You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
ATM, it's not possible - without subclassing Subscriber - for user-land operators to correctly chain the closed property of the subscriber received by the Observable ctor's callback.
We should export something like the internal OperatorSubscriber, to facilitate the more straightforward creation of user-land operators. IMO, we should export something that accepts callbacks using a config object - rather than the positional arguments.
Also, as raised in #2314, we should use a narrower type - e.g. a SubscriberLike - that does not expose as many implementation details as Subscriber.
The text was updated successfully, but these errors were encountered:
Some months ago core team meeting had discussed similar to this - make Observable itself as strongly public interfaced one which can create operators, and all of rxjs operator itself try to utilize it to create operators instead of rely on some internals. Discussion concluded we'd like to setup monorepo-ish (yes, again) to make rxjs/Observable itself as explicitly seperated import site operators can import.
In my opinion, regardless of publishing rxjs/observable or not - it may worth to explore these paths in a long run. I wanted to try this after we get concluded how to organize our repo for separated packages. (then we can make observable as internal packages to all operator enforced to follow its public interfaces)
Not directly related, but I guess this is somewhat related topics.
ATM, it's not possible - without subclassing
Subscriber
- for user-land operators to correctly chain theclosed
property of the subscriber received by theObservable
ctor's callback.We should export something like the internal
OperatorSubscriber
, to facilitate the more straightforward creation of user-land operators. IMO, we should export something that accepts callbacks using a config object - rather than the positional arguments.Also, as raised in #2314, we should use a narrower type - e.g. a
SubscriberLike
- that does not expose as many implementation details asSubscriber
.The text was updated successfully, but these errors were encountered: