On Feb 8, 10:09*am, "John Peterson" <j0... (AT) comcast (DOT) net> wrote:
http://forums.oracle.com/forums/thre...sageID=3575293

Although I personally would have expected the
user_constraints.deferred column to reflect that. Is that a bug or
documentation insufficiency?

Give some ddl/dml so we can all be sure to be on the same page.

jg
--
@home.com is bogus.
http://ostatic.com/blog/oracle-cuts-...ssibility-work

John Peterson (j0hnp (AT) comcast (DOT) net) wrote:
: Hello!

: First time poster in this forum -- please forgive me if this is the wrong
: place for my question. I've exhausted a web search on this issue, and was
: hoping that this might be a more targeted approach.

: I am trying to programmatically determine the current session's constraint
: state (immediate, deferred, or default).

: I have a procedure that I'd like to implement which would essentially
: temporarily set the constraints "deferred" (e.g. SET CONSTRAINTS ALL
: DEFERRED), perform some work, and then restore the constraint setting to the
: original state (e.g., immediate). However, I'm having a difficult time
: identifying how to determine the current state of the constraints.

: Any help would be very much appreciated!

Assuming that the developer always use the ALL keyword when the state is
set/unset, you could try using a flag table with a deferable unique
constraint and then write a non-unique value into that table. An
exception shows that constraints are not deferred (and handling the
exception prevents it from rolling anything else back so the test is safe
to use). If there's no exception then delete the value from the flag
table and continue as normal.

If the purpose is to set and restore the original setting within a single
application then perhaps it would be easier and just as useful for an
application to simply use a utility procedure that uses its own session
variable to track the state and restore it after an equal number of sets
and unsets.


$0.10

"joel garry" <joel-garry (AT) home (DOT) com> wrote

AHA! That's *exactly* what I'm looking for! Thank you!

Can I create a view based off of this x$ view, that I can then grant to a
"standard" user of the system? I'm guessing so...but I guess I'll play
around. :-)

"John Peterson" <j0hnp (AT) comcast (DOT) net> wrote

Well, nuts. Upon further review, this isn't *quite* what I'm looking for
(it's close!).

As it turns out, this works when the ALTER SESSION syntax is used.

However, it *doesn't* work when the SET CONSTRAINTS syntax is used. <sigh>

Ideally I would be able to determine *both* aspects. I wonder if there's an
x$ view that deals with transactions? When I review the list here:

http://yong321.freeshell.org/computer/x$table.html

I don't see anything that's immediately obvious.

Why does this have to be so hard!

"Malcolm Dew-Jones" <yf110 (AT) vtn1 (DOT) victoria.tc.ca> wrote

Thanks, Malcolm! Yeah -- I toyed with the idea of having a utility
procedure -- but we have application code that can't be easily retrofitted
to call out the new utility procedure. Which is why I was hoping to be able
to make the determination via the RDBMS metadata.

But, as I'm discovering, that's all but impossible. :-(

On Feb 9, 4:31*pm, "John Peterson" <j0... (AT) comcast (DOT) net> wrote:
Well, that's probably because the difference isn't trivial, the
session can go over many transactions, but the set constraints is just
a transaction.

Presumably, your code knows when it is in a transaction, rather than
having to figure it out. I don't deal with this, because the language
I use has a built-in to do it - so I know it is possible, but no clue
how, and it quite possibly is just tracking it with its own
variables. But I have to deal with loops unwinding levels of pseudo-
transactions until a rollback will actually work, so it's always
something.

jg
--
@home.com is bogus.
http://latimesblogs.latimes.com/.a/6...a80b970c-800wi