RE: a need for speed

rpg400-l-request@xxxxxxxxxxxx wrote:

>   5. RE: a need for speed (Joe Pluta)
>
>> From: Tom Liotta
>> 
>> AFAIK, Booth's method would "work", but Scott's _could_ easily be
>faster.
>
>Scott's won't work.  In the data given, he'd get a record of 0001/0092
>for terminal JSN, not two separate records as requested.

Ah, yes. It's not so much the range of suffixes as it is all of the various 
sub-ranges with a 'terminal' group. In some ways, it's determining which 
suffixes aren't used for a terminal. I see.

Every record indeed needs to be read and program logic needs to detect when 
continuity breaks occur on sub-ranges. So if I understand, the terminal could 
have suffixes 0001, 0002, 0003, 0005 and 0006. That results in two sub-ranges 
-- 0001-0003 and 0005-0006 -- within that terminal.

And 'terminal' causes a level break so that consecutive suffixes that cross 
groups still cause sub-range breaks.

There doesn't seem to be a good way out. Booth's method works for marking the 
'terminal' boundaries even though every individual record must still be... 
hmmm...

Well, actually, it shouldn't be necessary to READ many records if the index is 
built and includes suffix -- SETLL could be used just to see if an index entry 
exists. Then increment suffix in the search key and do another SETLL. Continue 
until no hit and then READ the next higher record to get the new search values. 
The key before the no-hit condition would mark the upper end of a sub-range. 
The record READ after the no-hit would be the lower end of the next sub-group 
and possibly mark a 'terminal' boundary also.

Might end up with _relatively_ very few records actually read.

Tom Liotta