Help with creating a complex view

Question

Greetings SQL Guru's
We are improving one of our tables for a future release, but we
must support the legacy look and feel of the table because we
have 100's of legacy apps that access the data in the table. The
new system is being deployed with Oracle 10g V10.1.
Our Old REMARKS table that must be turned into a view feed from the new
N_REMARKS table.
UIC     varchar2(6)    PK
RLT     char(3)        PK
LABEL   varchar2(8)    PK
SEQNO     number(1)      PK
REMARK  varchar2(2000)     -- Size is From Oracle 7.3.2 which we stil have in prodcution
New Table definition for N_REMARKS (N for new)
UIC           varchar2(6) PK
RLT           varchar2(3) PK
LABEL         varchar2(8) PK
REMARK_TYPE   char(1)     PK
REMARK        CLOB
[pre]
The remark coming in can be up to 5K long. The SEQNO in the org,
definition allows us to break it up into segments. The REMARK data
itself has three remark types (M-mandatory, O-Optional, G-generated)
that are basically concatenated together and then parsed out everytime
they are processed, not very efficient. The need for a remark is based
on other data in the database so as it relates to the REMARKS table each
type may or may not be there.
Our new structure basically stores the individual remark type in their own
row. Our tool has logic to ensure that the aggregated length does not go
over 5K.
I need a view that can aggregate the remarks together and split them in
2K chunks while incrementing the SEQNO.
Say I have remarks in the N_REMARKS table (For simplicity lets say the
REMARKS field length is 50 characters):
If the N_REMARKS table has
[pre]
UIC     RLT     LABEL    REMARK_TYPE   REMARKS
XXXXX   123     ABCDE      M             MANDATORY REMARK SEGMENT
XXXXX   123     ABCDE    O             OPTIONAL REMARK SEGMENT
XXXXX   123     ABCDE    G             GENERATED REMARK SEGMENT
YYYYY   345     WXYZ     O              SECOND OPTIONAL REMARKS
The New REMARKS view should display
UIC     RLT     LABEL    SEQNO         REMARKS
XXXXX   123     ABCDE      1             MANDATORY REMARK SEGMENTOPTIONAL REMARK SEGMENTGEN        
XXXXX   123     ABCDE    2             ERATED REMARK SEGMENT
YYYYY   345     WXYZ     1              SECOND OPTIONAL REMARKS
[pre]                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   

Answer 1

I had a go, this works for your example (I assumed the remark order is irrelevant, but if not you can change the order by remark_type to order by whatever you want):
  1  select uic, rlt, label, rown-1 AS seq, remark
  2  from (select uic, rlt, label, trim(r) as r, rown
  3        from (select uic,rlt,label,remark_type,to_char(remark) remark,
  4                     row_number() over (partition by uic, rlt, label order by remark_type) as rown
  5              from n_remarks) t
  6        model return updated rows
  7        partition by (uic, rlt, label)
  8        dimension by (rown)
  9        measures (remark r)
10        rules
11          (r [ANY] order by rown desc = r[cv(rown)]||r[cv(rown)+1]))
12  where rown=1
13  model return updated rows
14  partition by (uic, rlt, label)
15  dimension by (rown)
16  measures (r remark)
17  rules upsert iterate(100) until (((iteration_number+1)*50)>length(remark[1]))
18*  (remark[iteration_number+2]=substr(remark[1],1+(iteration_number*50),50))
UIC    RLT LABEL           SEQ REMARK
XXXXX  123 ABCDE             1 GENERATED REMARK SEGMENTMANDATORY REMARK SEGMENTOP
XXXXX  123 ABCDE             2 TIONAL REMARK SEGMENT
YYYYY  345 WXYZ              1 SECOND OPTIONAL REMARKSHowever this may fall over depending how long your remarks are (and I guess being clobs they're pretty long - so it probably will). That being the case I think the best solution is a pipelined function, then you can use PL/SQL to easily do what you want.
hth
Martyn
Message was edited by:
martyn