Re: Do I need DDS here?

[CRPence <CRPbottle@xxxxxxxxx>]

  Beyond optional support being added to CREATE INDEX in V6R1 for both 
expressions to define derived key and a WHERE clause to define key 
selection, V6R1 adds the ability to name the columns that will define 
the record format of the INDEX.  This is done with the RCDFMT clause 
using ADD column-name; albeit not robust for layout, since the columns 
named as key fields are always first:
http://publib.boulder.ibm.com/infocenter/systems/scope/i5os/topic/db2/rbafzxcindx.htm

  Thus the ability exists to create much the same as a DDS keyed-LF, 
but with SQL.  Some non-join VIEW *FILE objects will be able to be 
replaced by a more complex INDEX *FILE object for use in RLA.  A SQL 
INDEX is still only over a single TABLE [syntax: 
>--ON--table-name--...], which defines an Access Path over the _data_ 
in the dbfile.mbr.  An INDEX can not be created on a VIEW, nor has 
support been added for an ORDER BY clause [or similar] to ensure an 
implied order for a non-SQL program opening the file.  The same /stupid/ 
limitation will exist.

  I am not aware if non-key selection is dynamic as in DYNSLT [I think 
probably], or if the selection is always part of the access path itself. 
 If not dynamic, the relationship between key-expression and the WHERE 
search-condition is much closer than I would infer from the doc.

Regards, Chuck

rob wrote:
> V6R1 added some ability to create "sparse indexes" to emulate some
> of the last holdouts that DDS could do that DDL couldn't.  And the
> only real use for sparse indexes is for RLA as the optimizer ignores
> them.  There, I think the sparse indexes just add select/omit logic
> to a SQL index.  I wonder if it's supported in a view also?  Watch
> it have the same stupid limitation DDS has that you can only have a
> key on field(s) from one of the underlying tables.