HASH.HELP has improved over the years, but I sure wouldn't accept its
results blindly! The most interesting part of your post, to me, is the
variation in type recommendations -- type 10 v.s. 18.
As you suggest, the modulo should be made a prime whether it's
calculated by HASH.HELP or some other source.
For the type, the basic problem with HASH.HELP is that it uses only an
analysis of the record key patterns. This is inferior to a simulation
approach. Of the available Universe tools I'd choose HASH.AID, testing
all the types 2 through 18, and then looking at the resultant data (LIST
HASH.AID.FILE) for smallest overflow and tightest hashing. HASH.AID
isn't perfect as it doesn't report empty groups but at least it is a
simulation rather than just eyeballing the record keys.
Of course, the main thing is to keep those files sized! An implemented
non-optimum resize is better than a theoretical perfect solution.
Jeff Fitzgerald
Fitzgerald Long, Inc.
www.fitzlong.com
-Original Message-
From: Jerry Banker [mailto:[EMAIL PROTECTED]
Sent: Wednesday, June 20, 2007 8:58 AM
To: u2-users@listserver.u2ug.org
Subject:
I just came across what may be a bug in HASH.HELP. I ran HASH.HELP on a
static hash file and it told me the smallest modulus should be 80015
(yes, not a prime number), smallest separation or 4, and the best type
of 10.
I then ran HASH.HELP.DETAIL on the same file and got 80021, 4, and 18
respectively.
In this second case the modulus is a prime number. No changes were being
made on the system because it is our test system and I was the only one
on it at the time. This was on Solaris 8 running Universe 10.1.12 in
Reality flavor.
---
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/
---
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/
---
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/
