I decided to try attacking another part of this project, and opened a PF
with my JDBCR4 test-bed, without bothering to take out the code
activating extended metadata.
And I GOT NONBLANK VALUES FOR SCHEMA AND TABLE NAMES.
I tried it again with a simple logical.
I STILL GOT NONBLANK VALUES FOR SCHEMA AND TABLE NAMES.
So then I went back to the join-logical I'd been testing with. Still
blank on both for that.
After sending the above to Scott Klement, (I thought I was sending it to
the List), I got back a response, that DDS join-logicals were "a weird
peculiarity of the DDS/native record access world" that "do not exist in
the SQL world."
That seemed rather odd, so I tried it with a join that had been
generated from SQL similar to:
CREATE VIEW WTADDRV1 (RAETID, RADTYP, ADID, ADCNTR,
ADCITY, ADCITSORT, ADDR1, ADDR2, ADDR3, ADDRSORT, ADPROV, ADPROVSORT,
ADPCDE, ADCTY, ADCTYSORT, ADCTYNUM, ADUSTATE, ADSTATNUM,
ADUZP5, ADUZP4)
AS SELECT R.RAETID, R.RADTYP,
A.ADID, A.ADCNTR,
A.ADCITY, A.ADCITS, A.ADDR1, A.ADDR2, A.ADDR3, A.ADDRS, A.ADPROV, A.ADPROS,
A.ADPCDE, A.ADCTY, A.ADCTYS, A.ADCTYN, A.ADUSTA,
A.ADSTATN, A.ADUZP5, A.ADUZP4
FROM WTADDR A LEFT JOIN WTRADR R ON A.ADID = R.RADID
The real file has a few additional fields, most of which were calculated
using UDFs. And when I opened it in my JDBCR4 test-bed,
ALL THE FIELDS HAD CORRECT NONBLANK SCHEMA AND TABLE NAMES, EXCEPT FOR
THE UDF-CALCULATED ONES THAT, NOT SURPRISINGLY, HAD BLANK VALUES.
I don't doubt Scott's assertion, given his far greater experience with
such things (not to mention that I'd just confirmed it empirically), and
using JDBC to access a native join-logical on the same box amounts to
"carrying coals to Newcastle," but that doesn't mean I *understand* it.
Can anybody explain *why* it is that a native, DDS-generated,
join-logical is somehow fundamentally different from an SQL join-view? I
would have thought that in terms of database internals, making them
fundamentally different would amount to "reinventing the wheel."
--
JHHL
midrange.com
