At this point, Oracle isn’t much help and things are getting worse, so I’ve decided if index maintenance is the issue, there’s gotta be a better way to store the index information.  Maybe partitioning the table with a hash partition and locally partitioned indexes… then the answer comes to me in the form of Oracle’s own documentation.

Globally partitioned indexes. Just like partitioned tables, a database can have partitioned indexes, local OR global.  Locally partitioned indexes require a partitioned table and, from what I can tell, the same partitioning parameters, whereas globally partitioned indexes can be on anything, even when on a partitioned table…. sounds messy.

Next up… Oracle Partitioned Indexes - Part 2 (choosing the partition type)

I’m not a DBA, and even though I pretend sometimes, I can’t even come close to pretending on this one. We have a primary database and a secondary database (physical standby) using DataGuard. With FAN support, we can actually fail over to the secondary database, and allow the OCI8 driver to avoid the giant timeout normally found in TCP, reconnect and pick up where it left off. I’m super excited to see this working… has anyone tried it yet?

CREATE TABLE temp_1 (
user_id NUMBER(3)
);
INSERT INTO temp_1
(user_id)
SELECT user_id
FROM table1;

SELECT user_name, user_info
FROM users, temp_1
WHERE users.user_id=temp_1.user_id;

And here’s using the WITH clause:

WITH temp_1 AS (
SELECT user_id
FROM table1
)
SELECT user_name, user_info
FROM users, temp_1
WHERE users.user_id = temp_1.user_id

Oracle creates a temporary table internally and uses it in the query. You might be thinking, “not a big deal, I do things like this all the time.” Here’s the kicker: There’s NO redo generated. For those of use with physical and/or logical standby’s, we NEED this kind of functionality to keep redo down on the backup databases. YAY!

One of the primary concerns with the nested set model is the requirement that N number of rows would have to be updated with each new item. This is a deal-breaker in my job and wouldn’t fly, so I’ve added an additional column to assist in the solution that identifies the level at which the item occurs.

Given the following data set:

Here’s the query I came up with:

SELECT max(ltrim(sys_connect_by_path(site_area, ' > '),' > '))
FROM (
SELECT
site_area,
site_link,
row_number() over (partition by 1 order by site_level) level_index
FROM hierarchical_data
START WITH id=8
CONNECT BY PRIOR parent_id=id
)
START WITH level_index = 1
CONNECT BY level_index = prior level_index + 1;

And the resulting data set:

Let me know what you think… and be nice.