Re: Memory Allocation Question

[Rich Duzenbury <rduz-midrange@xxxxxxxxxxxxxxxxxxx>]

On Tue, 2004-11-09 at 14:51, Tony Carolla wrote:
> I am working on a program that, for good reason, runs loooooong.  I am
> looking through our account files to find matching accounts, where two
> of Name, DOB, SS# match.  We frequently deal with the same client in
> the same day, but create a new account for them (hospital
> billing--long story).  My task seemed easy at first, so I wrote it as
> follows:

Hmm, seems easier to just create three indexes:
DOB + Name, SS + Name, DOB + SS

You could then read each record in the account file, and then find the
potential matches with three sets of setll/reade loops.  

Now the I/O's get a lot fewer: 3.5M for the whole customer file, plus
lots of setlls with (mostly) non-matching reade's.  Additional reads
only when you find matches.   

I would probably write each account to a work file as it is 'seen' so
that it can't be listed more than once, so 3.5 million writes (or
perhaps updates to a 'seen' flag) as well.


--
Regards,
Rich

Current Conditions in Des Moines, IA
Broken Clouds
Temp 62.6F
Winds out of the Southwest at 16mph