andrewmacmurray/elm-concurrent-task

Run a tree of Tasks concurrently, call JS functions as Tasks (Task Ports).

elm-concurrent-task's Issues

ConcurrentTask.Time.getZoneName references wrong js function

https://github.com/andrewMacmurray/elm-concurrent-task/blob/1.0.0/src/ConcurrentTask/Time.elm#L86C29-L86C29
uses builtin:timeZoneOffset instead of builtin:timeZoneName.
This seems unintentional?

Just something I've noticed looking through the code, nothing that's bitten me. Much love for making this package 💙

Copy API from `Time`

How do you feel about the idea of copying the API from elm/time inside ConcurrentTask.Time?

The idea is that then you can import ConcurrentTask.Time as Time and just use it as if it were the normal Time module.

Http Bytes compatibility

It would be nice to make the ConcurrentTask.Http module compatible with elm/bytes.

A popular way seems to be encoding / decoding a Base64 string with a package like this. This however loses most of performance gains of using Bytes.

I'd be curious if anyone has any opinions / alternatives to this.

implementation question about consuming JSON body

For expect JSON, why does the http body get decoded with .text(). instead of .json()
https://github.com/andrewMacmurray/elm-concurrent-task/blob/1.0.0/runner/http/fetch.ts#L30?

Just something I've noticed looking over the code. Is this intentional?

Multiple pools per ports

Currently it's assumed that you'd use 1 task pool per pair of ports.

However, this is not the case with something like a nested TEA single page app, where it makes sense to have a task pool per page. It's unsafe currently to do this:

• A task is in-flight from one pool (say internally it has an attempt id of 0).
• The user switches pages and immediately starts a task.
• The attempt ids would overlap and there would likely be a weird race condition.

A possible solution would be to add ConcurrentTask.namedPool where you could give a pool a unique name / id - this would prevent race conditions as the runner would know which task pool the responses are associated with.

There may be a way for the pool to initially generate a random id internally before it does any work, but would have to experiment with that and hopefully not break the types