Walden H. Leverich wrote:
CRPence wrote:
IMO the most obvious reason, is to enable ISAM versus SQL access
to the data via...
OK, but then create an LF over the table. Why take a SQL concept
(view) and try to use it in HLL-IO? If you want a LF create an LF,
complete w/the necessary DDS. I don't believe there's a restriction
to creating an LF over a PF that is a SQL table, is there?
A DDS LF, simply put, does not support the many desirable features
and mappings that are available via an SQL SELECT. A DDS LF does not
support any of, for example, HEX(), character to numeric, CASE, GROUP
BY, AVG, date arithmetic, scalar UDF, et al.
FWiW, given my choice, an ORDER BY in a VIEW would have more
intuitively satisfied the concept; i.e. rather than an INDEX that is
over a VIEW. Regardless, my points are both that ordering the result of
a VIEW may be desirable for ISAM and that OPNQRYF of a VIEW may be able
to effect that in many cases, to accomplish what is desired.
The better question I think is: Why *not* take advantage of the SQL
concept (VIEW) and use it to remove logic from the HLL, when it can be
replaced easily by what the database already provides? Since CLP is
very limited in its file I/O, it can obviously benefit significantly
more from being able to push the logic down to the database than a
language like RPG. However at the same time, there is no reason RPG
should not be able to benefit similarly.
Oddly, I seem to have failed to paste additional text I had composed
separately, that mentioned that DB2 for i [with IBM i 6.1] offers some
new function for CREATE INDEX. I will make a separate reply to the OP.
That new function enables an SQL INDEX to be created with some mapping
[expressions; e.g. HEX(), etc.] and selection [WHERE clause], very
similar to the DDS LF derived [keyed] [select/omit] indexes. Note: the
capability of the optimizer to use these indexes [until the next
release] may actually be more limited than for what capabilities can be
achieved using them with RLA. Refer to the revised SQL syntax:
http://publib.boulder.ibm.com/infocenter/systems/topic/db2/rbafzxcindx.htm
Regards, Chuck
midrange.com
