Find parent/child relationships At More Than 2 Levels

Question

Hello,
Does anyone have a solution to find parent/child relationship for data more than 2 levels deep?
I have a solution when there's a simple parent-child relationship but not when there's a grandparent-parent-child relationship or deeper.
Ex. I have a table company_parent_child that stores the relationship betwen a company and it's direct parent.
create table TEMP_COMPANY_PARENT_CHILD
PARENT_ID NUMBER(10),
COMPANY_ID NUMBER(10)
insert into TEMP_COMPANY_PARENT_CHILD values (1, 10);
insert into TEMP_COMPANY_PARENT_CHILD values (1, 11);
insert into TEMP_COMPANY_PARENT_CHILD values (1, 12);
insert into TEMP_COMPANY_PARENT_CHILD values (2, 13);
insert into TEMP_COMPANY_PARENT_CHILD values (10, 100);
insert into TEMP_COMPANY_PARENT_CHILD values (10, 101);
insert into TEMP_COMPANY_PARENT_CHILD values (10, 102);
insert into TEMP_COMPANY_PARENT_CHILD values (11, 103);
1->
___10->
______100,101,102,
___11->103
Companies 100, 101 and 102 are under parent 10 and grandparent 1.
I need to create such a view or another temp table so that when I pass the parent ID, I will pull all the children on all levels. In addition, and this is the tricky part, when I join this new temp table or view to another data table without any parameters, the data should not be duplicate, ie. each company ID should appear only once.
create table TEMP_JOIN
company_id number(10),
order_id varchar2(10)
insert into TEMP_JOIN values (100, 'a');
insert into TEMP_JOIN values (101, 'b');
insert into TEMP_JOIN values (102, 'c');
insert into TEMP_JOIN values (103, 'd');
insert into TEMP_JOIN values (10, 'e');
insert into TEMP_JOIN values (11, 'f');
insert into TEMP_JOIN values (12, 'e');
insert into TEMP_JOIN values (13, 'f');
Thanks.

Answer 1

start by learning CONNECT BY/START WITH. once you've
written a query to read the grandparent-parent-child
relationship, then come back with more questionsYes. we did look heavily into connect by/start with, in fact along with "connect_by_iscycle","connect_by_isleaf","connect_by_root" as well.
Our dilemma is that when a joint is made between those two tables mentioned above TEMP_COMPANY_PARENT_CHILD and TEMP_JOIN, we are not able to create a view that would contain distinct company_ids, each mapped to a unique order id.
The problem is we cannot have this type of joint when there are "n" level relationship between companies (or company_id). Basically, I think we should have our unique order id mapped to a unique key. This unique key should be a specialized key that we can know at anytime the entire path of the ancestry which we can know by sys_connect_by_path(company_id,'/') path.
How do we know which path to take. The best bet is to "connect_by_isleaf" and just have the distinct "deep" path which form the specialized unique key. If you need help on this let me know. (A hint, sort by LEVEL and then do a rank after partitioning by company id and then filter the records by rank = 1, try this one!!!)
So, we will eventually have a joint (say Table X) like
PATH ORDER_ID
/1/10/100 a
/1/10/101 b
/1/10/102 c
/1/10 e
/1
/1/11/103 d
/1/11 f
I think this is the best view we can have to maintain a joint with no repetition along PATH as well as ORDER_ID. If you have any other thoughts, let me know.
Then you query by path using INSTR to pull records by company_id.
for example, if you want to get all the children for company_id "10" you would just say
select * from X where INSTR(PATH,10,1,1) <> 0
or if you want to get all the children for company_id "11" you would just say
select * from X where INSTR(PATH,11,1,1) <> 0
What do you think? Has anyone used the path information for traversing the tree? Or is there any article that tells us how to make effective use of sys_connect_by_path(company_id,'/') path.
Thank you. Hope it made sense!