On Tue, 13 Sep 2011 09:58:12 +0200, DUGALEIX Michaël wrote:
>Well, its key is not currently in the master (which has only one key at 
>this moment, shows PIPEDEMO), but how does "lookup" know that the key 
>WILL NOT be in the master later ?

LOOKUP doesn't do time travel.  It doesn't care what will be there later.

>PIPEDEMO shows me that when "lookup" gets rid of "KEY1_B_3", all the 
>keys have not yet arrived in the master.
>They seem to get to the secondary input of "lookup" only just after "not 
>fanout" had read them. Very slowly.

Not secondary:  tertiary.  (That is:  input 2, not input 1.)  Notice the
pipeline in Rob's example that has just the \ l: \ label and nothing
else.  That's stream 1 of LOOKUP, connected to nothing.

Records on input 1 are all added to the reference before anything else is
read.  Records on input 2 are added to the reference as they arrive
later.

¬R      Blood is worthless, outside its original container.
http://users.bestweb.net/~notr/davidcar.html     --Don Rauf

Reply via email to