--- Paul Dorfman <paul_dorfman@HOTMAIL.COM> wrote: > However, the peekc/poke couple has proven to be practically useful to > make a > number of related programs run faster than they used to, thus making > it > possible to fit them in the production time window. Fancy a huge SAS > data > set with numeric variables variables V1-V3600 (actually representing > 36 > month with 100 data elements in each - the design stank, but I was > not at > liberty to change it) and close to a million rows. Each observation > is > supposed to be a sliding window, i.e. when 100 data elements Z(*) for > a new > month arrive, they are inserted from right to left as in a queue > displacing > all elements 100 places downwards, the lowest 100 elements (the > oldest > month) falling off and being discarded. It is very easy to do like > this: > > array v (3700) v1-v3600 z1-z100 ; > do j = lbound(v) + 100 to hbound(v) ; > v (j - 100) = v(j) ; > end ; > > However, it requires moving one element at a time. Now as an > exercise, I > offer everyone to think it over how the peekc/poke team can be used > to make > this process of sliding 3600 elements down the array *as a whole* > faster, > easier, and more naturally aligned with the problem's > stream-of-the-consciousness. >

Umm, write an IML program so that I can use an assignment like

v = z || v[,1:3500];

Oh, that wasn't the answer? Gosh, darn, there are just too many good choices in SAS!

Dale

===== --------------------------------------- Dale McLerran Fred Hutchinson Cancer Research Center mailto: dmclerra@fhcrc.org Ph: (206) 667-2926 Fax: (206) 667-5977 ---------------------------------------

__________________________________________________ Do you Yahoo!? Yahoo! Mail Plus – Powerful. Affordable. Sign up now. http://mailplus.yahoo.com