Sort this !

Question

Hi guys,
Its not a HTMLDB problem but a sql "order by" problem which is causing me some grief, thought would post it up here if anyone has any good ideas :-)
Please take a look at :-
http://htmldb.oracle.com/pls/otn/f?p=26396:1:12152463845138080636
I have a report which is displaying my sort_test table. Please ignore the first column (my_sort) as it is not part of the table (I just used it to display the records in right order for demonstration).
The first 14 rows (my_seq 1 to 14) is how the data currently exists and rows 16-29 is how I want them to be sorted. This data has been downloaded from a fixed length file having record types like 'H' for student header, 'M' for student matching ,'N' for student note, 'C' for student course and 'S' for student subject.
The file should always have records in the order of
H,M,N,C,S i.e first the header, then student matching, then student note, then student course and lastly subjects. A student may have 0-n courses, a student may have 0-n notes, a course may have 0-n subjects.
I am using recid to uniquely identify a student header.Therefore, the first order by clause should be rec_id. The "seq" field is just a surrogate key and means nothing.
Within 'M' the notes ('N') should be sorted by note_seq.
Within 'M' the courses should be ordered by course_cd alpha ascending.
Within 'C' the subjects should be sorted by
1.unit year
2. unit sem start dt
3. unit cd
Has anyone got any ideas on how to achieve this "messy" sort ?
thanks

Answer 1

I guess, I will have to tell you the whole story then.
What I am actually trying to do is to read a fixed length file produced by a third party tool and randomly add subjects/notes to it because it is missing some of them.
The file would look like this :-
H3020THIS IS THE FILE HEADER 20050503212820Q2000892 259619930Y001
MSCB97513769 PETER SMITH 19790123N0002
N100 ThIS is note one
N200 ThIS is note two
N300 ThIS is note three
C3020RBANT Bachelor of Arts (Nature Tourism) 08070101101NYN20000200510350Y026
SBBO1AE WORLD ENVIRONMENTS 20001000123C 301090001
C3020RBAS Bachelor of Applied Science 01000001101NYN19971199920300N024
SBBI110 BIOL110 CELL & ORGANISM 19971000125N 201090001
SBCH110 CHEM110 CHEMISTRY 1A 19971000125D 301050001
H3020THIS IS THE FILE HEADER 20050503212820Q2000895 259619930Y001
MSCB2527237 MICHAEL WILSON 19810807N0001
C3020ABA Bachelor of Arts 09000001101NYN20021200420191N011
SENG11TOT TEXTS OVER TIME 20021000125B 309152301
SLIN11NLA THE NATURE OF LANGUAGE A 20021000125D 309152101
I still have not started coding for it and am currently designing a solution for it. I am thinking on the lines of something like this :-
1. Read the entire file into a temp table with just 2 columns - recnum and text. recnum is a sequence number assigned when read from utl_file and text is the whole line string from the file.
2. Loop through all the records in this temp table (order by recnum) and insert them into a staging table, which would look like the sort_test you have seen before.
Therefore, if I read line 'M' then I just extract the student id from the line and put into the student_id field for REC_TYPE = 'M'. for 'C' ,I extract the course_cd, for 'S' I extract unit_code etc.
3. Then for a student id + course code I will try to find if the file is missing any units and notes, if yes, then insert them with the current recid and so on.
4. Lastly, I just need to sort this staging table and then write them to a new file.
Now, to answer your question..to be very honest I didn't even think that far :-), obviously I need to look at my design again. May be when I read a 'S' line and insert the unit_code etc into the file,I also need to insert the student_id and course_cd as well in the same line that I know that this unit belongs to this course code.that way I would be able to sort by course_cd,unit_cd etc ..
Do you think what I have designed above is going to work ? Can you suggest any better solution here ?
thanks