Combine latest on some about streamz

[jrmlhermitte]

I think for simplicity, maybe combine_latest should just assume that you only emit on the first child, which means changing this line in combine_latest:

        if not self.missing and who == self.children[0]:

thus this will emit once:

s1 = Stream()
s2 = Stream()
sout = s1.combine_latest(s2)
sout.map(print)

s2.emit(1)
s1.emit(1)
s2.emit(1)
s1.emit(1)
s2.emit(1)
s1.emit(1)

which only prints on emitting from s1.

In the long run, I think a trigger should be added somewhere. Perhaps add a trigger flag to the buffer stream element?

What do you think? I think it's simpler than anything else I can think of. Also, we should fork streams itself (separate from event model). It could be on your or my branch, up to you.

[mrocklin]

There are definitely cases where people want to emit on updates from multiple children. I think that one would take in a list of booleans in a trigger= or emit_on= keyword for which children should activate an emit.

…

On Mon, Jun 12, 2017 at 10:52 AM, Julien Lhermitte ***@***.*** > wrote: I think for simplicity, maybe combine_latest should just assume that you only emit on the first child, which means changing this line in combine_latest: if not self.missing and who == self.children[0]: thus this will emit once: s1 = Stream() s2 = Stream() sout = s1.combine_latest(s2) sout.map(print) s2.emit(1) s1.emit(1) In the long run, I think a trigger should be added. Perhaps add a trigger flag to the buffer stream element? What do you think? I think it's simpler than anything else I can think of. Also, we should fork streams itself (separate from event model). It could be on your or my branch, up to you. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#22 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AASszNoW7QmX8K82PjNuecizgG-3bscfks5sDVCrgaJpZM4N124G> .

[CJ-Wright]

I like the idea of a kwarg for this.

I guess it would be like:

emiting_children = [child for child, t in zip(self.children, trigger) if t is True]
if not self.missing and who in allowed_emiting:

[jrmlhermitte]

Sorry yes, that was too restrictive. I like that too.
Maybe have the keyword be the list (or one element) of numerically ordered children as a compromise?
(so s.combine_latest(emit_on=0) would be fine)
I suggest a PR somewhere? Maybe @CJ-Wright self-PR from your feature branch to your master and we can work from there? (Unless @mrocklin you want to do it)

[CJ-Wright]

Closed by #32