Great ideas there Rob, thanks. I will pursue them.
--
bc
On Jun 8, 2016 4:25 AM, "Rob van der Heij" <[email protected]> wrote:

> On 7 June 2016, 12:27:26, Bob Cronin wrote:
>
> > So, I am looking for ideas on how to choose the first VALID address on
> that
> > list rather than just the first address. As usual the selection logic is
> in
> > the middle of a huge pipeline with all sorts of things going on, so it
> will
> > be important to not disturb the timing in any way.
>
> Assuming you have one record with the list of addresses to probe, your
> filter
> should simply output the first one in the list that meets the criteria,
> right?
>
> There are a few trigger words in that. The "first in the list" suggests a
> pipe
> with "juxtapose count" to collect all valid ones from the list, and ignore
> all
> but the first one. Disadvantage is that you evaluate all of them, even
> when the
> first was good enough. If that is a concern (like when you would look up
> the
> address through DNS) a feedback loop with "fanintwo" would be appropriate.
> Or
> a sipping pipeline.
>
>    \ *:
>    | o: fanout
>    | spec number 1.10 r
>    | j: juxtapose count
>    | unique 1.20 first
>    | substr 21-*
>    | *:
>    \ o:
>    | split ,
>    | .. sift bad ones out
>    | j:
>
> As for sifting out invalid addresses from the list, you'd probably use some
> rough selection stages to drop the bogus stuff. Add more over time or after
> sifting through some history data. Things like this are pretty effective:
>
>    | locate fs @ ws . substr w2 of f2        /*  @ plus 2 qualifiers  */
>
> Sir Rob the Plumber
>

Reply via email to