarking that implementation with only using `WeakRef` for consumers and had almost...