I've been lurking, following this thread, and I guess I have to stick my 2 cents in. I've worked with MV for 33 years, from Reality to jBASE, and one of the best features ever added was secondary indexes. Choices depend on the platform, the iron involved, size and structure of the database and how often you need a particular select, but this is an approach I have used with success. I like to use an index to cull the file down as much as possible, then use that select to drive the next index select, etc. (if the platform allows it and it's appropriate). Then, when the final list is culled down to the records I need, I use another index select to put the list in the sequence I want. The key can be constructed of multiple fields to get just the desired sequence. No brute force required. One advantage of multi-field keys is small nodes, which means more efficient index updating and selects. Now, you can't go crazy and index everything in sight, but you can design indexes to give you the maximum bang for your buck. At least, this approach has worked well for me. Like it, hate it, throw rocks at it, it's all OK by me.

BTW, it's Sunday after Christmas - what are we doing thinking about work today?

Merry Christmas and Happy New Year to all,

Charlie Noah
Charles W. Noah Associates
cwn...@comcast.net

The views and opinions expressed herein are my own (Charlie Noah) and do not necessarily reflect the views, positions or policies of any of my former, current or future employers, employees, clients, friends, enemies or anyone else who might take exception to them.


On 12-26-2010 6:50 PM, Robert Houben wrote:
I should also clarify that we have a lot of customers we support on mvBase, 
mvEnterprise, D3 and lots of other platforms, and yes, we do have PI Open 
customers...  I recognize that the topic here is U2, but even there, we have 
customers who use our products running on Universe 5 (don't ask, we just do...) 
 So, I find that I often have to assume a lowest common denominator when 
building a reusable solution.  I may not always be aware of the current state 
of a particular platform.  I'd love to be corrected if my understanding of 
limitations is out-of-date!  Last I knew, if you wanted to sort an MV file by 
more than one field, regardless of how many indexes you had, you got to pick 
one of them, and you would settle for brute force for the others.  This was 
true, last I knew, of EVERY MV platform I knew of that had indexes.

-----Original Message-----
From: u2-users-boun...@listserver.u2ug.org 
[mailto:u2-users-boun...@listserver.u2ug.org] On Behalf Of Robert Houben
Sent: Sunday, December 26, 2010 4:42 PM
To: U2 Users List
Subject: Re: [U2] Migration

Should have clarified "when you sort *multiple* fields that are indexed".  I 
still haven't heard anyone tell me that either UV or UD now support more than one indexed 
field.  Let me know if this has changed...

-----Original Message-----
From: u2-users-boun...@listserver.u2ug.org 
[mailto:u2-users-boun...@listserver.u2ug.org] On Behalf Of Wols Lists
Sent: Sunday, December 26, 2010 4:33 PM
To: u2-users@listserver.u2ug.org
Subject: Re: [U2] Migration

On 24/12/10 15:50, Robert Houben wrote:
SQL will beat MV every time when you sort fields that are indexed.
Huh? Ime (UniVerse), that's wrong.

Indexes are b-trees, which you can walk, and the contents of the index are 
sorted. afaik you would have been right about PI, but that's long dead. Dunno 
about UniData, but UV is a lot of sites where MV will equal SQL ... :-)

  For direct reads, MV seems to have a slight advantage.  Inserts and
updates that affect indexed fields are slower in SQL (inserts are
painfully slow if you fail to size your SQL table well, but try
"inserting" millions of records into a file with a modulo of 1...)
Been there, done that. But that's why most places use dynamic files nowadays. 
:-)

Cheers,
Wol
_______________________________________________
U2-Users mailing list
U2-Users@listserver.u2ug.org
http://listserver.u2ug.org/mailman/listinfo/u2-users
_______________________________________________
U2-Users mailing list
U2-Users@listserver.u2ug.org
http://listserver.u2ug.org/mailman/listinfo/u2-users
_______________________________________________
U2-Users mailing list
U2-Users@listserver.u2ug.org
http://listserver.u2ug.org/mailman/listinfo/u2-users

_______________________________________________
U2-Users mailing list
U2-Users@listserver.u2ug.org
http://listserver.u2ug.org/mailman/listinfo/u2-users

Reply via email to