RE: Indices & Packages

[eydeak]

RE: Indices & Packages

Nothing happens. The index is not created until it is generated.



Ellen Deak
Senior Analyst/Programmer
Cooper-Standard Automotive
OneWorld Xe (B7333, Service Pack 22)
AS/400 DB2

 

[scottw]

Re: Indices & Packages

Good Afternoon, First, just add the index and generate the indexes. This actiondoes not remove the data.

Secoond on an upgrade you need to remember the index.

Third, copy of the F4211 (F554211) through OMW, copy the data from the F4211 into the copy. Add an index to the F554211, generate the index and test to see if there are any bad side effects. The later might the first thing you do if you do not trust what you are being told to do.

--
Senior Technical Consultant
17611 E. Tennessee Pl.
Aurora, CO 80017
Work 303-337-1906
Cell 303-885-1405
Fax 303-306-1800

 

[DBohner-(db)]

RE: Indices & Packages

Hey, I just thought of 'what happens' if the index is not generated...

Haven't tried it yet - but here is a scenario...

Internal to you application/ube, you fetch based on an index that isn't
there. An SQL statement is initiated and you get data back... BUT - there
wasn't an Index. Either the fetch did a table scan or the database was
smart enough to use an already existing index. Table Scans can be
INTENSE....

I'd create the index.

Daniel Bohner
ExistingLight, LLC
www.existinglight.net
[email protected]
208.250.1917





Looking for work out west
Daniel Bohner | JDE Developer | Freelance-JDE/Web/Photo
www.ExistingLight.Net | [email protected]
JDE XE | SP20 | AS/400 | V5R1

 

[wainwrr]

Re: RE: Indices & Packages

I wouldn't want to speak to any other DB but with ORACLE - just because a table exists which matches your selection criteria, it won't necessarily be used. Usage depends on the type of optimisation you have in place and the statistics of the table. Sometimes a full table scan will be faster/more efficient than using an index.


<Start crude approximation>
Consider getting multiple rows (because of wild cards in the selection say) with an index that approaches being a unique key. You'd end up reading every block of the index and then every block of the data. Actually doing a full table scan you just read all the blocks of the data - less work
<End crude approximation>