On Wed, Jul 31, 2013 at 05:01:35PM +0300, Javor Kliachev wrote:

Hello,

In the documentation it is stated, that pair sets defined like this  
[(1..100,123)] are in fact translated to
[(1,123),(2,123),(3,123),...,(100,123)] and hence there is an  
considerably increased memory footprint.

Is this also true if  the right member of the pair is defined as a range  
or as *:

[(123,*)]  =?=  [(123,0),(123,1),(123,2),...,(123,65535)]

Should I consider an eventual intensive memory usage in both cases?

No. Pair sets like [(123,10..100)], [(123,*)] or even [(10..100,*)] are
efficient. Think about it like the value on the right side sudivides the
value on the left side, therefore (123,10..100) is one interval from 
(123,10) to (123,100), while (10..100,123) is many small (single-pair)
intervals.