RE: access path

["jt" <jt@xxxxxx>]

I probably can't answer at the level of detail You might be looking for,
Leif, ole buddy.  (Looooong time and been meaning to call...  But finalizing
dissolution and needing to find place to live and all.. well.. hope Ya
understand.  Mebbe next week or so.)  Been a long time since I dumped one
with DMPSYSOBJ, but I believe at MI level it's an object type of index and
several other associated system-objects that make up a Logical File.

Conceptually it's an index, anyway, with the key fields and (I assume) a
pointer to a record in a Physical File.  The access path contains the
pointers to records that satisfy the Select/Omit criteria.  If keyfields and
SLTOMT and based-on PF are all the same, the system will implicitly share
that ACCPTH without the coder/DBA having to do anything a-tall.  (To
simplify, same principles that apply to PF data members also apply to LF
DTAMBRs.)

A LF has to be rebuilt for 3 or 4 reasons that I know of:

1) It's been created or changed to ACCPTH(*REBLD), to save the overhead of
adding index entries on each write (or update of key/SLTOMT fields)
2) It's been created or changed ACCPTH(*DLY), and the number of added
records exceeds limit (10%?)
3) The LF has been damaged (although I don't recall actually seeing this in
20-some years)
4) The based-on physical file has been restored, without the corresponding
logical being restored (which is always the case when the PF and LF are not
in the same library)

There may be other reasons, 'course.  Regarding point #3, I once had a VP
who got hired in over me to head department, who was enthralled with
Unix/NT.  He once asked me how often the 400 lost a data base due to
"corrupted pointers"...  When I said "never", He paused fer a few secs and
said "Hmmm... ... never?" so I repeated my answer.  In that same *nix shop
now, (with new VP btw), restoring a corrupted database is part of the
pre-testing for any Oracle DBA looking for a job.

Hmmm... ...  (This isn't a slam on *nix, btw, because they're using old *nix
on old non-IBM hardware, AFAIK, which may be a factor... ? )




| -----Original Message-----
| [mailto:midrange-l-bounces@xxxxxxxxxxxx]On Behalf Of Leif Svalgaard

| I have looked far and wide, but I'm still baffled: what the
| h*** is an "access path"? where is it? what does it do?
| Why must it be "rebuilt" at times and what does that entail?