Thursday, May 26, 2005
The documentation.
Ever read the contents of $ORACLE_HOME/rdbms/admin?
Loads of good stuff in there. Right now I'm evaluating dbms_metadata - and sadly coming to the conclusion that you don't want to use it to dig out schema ddl.
So I've logged at least one tar ORA-31608 when extracting grant information doesn't mean I errored out, it probably means no grant information to return.
In the mean time Ive dug into catmeta.sql and dbmsmeta.sql lots of interesting stuff in there. Interesting in the sense of Oh thats what that view does and interesting in the sense of oops had to withdraw that bit while the XDK was fixed. Anyway given a bit of R & D time read some of the header files in $ORACLE_HOME/rdbms/admin It'll be more interesting and understandable than you think.
I think I'll leave the last word to catmeta.sql
-- view to get a simple set of column attributes.
-- NOTE: Originally we had just an attribute called 'name' that was either
-- c.name or attrcol$.name selected via a DECODE of c.property bit 1.
-- However, for an as yet unexplained reason, this causes a full table scan
-- on col$ in outer views that used this view.
don't you just want to know why?
1 Comments
Loads of good stuff in there. Right now I'm evaluating dbms_metadata - and sadly coming to the conclusion that you don't want to use it to dig out schema ddl.
So I've logged at least one tar ORA-31608 when extracting grant information doesn't mean I errored out, it probably means no grant information to return.
In the mean time Ive dug into catmeta.sql and dbmsmeta.sql lots of interesting stuff in there. Interesting in the sense of Oh thats what that view does and interesting in the sense of oops had to withdraw that bit while the XDK was fixed. Anyway given a bit of R & D time read some of the header files in $ORACLE_HOME/rdbms/admin It'll be more interesting and understandable than you think.
I think I'll leave the last word to catmeta.sql
-- view to get a simple set of column attributes.
-- NOTE: Originally we had just an attribute called 'name' that was either
-- c.name or attrcol$.name selected via a DECODE of c.property bit 1.
-- However, for an as yet unexplained reason, this causes a full table scan
-- on col$ in outer views that used this view.
don't you just want to know why?
1 Comments:
"...for an as yet unexplained reason..." That really doesn't leave a warm and fuzzy feeling.
That reminds me of a time when I commented a piece of production code that said, "I really don't understand this code. It seems to be working, so please do not touch it." This was really early in my career. The project lead later found my comments and gave me a sturn lecture.
Post a Comment
That reminds me of a time when I commented a piece of production code that said, "I really don't understand this code. It seems to be working, so please do not touch it." This was really early in my career. The project lead later found my comments and gave me a sturn lecture.
