On Friday, October 4, 2019 at 5:40:08 PM UTC-4, ohir wrote:
>
> On Fri, 4 Oct 2019 13:52:19 -0700 (PDT)
> T L <tapi...@gmail.com <javascript:>> wrote:
>
> > One priority-order select block is not only cleaner than two
> random-order
> > select blocks, but also more efficient.
>
> It is neither.
>
> 1. It is not cleaner, because you would need to somehow denote priorities.
> If you think now about "in written order", you should think of the most
> frequent
> cases that you do not want to have all five, ten, twenty cases be serviced
> on order. Not to mention that down the road someone might introduce an
> unnoticeable bug for pure aesthetic reasons.
>
> 2. It is not more efficient, it is less efficient because every select
> operation
> would need at least a "priority comparison" op, or - in general case - a
> sort
> op, because not only two channels can be ready, it may happen that all of
> them will. This would have to run for every select then, unless
> introduced by a "select_with_priorities" keyword.
>
>
select(ordered) {
case <-ctx.Done():
return ctx.Err()
case out <- v:
}
is undoubtedly more efficient than
select(randomized) {
case <-ctx.Done():
return ctx.Err()
default:
}
select(randomized) {
case <-ctx.Done():
return ctx.Err()
case out <- v:
}
Not only one less try-receive operation is saved,
but also is the step 2 described in
https://go101.org/article/channel.html#select-implementation saved.
> 3. Others tried to no avail, I will too in other words:
>
> The select switch is deliberately randomized because language creators
> envisioned that otherwise someone would be "smart" and will use
> ordering to impose a **dependency** where it should not be made.
>
> Channel operations serviced by select are treated as sourced
> from **independent** events. Period.
>
> If your out <-v must be dependent of ctx.Done not being ready you
> must resolve this dependency by other means. Eg. by selecting for it
> in separate select before.
>
It is not a must. It just tries to select ctx.Done if possible.
In fact, if the two evens are independent with each other,
there is no way to absolutely ensure the must.
>
> Hope this helps,
>
You can read the first comment in this thread for another use case of
priority-order
select block.
>
> --
> Wojciech S. Czarnecki
> << ^oo^ >> OHIR-RIPE
>
--
You received this message because you are subscribed to the Google Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to golang-nuts+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/golang-nuts/0c9aaa9b-ea9d-49f5-a1b4-4f94ee0de424%40googlegroups.com.
