Thank you for the information, Wally! :) Again, though, I'm on the
hunt -- I like having references for my forgetful mind. Is there a
piece of UniData documentation or online help that describes the
ONLINE keyword in relation to BUILD.INDEX? It doesn't seem to show
when I do a HELP BUILD.INDEX.
On 2/7/07, Wally Terhune <[EMAIL PROTECTED]> wrote:
I don't agree with Shaun's analysis of UniData and indexes - though perhaps
we are looking at it slightly differently.
1) CREATE.INDEX just creates the index file and puts a flag in the source
file header indicating that indexes exist - and how many.
The details about the specific fields and their 'formula' (D type =
attribute, I type = expression) are stored in the header of the index file
by the CREATE.INDEX command.
There are no locking or file open issues regarding CREATE.INDEX that I am
aware of. You can run LIST.INDEX before you BUILD.INDEX and confirm that
you got what you asked for.
2) BUILD.INDEX (prior to UniData release 7.1) requires exclusive access to
the primary file during the build process. This is to ensure that when the
build is done, the contents of the index are 100% consistent with the
contents of the primary file. This is done with UniData internal locks
(exclusive lock on the entire file for updates - the file can still be
listed during BUILD.INDEX, but not updated).
For Recoverable files (RFS), no udt process can have the file OPEN. If
BUILD.INDEX cannot get an exclusive lock (or for RFS files - other
processes have it open), BUILD.INDEX will fail - producing an error
message. LIST.INDEX can be used to confirm the 'built' status of your
indexes.
3) At release 7.1.0, we added the ONLINE keyword to the BUILD.INDEX
command. This changes the underlying build operations considerably, but
allows you to continue to write records to the primary file during
BUILD.INDEX. When you use the ONLINE keyword, the build is significantly
slower - but your 7/24 application use can continue unimpeded.
Wally Terhune
U2 Support Architect
IBM Information Management
4700 South Syracuse Street, Denver, CO 80237
Tel: 303.773.7969
Fax: 303.773.5915
[EMAIL PROTECTED]
<[EMAIL PROTECTED]
erguson.com>
Sent by: To
[EMAIL PROTECTED] <u2-users@listserver.u2ug.org>
stserver.u2ug.org cc
Subject
02/07/2007 11:19 [U2] Index Rebuilding
AM
Please respond to
[EMAIL PROTECTED]
er.u2ug.org
I'm on digest, so sorry for the delay. Indexes on UniVerse and UniData
are different. UniVerse will stop you from creating them if the file is
open by any process
(very good!). UniData will go on it's merry way and create the index
even if the file is open by another process. You run the BUILD.INDEX
command and all looks
fine. Then you add more records to the file and the index appears
broken. Your new records are not accounted for. On UniData, you must
be absolutely sure
that the file is not open when you create the index. On unix, I use
"fuser -u [path to file]". I do not know the equivalent on Window$.
Indexes have become very
stable and accurate if created properly. As mentioned earlier, best not
to use "I" or "V" dictionary items, if possible (especially TRANS or
file OCONV's), and
better to use the key and null in a boolean expression (using NO.NULLS
at creation) instead of a "1" and "0". Just my 2 cents...
Shaun
-------
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/
[demime 1.01d removed an attachment of type image/gif which had a name of
graycol.gif]
[demime 1.01d removed an attachment of type image/gif which had a name of
pic23157.gif]
[demime 1.01d removed an attachment of type image/gif which had a name of
ecblank.gif]
-------
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/
